Research ArticleSAI Safety Application Identifier Algorithm atMAC Layer for Vehicular Safety Message Dissemination OverLTE VANET Networks

Shuja Ansari 1 Marvin Saacutenchez1 Tuleen Boutaleb 1 Sinan Sinanovic 1

Carlos Gamio1 and Ioannis Krikidis2

1School of Engineering and Built Environment Glasgow Caledonian University Glasgow UK2Department of Electrical and Computer Engineering Faculty of Engineering University of Cyprus Nicosia Cyprus

Correspondence should be addressed to Shuja Ansari shujaansarigcuacuk

Received 25 August 2017 Revised 4 January 2018 Accepted 30 January 2018 Published 27 February 2018

Academic Editor Orhan Gazi

Copyright copy 2018 Shuja Ansari et al This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

Vehicular safety applications have much significance in preventing road accidents and fatalities Among others cellular networkshave been under investigation for the procurement of these applications subject to stringent requirements for latency transmissionparameters and successful delivery ofmessages Earlier contributions have studied utilization of Long-TermEvolution (LTE) undersingle cell Friis radio or simplified higher layer In this paper we study the utilization of LTE under multicell and multipath fadingenvironment and introduce the use of adaptive awareness range Then we propose an algorithm that uses the concept of qualityof service (QoS) class identifiers (QCIs) along with dynamic adaptive awareness range Furthermore we investigate the impact ofbackground traffic on the proposed algorithm Finally we utilize medium access control (MAC) layer elements in order to fulfillvehicular application requirements through extensive system-level simulationsThe results show that by using an awareness rangeof up to 250m the LTE system is capable of fulfilling the safety application requirements for up to 10 beaconss with 150 vehiclesin an area of 2 times 2 km2 The urban vehicular radio environment has a significant impact and decreases the probability for end-to-end delay to be le100ms from 93ndash97 to 76ndash78 compared to the Friis radio environment The proposed algorithm reducesthe amount of vehicular application traffic from 21Mbps to 13Mbps while improving the probability of end-to-end delay beingle100ms by 20 Lastly use ofMAC layer control elements brings the processing ofmessages towards the edge of network increasingcapacity of the system by about 50

1 Introduction

Cellular Networks are ubiquitous technologies that haveevolved in order to satisfy the continuous increasing trafficand quality of service (QoS) demands required by existingand future user applications Operators currently utilize oneor several standardized technologies as the application mar-ketrsquos demand increases For instance voice and data applica-tions requiring low latency can be served by wideband codedivision multiple access (WCDMA) while higher data rateswith lower latency can be provided by introducing evolveduniversalmobile telecommunications service terrestrial radioaccess networks (E-UTRAN commercially referred to asLong-Term Evolution (LTE)) [1]

Vehicular communications presiding under the umbrellaof cooperative intelligent transportation systems (C-ITS)sets forth stringent requirements in terms of latency andsuccessful packet delivery There are mainly three applica-tions in vehicular networks as categorized by [2] safety(eg cooperative forward collision warning to avoid rear-end collisions) transport efficiency (eg traffic light opti-mal speed advisory to assist the driver to arrive during agreen phase) and informationentertainment (eg includingremote wireless diagnosis to make the state of the vehicleaccessible for remote diagnosis) Safety applications alsoreferred to as driver safety applications are envisioned to helpprevent fatalities on the road by sending periodic messages(beacons) in the vehicles neighborhood These applications

HindawiWireless Communications and Mobile ComputingVolume 2018 Article ID 6576287 17 pageshttpsdoiorg10115520186576287

2 Wireless Communications and Mobile Computing

Table 1 Safety application requirements

Safety application Message type Communication mode Minimum frequency Criticallatency

Transmissionrange

Intersection collisionwarning CAM Broadcasting periodic

messages 10Hz sim100ms 150m

Lane change assistance CAM Cooperation awarenessbetween vehicles 10Hz sim100ms 150m

Slow vehicle indication CAMDENM Broadcasting state periodically 2Hz sim100ms 200mTraffic light speedadvisoryviolation CAMDENM Broadcasting periodic

messages 2Hz sim100ms 150m

Overtaking vehicle warning CAM Broadcasting overtaking state 10Hz sim100ms 300m

Head on collision warning CAM Broadcasting periodicmessages 10Hz sim100ms 200m

Collision risk warning CAMDENM Time limited periodicmessages on event 10Hz sim100ms 300ndash500m

Cooperative forwardcollision warning CAM Cooperation awareness

between vehicles 10Hz sim100ms 150m

Emergency vehicle warning CAMDENM Broadcasting periodicmessages 10Hz sim100ms 300m

Cooperative mergingassistance CAM Cooperation awareness

between vehicles 1Hz sim1000ms 250m

Speed limits notification CAM Broadcasting periodicmessages 1ndash10Hz sim100ms 300m

Motorcycle approachingindication CAM Cooperation awareness

between vehicles 2Hz sim100ms 150m

and technologies have been subject of much research sincethe allocation of the dedicated short-range communication(DSRC) licensed spectrum on 59GHz frequency band [3]In order to utilize the DSRC spectrum equipment mustbe compliant with the IEEE wireless access in vehicularenvironment (WAVE) standards suite that includes IEEE80211p designed for vehicular environment Utilizing thisstandard vehicular ad hoc networks (VANET) can be createdon roads having the advantage of workingwithout the need ofany fixed infrastructure known as vehicle-to-vehicle (V2V)communications or with the installation of road side units(RSUs) the latter can be used for vehicle-to-infrastructure(V2I) communications to extend the type of services andapplications

Provisioning of V2I using IEEE 80211p requires instal-lation of RSUs which means deployment and maintenanceof a dedicated network is required [4 5] This has motivatedseveral authors to propose the utilization of LTE as analternative to provide V2I applications and furthermoreinvestigate if LTE can satisfy the stringent requirements withV2V applications [6 7]

In continuation of the previously contributed work fromvarious authors described in Section 11 contributions of thispaper include the following

(i) Analysis of intercell interference and handoverimpact on latency of the end-to-end safety systemsand studying the impact of vehicular urban radioenvironment employing multipath fading channelsunder European (Glasgow) and American (Manhat-tan) mobility traces

(ii) Investigation of variable awareness range that can beused over LTE to fulfill various vehicular applicationrequirements (see Table 1)

(iii) Investigation of regular cellular trafficrsquos impact onvehicular applications running on LTE networks

(iv) Proposal of safety application identifier (SAI) algo-rithm utilizing dynamic selection of transmissionparameterswithin the impact of the above-mentionedcontributions

(v) Utilization of medium access control (MAC) layercontrol elements with the proposed SAI in orderto reduce latency of the network while increasingcapacity for vehicular application use

In light of these contributions related literature is pre-sented in the following section

11 RelatedWorks When utilizing LTE uplink transmissionsfrom vehicles to a centralized server are commonly doneusing unicast [9] For the downlink forwarding of thereceived information can be delivered to multiple vehiclesusing either unicast broadcast or multicast transmissionsBroadcastmulticast requires transmissions to be received byusers in the worst cell quality conditions (eg at the celledge) and results in lower average spectral efficiency (bpsHz)compared to unicast transmissions [1 Chapter 1] In spite ofthis if the number of receiving users of the message is largerthan the spectral efficiency reduction relative to unicast itis more efficient to use broadcastmulticast transmissions inorder to cater capacity blocks

Wireless Communications and Mobile Computing 3

Broadcast and unicast transmissions with LTE have beeninvestigated in [10ndash12] Vinel [10] used a simplified model inorder to compute analytically the packet delivery ratio (PDR)assuming fixed number of neighbors (50 vehicles) in a timedivision duplex (TDD) LTE single cell system Due to theoverall system complexity of LTE system-level simulationshave become an indispensable tool for predicting the per-formance over more realistic scenarios In [11] simulationswere used to compare IEEE 80211p with frequency divisionduplex (FDD) LTE single cell system performance usingFriis radio propagation environment on a Manhattan girdmobility model Their performance evaluation showed thatLTE is suitable for VANETs however the use of single cellsimplified higher layers and Friis radio environment does nottest the systemunder extreme propagation scenariosWhen itcomes to system-level simulations choice of mobility modelscarries much importance Grzybek et al in [13] studied andcompared various mobility generators for VANETs propos-ing a mobility generator for realistic traces called vehiluxThis generates traces considering the spatial temporal andbehavioral aspects of traffic distribution Since the scope ofthis paper is within the transmission requirements of safetyapplications the trace generator adopted is routes mobilitymodel [14]This generates traces with geographic restrictionson top of spatial temporal and behavioral aspects allowinga fixed number of vehicles in the area throughout thesimulation time

Kato et al [15] defined the concept of data freshnessand used this parameter to evaluate the performance of LTEnetworks with VANETs If a vehicle application transmitsa packet at a frequency of 10Hz that is a message istransmitted every 100ms the freshness requirement for thatspecific application would be 100ms They observed thatit is possible to achieve a freshness of 100ms by havingvehicular application server at the edge of LTE networkhence discarding the message if it arrives later than thebeacon interval time Authors evaluated the impact of serverlocation by calculating the amount of network usage Theyconcluded that by reducing the round trip time (RTT) ofthe system network usage can be reduced to half hence byplacing the server close to the eNodeB or even the LTE core itcan have a significant impact on the scalability of the system

In order to utilize broadcast functionality on LTEoperators must implement multimedia broadcast multicastservices (MBMS) [1 Chapter 14] Without MBMS currentlyonly low rate public warning system (PWS) messages can besent through the broadcast capability of LTE [16 17] Hencein some situations unicast is the only available schemeFor instance the experimental study in [6] used TCPIPping application to measure the end-to-end delay in orderto evaluate the suitability of LTE for VANET Also unicastfor cooperative awareness message (CAM) transmissions inmulticell environment with LTE has been studied in [12] Inthat study it was found that the downlink was the bottleneckdue to high traffic load However the study was oriented todetermine the capacity limit therefore the downlink trafficwas modeled by fixed size periodic packet transmission thehandover effect was neglected and did not include the end-to-end system Previous performance evaluations have also

suggested that the use of LTE for vehicular communications issuitable but without any centralization it can put enormousload on the network [18] In the pursuit of centralizing vehic-ular communications in LTE group formation and MBMSimplementation have been proposed in [18 19] Groupformation or clustering has shown promising performanceHowever clustering relies on relaying transmissions whichcan pose a privacy and security issue [20]

3GPP release-12 [21] includes device-to-device (D2D)communication under LTE-Advanced which can be a poten-tial candidate technology for V2V communications Bazzi[22] investigated the use of LTE direct communications(D2D) using full duplex radios in order to enable vehiclesto receive and transmit at the same time Results from theirdeveloped analytical framework showed promising reductionof LTE uplink resources as compared with using unicastand broadcast However the proposed framework exhibitsminimum interference only when the awareness range is setto 100m In [23] authors have proposed a scheme that utilizesan unused 16-bit long MAC Control Element inside thebuffer status report (BSR) request for the network to identifythe receiver and transmitter This element uses the spacereserved for future use and is indexed in the MAC ProtocolData Unit (PDU) subheader by the logical channel identifier(LCID) By doing so packets sent from the transmitter getto the receiver via eNodeB instead of traversing throughthe evolved packet core (EPC) saving the round trip timefrom the LTE core network (CN) to the access network Asimilar concept has been adopted in our work exploringthe possibility of using MAC control elements for messagedissemination within vehicular networks in order to reducelatency

The remainder of this paper is organized as follows inSection 12 the vehicular safety requirements and awarenessrange are introduced followed by the proposed algorithmin Section 2 the system model used for LTE along withthe performance measures is then included in Section 3finally the simulation results and conclusions are discussedin Sections 4 and 5 respectively

12 Vehicular Safety Applications In vehicular safety applica-tions the consequence of failure in message delivery withina minimum delay and awareness range may result in a fatalaccident [24] In Table 1 we show a survey of vital active roadsafety applications along with their requirements for criticallatency beacon frequency and transmission range [25ndash29]

Overall these applications require a critical latency lessthan or equal to 100ms and they differ in the message typeminimum frequency and transmission range requirementsWithout any broadcast routing algorithms the packet deliv-ery rate for VANETs is generally low ranging between 60and 80 [30] However the standards do not define anacceptable packet delivery rate

Awareness range previously elaborated in [9 22 31]is the geographical area around the vehicle where all theneighbors are to bemade cognizant of the vehicle Dependingon the applicability of the message this range can vary Thereare two types of messages involved in VANETs classifiedas CAMs [26] and Decentralized Environment Notification

4 Wireless Communications and Mobile Computing

Table 2 Safety application requirements and SAIs

Safety application Message type Transmitted data Criticallatency

Transmissionfrequency BF

Transmissionrange 119877 SAI

Intersection collisionwarning CAM Vehicle type position heading velocity

acceleration yaw rate sim100ms 10Hz 150m

1Lane change assistance CAM Position heading velocity accelerationturn signal status sim100ms 10Hz 150m

Cooperative forwardcollision warning CAM Vehicle type position heading velocity

acceleration yaw rate sim100ms 10Hz 150m

Slow vehicle indication CAMDENM Vehicle type position headingacceleration velocity sim100ms 2Hz 200m 2

Traffic light speedadvisoryviolation CAMDENM Signal phase timing position direction

road geometry sim100ms 2Hz 150m3

Motorcycle approachingindication CAM Vehicle type position heading velocity sim100ms 2Hz 150m

Overtaking vehiclewarning CAM Position velocity yaw rate acceleration sim100ms 10Hz 300m 4

Head on collisionwarning CAM Vehicle type position heading velocity

acceleration yaw rate sim100ms 10Hz 200m 5

Collision risk warning CAMDENM Vehicle type position heading velocityacceleration yaw rate sim100ms 10Hz 300ndash500m 6

Emergency vehiclewarning CAMDENM Position heading velocity acceleration sim100ms 2Hz 300m 7

Cooperative mergingassistance CAM Curve location curvature slope speed

limit surface sim1000ms 1Hz 250m 8

Speed limits notification CAM Velocity acceleration position speedlimit heading sim100ms 1ndash10Hz 300m 9

Message (DENMs) [27] CAMs are periodic messages thatprovide information like the presence position and sensorinformation and are expected to be received by all thevehicles within the awareness range DENMs are used forcooperative road hazard warnings that are broadcasted intheir relevance area whenever a hazardous event occurs [32]The work in [33] proposes the use of dynamic adaptation ofcommunication radius in order to maximize sum rate weadopt the same concept and apply it to dynamically adaptingcommunication range in order tominimize end-to-end delayFigure 1 illustrates the awareness range119877 for vehicle 119894 Noticethat vehicle 119896 is within the awareness range of 119894 since thedistance 119889119894119896 le 1198772 Safety Application Identifier Algorithm

Message delivery and reliability are a major concern in vehic-ular networks due to which a differentiated QoS mechanismis proposedThismechanismworks on the principle of index-ing various applications according to their requirementsmotivated from QoS class identifier (QCI) implemented inLTE networks [34] In Table 2 each application fromTable 1 iscategorized and assigned an SAIThis SAI which is a numberranging between 1 and 9 is included in the transmitted packetappended before the IP header by the application layer asshown in Figure 2 The categorization and assignment of SAIfor ourmodel are further discussed in Section 3This conceptis implemented in the form of an algorithm proposed in thefollowing subsection

Input VSMs Vehicles rarr ServerOutput VSMs Server rarr Vehicles

System Setup(1) while Server larr VSM119894 do(2) Server locates all vehicles(3) VSM119894 rarr (SAI119894Position (119889119896))(4) SAI119894 rarr (119877119894BF119894)(5) (119877119894 119889119896) rarr Distance (119889119894119896)(6) F119894 = forall119896 119889119894119896 lt 119877119894 119894 = 119896(7) Server rArr Vehicles isin F119894 at BF119894(8) end while(9) return F119894

Algorithm 1 SAI Algorithm

21 SAI Algorithm With the concept of safety applicationindexing we propose an algorithm that does not includethe use of group formation or IEEE 80211p The proposedalgorithm is a process that is carried out on the vehicularsafety application (VSA) server elaborated in Algorithm 1When the vehicles start transmissions the packet is carriedfrom the eNodeB to the VSA server via the CN This packetcontains vehicle location (119889119896) and corresponding vehicularsafety message (VSM) As soon as the server starts receivingthese messages it locates all the vehicles being served bythe network forming a virtual geographical map of theserved area The VSA server while receiving the packets

Wireless Communications and Mobile Computing 5

S-GWP-GWVSA

Server

Awareness Range

VSM

S-GW serving gatewayP-GW PDN gatewayVSA vehicular safety application

idik

k

R

Position dkSAIi (Ri BFi)

Figure 1 LTE multicell system and awareness range

will also extract the SAI information from the receivedpacket This SAI information will then be checked againstthe database stored in the server (Table 2) With this SAIthe server retrieves the transmission beacon frequency BF119894and awareness range 119877119894 requirement of the application thatvehicle 119894 is utilizing Using the required awareness range theserver determines the forwarding set of vehicles (neighboringvehicles) by

F119894 = forall119896 119889119894119896 lt 119877119894 119894 = 119896 (1)

where 119889119894119896 is the distance from vehicle 119894 to the neighboringvehicle 119896 Recalling the concept of data freshness defined by[15] the VSA server will discard the packet if it arrives laterthan the beacon inter arrival time (1BF119894)22 MAC Layer Inclusion This work proposes the inclusionof SAI from application layer to MAC Protocol Data Unit(PDU) control elements This inclusion in MAC layer bringsthe processing towards the base station decreasing latencyand reducing the vehicular traffic load on themobile network

In cellular networks eNB uses the cell radio networktemporary identifier (C-RNTI) to uniquely identify UEs Asshown in Figure 3 a unique C-RNTI is assigned to everyuser by the eNB during the initial random access procedurewhich is used for identifying the radio resource control(RRC) connection and for scheduling purposes Coding anddecoding of physical downlink control channel (PDCCH)for a specific UE is based on its C-RNTI UEs initiatecontention based access to the network by transmitting apreamble sequence on the physical random access channel

(PRACH) to which the eNB responds with a C-RNTI onthe physical downlink shared channel (PDSCH) In vehicularcommunications since the awareness range is a function ofvehiclersquos geographical location if the transmitterrsquos location isunknown to the network an additional message containingtransmitterrsquos location is required

So for our proposed system model signaling shown inFigure 3 the transmitter registers with the network in thesame way as discussed above and it includes its location inthe RRC connection request message in its establishmentcause as mo-Data transmitted via the physical uplink sharedchannel (PUSCH) This location is introduced in the RRCconnection request message as a new information elementin line with the 3GPP specifications for RRC connectionrequest [21] At the same time all the UEs already registeredto the network update the eNB of their location using thenetwork assisted global navigation satellite system (GNSS)location update message as specified in [35] Once thevehicle has a beacon ready for transmission it sends outa scheduling request (SR) in the physical uplink controlchannel (PUCCH) The eNB replies back in the downlinkwith the grant for sending BSR After receiving this grantthe vehicle will send out the BSR in PUSCH containing theMAC PDU SAI will be in the unused 16-bit long MACControl Element inside the BSR This element utilizes spacecurrently reserved for future use and is indexed in the MACPDU subheader by the logical channel ID (LCID) value equalto 01011 [36] This new element now referred to as SAIis appended to the existing LCID values such as commoncontrol channel (CCCH) C-RNTI and the padding as shownin Figure 4 The eNB will then use the SAI to determine the

6 Wireless Communications and Mobile Computing

PayloadSAIIP Header

PDCP SDUPDCP header

RLC SDURLC header

MAC SDUMAC header

Transport block CRC

Location coordinates + safety message

PayloadIP SAI

Vehicular safety message

Inclusion of SAI

Header compression

PDCP

RLC

Ciphering

Segmentation

MAC

Multiplexing

PHYOne TB per TTI (1 ms)

Figure 2 SAI inclusion in LTE message structure

transmission parameters allotting transmission resources tothe sender and receiver vehicles

23 Scalability of the System Kato et al [15] emphasizedthe significance of placing the VSA server close to the edgeof LTE network It can also be argued if a single serverwould be able to serve other locations that is multipleeNodeBs or more servers would be required giving rise tothe question of system scalability For the implementationof SAI algorithm at the VSA server the location would playa vital role Considering the data freshness concept ideallocation of the server would be at the EPCTherefore havingmultiple EPCs serving their respective geographical locationswould have their own VSA servers with a similar approachas is for mobility management entity (MME) reducing theRTT while increasing the system capacity and eventuallymeeting the strict transmission requirements for vehicularsafety applications

In the case of having SAI appended in MAC layer theimplementation of SAI algorithm is brought towards theeNodeB reducing the RTT and eventually increasing thesystem capacity For vehicles being served by neighboringeNodeBs the message would be forwarded by the eNodeBserving the source vehicle to the respective neighboring

eNodeB via the X2 interface However if the neighboringeNodeB does not have an X2 interface or is residing in adifferent EPC then the VSA server would come into play Forsuch a scenario the centralization of the server will again besignificant

3 System Model

The network is assumed to be composed of 119873 vehiclesuniquely identified by their number 119894 (119894 = 1 sdot sdot sdot 119873) Vehiclesare assumed to use FDD LTE transceivers with 2 times 10MHzbandwidth uplink carrier frequency 1715MHz and downlinkcarrier frequency 2115MHz (band 4) [34 Table 55-1] Werefer to vehicles as LTE UEs

The work in this paper builds on to the performanceevaluation from [11] adding multicell and multipath alongwith Extended Vehicular A (EVA) fading environment Thenetwork modeled is a 2 times 2 km2 area of Manhattan Grid(MG) and Glasgow city center (GCC) shown in Figures5 and 6 respectively implemented in ns-3 [37] Mobilityof the vehicles in the network is generated using routesmobility model which assigns each vehicle with a routegenerated using Google maps API [14] Along with themobility model the site configuration is also changed for

Wireless Communications and Mobile Computing 7

Vehicle 1 eNB Vehicle 2 Vehicle N

Initial networkaccess

Registered to network

Registered to network

PRACH RA preamble

PDSCH RA responseAssigns C-RNTI 1

PUSCH RRC connection request

PDSCH connection setup

PUSCH RRC connection setup complete

Contains eUE 1location

Assuming norandom access

contention

Network assisted GNSS location update

Network assisted GNSS location update

Beacon for transmission

PUCCH (SR)

PDCCH UL grant for BSR

PUSCH (BSR containing MAC PDU)Contains SAI

Resource allocation

PDCCH TX grant PDCCH RX grant PDCCH RX grant

PUCCH (ACKNACK)PHICH (ACKNACK)

PUSCH beacon transmitted to nUE 2 middot middot middot nUE N

Figure 3 LTE signaling with MAC control elements in BSR

Header Controlelements Payload (SDU) Padding

Subheader1

Subheader2

SubheaderK

MAC Ctrlelement 1

MAC Ctrlelement 2

MAC Ctrlelement N

R ER LCID = 01100 Safety application identifier

MAC PDU

Figure 4 SAI in MAC control elements

8 Wireless Communications and Mobile Computing

500m

S-GWP-GWVSA server

400m

500m1 km

500m

Figure 5 MG model covered by 4 sites and 3 cellssite with 1000mintersite distance (ISD)

GCC This change in the network design considers a morerealistic network model employing the site configurationused by UKrsquos mobile operator EE in Glasgow [8] The LTEfunctionality is implemented through the LTE EPC networksimulator (LENA) module [38] that includes highly detailedeNB and UE functionalities as specified by 3GPP standardsTo manage handover and Internet connections eNBs areinterconnected through their X2 interfaces and connected tothe EPC through S1-U interface Interconnection from the P-GW to the VSA server residing in the EPC is modeled usingan error free 10 Gbps point-to-point link and TCPIP version4

The granularity of the LENA module is up to resourceblock (RB) level including the user plane control planeand reference signals used for coherent reception The safetymessages are UDP packets and often might be segmentedat the radio link control (RLC) and MAC layers to betransmitted on the transport shared channel (S-CH) passedto the physical (PHY) layer The size of transport blocks(TB) used at the PHY layer is decided by the MAC sublayerdepending on the channel quality and scheduling decisionsand may be transmitted using several RBs utilizing a mod-ulation and coding scheme (MCS) [38] Transmissions arescheduled every 1ms corresponding to one subframe usingProportional Fair Algorithm which is one of the approachessuitable for applications that include latency constraints [1Chapter 12] For downlink the algorithm utilizes userrsquoschannel quality indicator (CQI isin [0 sdot sdot sdot 15]) measurementreports sent from the vehicles For uplink channel qualitymeasurements are done by the eNB through scheduling

periodic vehicle transmissions of sounding reference signals(SRS)The radio resource control (RRC)module in LENAns-3 assigns the periodicity of SRS as a function of the numberof UEs attached to an eNB according to 3GPP UE-specificprocedure [37]

Each site modeled by an eNB uses 10W (40 dBm) totaltransmission power and a cosine antenna sector model with65∘ half power beamwidth (HPBW) [39] and azimuthaldirection 0∘ 120∘ or 240∘ with respect to north and 1mintersector space We assume utilization of 2 times 10MHzbandwidth whichmeans there are119872 = 50RBs for the uplinkand for the downlink Therefore assuming that the radiopropagation channel is constant in subframe 119905 if vehicle 119894 isscheduled to receive from its serving cell 119895 on resource block119898 isin [0 sdot sdot sdot 49] the received signal power 119875119895119894 can be modeledas follows

119875119895119894 (119898 119905) = 119875119895 (119898 119905) 119866119895 (120579119895119894)119866119894 (120579119894119895)119871119895119894 (119898 119905) (2)

where 119875119895(119898 119905) is the transmitted power for resource block119898119866119895(120579119895119894) is the antenna gain of cell 119895 in the direction of user 119894119866119894(120579119894119895) is the antenna gain of user 119894 in the direction of cell 119895and 119871119895119894 is the path loss from cell 119895 to user 119894

In the single cell scenario there is no intercell interferenceand mutually exclusive RBs are assigned to users Howeverin a more general deployed urban environment the availablespectrum is utilized by multiple cells in the service area andthe intercell interference is one important limiting aspectespecially for cell-edge users [1 Chapter 12]

In order to allocate usersrsquo transmissions while at the sametime maximizing the signal to interference plus noise ratio(SINR) for each RB frequency reuse and handover are usedFrequency allocation has an impact on the achievable systemcapacity and delay since the transport block size MCS andthe transmission schedule are dependent on the SINR [37Chapter 9] In ns-3 if several RBs are used to transmit a TBthe vector composed of the SINRs on each RB is used toevaluate the TB error probability The SINR in RB 119898 whenuser 119894 is scheduled to receive from serving cell 119895 can bewrittenas follows

SINR119894 (119898 119905)= 119875119895 (119898 119905) 119866119895 (120579119895119894)119866119894 (120579119894119895) 119871119895119894 (119898 119905)1198730 + sumforall119897 =119895 (119875119897 (119898 119905) 119866119897 (120579119897119894) 119866119894 (120579119894119897) 119871 119897119894 (119898 119905))

(3)

where1198730 is the noise powerUEs report the SINR mapped to the CQI value corre-

sponding to the MCS (isin [0 sdot sdot sdot 28]) that ensures TB error ratele 01 according to the standard [40 Section 723] In ns-3 thedownlink SINR is used to generate periodic wideband CQIsfeedback (ie an average value of all RBs) and inband CQIs(ie a value for each RB) We set the CQI to be calculated bycombining the received signal power of the reference signalsto the interference power from the physical downlink sharedchannel as an approximation to (3)

For dynamic frequency allocation we use the distributedfractional frequency reuse algorithm that utilizes intercell

Wireless Communications and Mobile Computing 9

S-GWP-GWVSA

TCC

Figure 6 GCC model covered by 6 sites and 3 cellssite where eNB location is determined using UK operator EE mast data [8]

interference coordination (ICIC) [41] The eNB adaptivelyselects a set of RBs referred to as cell-edge subband (set to10RBs) based on information exchanged with its neighborsTherefore according to ICIC a vehicle should be allocatedin the cell-edge subband if its serving cell reference signalreceived quality (RSRQ) gets lower than minus10 dB

Handover decisions are also performed by eNBs based onconfigurable event-triggeredmeasurement reports includingthe RSRQ and the reference signal received power (RSRP)The measurements are performed by UEs usually in orthog-onal frequency division multiplexing (OFDM) symbols car-rying reference signals (RS) for antenna port 0 In ns-3 since the channel is assumed flat over an RB that iscomposed of 12 subcarriers all resource elements (RE) in anRB have the same power Hence assuming orthogonal RSreception the RSRP for cell 119895 measured by vehicle 119894 is givenby

RSRP119895119894 (119905) = sum119872minus1119898=0 sum119870minus1119896=0 (119875119895119894 (119898 119896 119905) 119870)119872= sum119872minus1119898=0 (119875119895119894 (119898 119905) 12)119872

(4)

where 119875119895119894(119898 119896 119905) is the received power of RS 119896 in RB 119898 119872is the number of RBs and 119870 is the number of RS measuredin the RB The average value is passed to higher layers every200ms therefore the index 119905 is omitted for simplicity

The RSRQ for serving cell 119895measured by vehicle 119894 can becomputed by the following [42]

RSRQ119895119894 = 119872 times RSRP119895119894RSSI119895119894

(5)

where RSSI119895119894 is the received signal strength indicator (RSSI)measured by UE 119894when served by cell 119895 and according to the

10 Wireless Communications and Mobile Computing

Table 3 LTE simulation parameters

Parameter ValueSimulation time 100 seconds

Road model Manhattan grid model (MG) (two way roads)Glasgow City Center (GCC) (2 km times 2 km)

LTE network 4 sites with 3 cellssite 1000m ISD for MG6 sites with 3 cellssite UK Operator EE mast data [8]

Transmission power eNB 40 dBm UE 23 dBmCarrier frequency DLUL 2115MHz1715MHzChannel bandwidth 2 times 10MHz (2 times 50RBs)Noise Figure eNB 5 dB UE 9 dBUE antenna model Isotropic (0 dBi)eNB antenna model 15 dB Cosine model 65∘ HPBWScheduling algorithm Proportional Fair

Handover algorithm A2A4RSRQ RSRQ threshold minus5 dBand NeighbourCellOffset = 2 (1 dB)

Frequency reuse Distributed Fractional Freq Reuse

Path loss model LogDistance (120572 = 3) and3GPP Extended Vehicular A model

Safety message format 256 bytes UDPNumber of vehicles 75 100 125 150Average vehiclersquos speed 20 kmh 40 kmhBeacon frequency 1Hz 10HzAwareness range (119877) 100m 250m 500m 750m 1000m

standard comprises the linear average of the total receivedpower observed by the UE from all sources includingcochannel serving and nonserving cells adjacent channelinterference thermal noise and so forth [42] The RSSImeasurement model in ns-3 is computed considering two REper RB (the one that carries the RS) [37] that is

RSSI119895119894 (119905)= 119872minus1sum119898=0

2 times (119873012 +sumforall119897

119875119897 (119898 119905) 119866119897 (120579119897119894) 119866119894 (120579119894119897)12 times 119871 119897119894 (119898 119905) ) (6)

Furthermore in order to perform handover decisions thereported RSRQ for neighboring cell 119897 when vehicle 119894 is con-nected through serving cell 119895 is computed by the following

RSRQ119897119894 = 119872 times RSRP119897119894RSSI119895119894

(7)

The RSRQs for serving and neighboring cells are usedwith theA2A4RsrqHandoverAlgorithm to trigger LTE eventsA2 and A4 [43 Section 554] Event A2 is set to be triggeredwhen the RSRQ of the serving cell becomes worse than minus5 dB(ServingCellThreshold parameter) While event A2 is trueevent A4 is triggered if a neighboring cell becomes betterthan a threshold which is set by the algorithm to a verylow value for the trigger criteria to be true Thus the mea-surement reports received by the eNB are used to consider aneighboring cell as a target cell for handover only if its RSRQ

is 1 dB higher than the serving cell (NeighbourCellOffsetparameter) Table 3 summarizes the simulation parametersfor LTE

31 Vehicular Radio Urban Environment As mentioned inSection 11 previous evaluations lack the use of multipathand multicell environments Use of proper channel modelingis essential in evaluating system models as it poses close toreality system degradations The system model presented inthis paper consists of 6 sites with 3 cells per site Users areassumed to have isotropic antenna (119866119894(120579119894119895) = 0 dBi) and thepath loss (119871119895119894) ismodeled using the Log-distance propagationwith shadow exponent 119899 = 3 [44] plus 3GPP EVA multipathfading modelL119895119894 [34] and is computed by the following

119871119895119894 (119898 119905) = 1198710 + 10119899 log10 (119889119895119894 (119905)1198890 ) +L119895119894 (119898 119905) dB (8)

where 1198710 = 20 log10(120582(41205871198890)) 120582 is the wavelength and1198890 is the reference distance assumed to be 10m in oursimulations Traces for EVA were generated using MATLABwhile modifying the classical Doppler spectrum for each tapat vehicular speed (V) of 40 kmh and 60 kmh as

119878 (119891) prop 1(1 minus (119891119891119863)2)05 (9)

where 119891119863 = V120582 These EVA traces specified by 3GPP in[34] improvise the NLOS conditions relative to the speed

Wireless Communications and Mobile Computing 11

of UE and the transmission bandwidth Simulation of themultipath fading component in ns-3 uses the trace startingat a random point along the time domain for a given user andserving cell The channel object created with the rayleighchanfunction is used for filtering a discrete-time impulse signal inorder to obtain the channel impulse response The filteringis repeated for different Transmission Time Interval (TTI)thus yielding subsequent time-correlated channel responses(one per TTI) This channel response is then processedwith the pwelch function for obtaining its power spectraldensity values which are then saved in a file with the formatcompatible with the simulator model

32 SAI Implementation Model Archer and Vogel in [45]carried out a survey on traffic safety problems in urbanareas It was concluded that overtaking tends to occur lessfrequently within urban areas where the speed limit isless than 50 kmh Lane changing however occurs quitefrequently within urban areas due to low speed limits Thenumber of rear-end accidents is greater within urban areasthan in rural areas and similarly due to higher level ofcongestion and higher number of traffic junctions there isa greater number of opportunities for turning and crossingaccidents within urban area According to another survey of4500 crashes by the Insurance Institute of Highway Safety[46] 13 of accidents are during lane change maneuvers12 are intersection collisions 22 are due to traffic lightviolation 9 are head on collisions 18 include left turningcrashes 12 are due to emergency vehicles parked on blindturns and 14 are due to over speeding In the light ofthese statistics for our adopted urban environment the SAIs(Table 2) we utilize in our simulations are 1 (25) 3 (22) 5(9) 6 (18) 7 (12) and 9 (14)

33 Background Traffic In order to model close to realscenarios we added background traffic into our simulationsModeling of such traffic is done bymaking 50of the existingvehicles use voice application and 25 stream videos Voiceapplication is modeled by implementing constant bit rate(CBR) traffic generator on G711 codec using onoffapplicationon ns-3 [37] Voice payload is 172 bytes and the data ratechosen is 688 Kbps The reason for using a high data rate isto cater worst case scenario where an operator would preferhaving a high quality voice application At the same timevideo streaming is being carried out using evalvid simulatormodule on H263 codec with a payload of 1460 bytes at a datarate of 122Mbps [47] Both these voice and video applicationsare assigned their respective QCIs

34 Performance Measures The primary performance mea-sures used are the end-to-end packet delay the packet deliveryratio and the downlink goodputThe goodput is defined as thenumber of useful information bits received at the applicationlayer per unit of time

In addition we use the notation |F119894| to represent theaverage number of neighbors calculated over the number ofreceivers whenever a beacon is disseminated The end-to-enddelay of a beacon (119863119894119896) is defined as the time measured fromthe start of its transmission at vehicle 119894rsquos application layer until

its successful reception at vehicle 119896rsquos application layer PDR isdefined as the beacon success rate within the awareness range119877 [48] that is for a beacon sent fromvehicle 119894 its PDR is givenby

119875119894 (119877) = sum119896isinF119894

1198681198941198961003816100381610038161003816F1198941003816100381610038161003816 1003816100381610038161003816F1198941003816100381610038161003816 gt 0 (10)

where | sdot | indicates the cardinality of the set and 119868119894119896 (isin 0 1)is set to 1 if the beacon is received by vehicle 1198964 Simulation Results

The success of cellular networks relies on the scalability ofthe system capacity achieved by reusing the spectrum Thecommon approach implemented by operators is to caterthe capacity needs as the traffic demand grows Thereforefirst we analyze the capacity limit which results from thesingle cell case compared to multiple cells using the Friisradio propagation environment and afterwards we analyzethe impact of the urban radio environment as a function ofthe awareness range

Figure 7(a) shows the cumulative distribution function(CDF) of the end-to-end delay when the network is com-posed of 100 vehicles moving at an average speed of 40 kmhThe service area is covered using a single cell with an isotropicantenna located at the upper left corner of Manhattan Gridmodel as utilized by [11] in a Friis radio environment wherevehicles transmit at 1 Hz beacon frequency We observe thatfor awareness range of up to 500m the end-to-end delay isle100ms with around 95 probability Probability of higherend-to-end delay increases more between 250m to 500mas compared to the increase from 500m to 750m That ismainly the result of the increase in traffic but also as theawareness range increases cars located on roads near the edgeof the Manhattan model will have neighbors towards the celldirection with better signal quality which is further exploitedby the scheduling algorithm Figure 7(a) also shows that thereis no significant increase in end-to-end delay from 750mto 1000m This is due to the border effect incurred by thecar located at the edge of the model having less number ofneighbors when increasing the awareness range as comparedto the cars located at the center of the cell Nevertheless theseresults show that up to 100 vehicles were possible for thesingle cell case at a beacon frequency of 1Hz under Friis radioenvironment In order to analyze the intercell interferenceimpact on the delay while increasing the number of vehiclesand beacon frequency we increase the network capacity bydeploying 4 sites with 3 sectorssite (the average number ofvehicles per sector is reduced to 125) Figure 7(b) shows thatthe end-to-enddelay is then further improved to below 50mswith 91 success rate when awareness range of 1000m at thebeacon frequency of 1Hz was utilized We can also noticethat in this case there are some beacons that experiencedlonger delays compared to the single cell scenario causedby low quality conditions of cell-edge users due to intercellinterference and handover However with 1000m 934 ofthe forwarded beacons were received with end-to-end delayle 100ms and 982 for 250m

12 Wireless Communications and Mobile Computing

0 50 100 150 200 250 3000

01

02

03

04

05

06

07

08

09

1CD

F

End-to-end delay (ms)

R = 100

R = 250

R = 500

R = 750

R = 1000

(a)

0

01

02

03

04

05

06

07

08

09

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 1

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 35

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 14

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 25

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 41

(b)

0

01

02

03

04

05

06

07

08

09

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 1

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 35

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 14

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 25

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 41

(c)

Figure 7 LTE end-to-end delay for 100 vehicles at a beacon frequency 1Hz and average speed 40 kmh with Friis radio environment (a)single cell (b) multicell (4 sites 3 sectorssite) (c) multicell with EVA fading radio environment

The results in Figure 7(b) show that under the Friis radiopropagation environment the capacity increases and thesystem performance is not significantly affected in terms ofthe end-to-end delay in presence of intercell interference andhandover since the scheduler effectively managed interfer-ence by time and frequency allocations However in an urbanenvironment it is expected that fading caused by buildingsand Doppler shift due to the relative velocity of vehicles havea very important impact on the received signal quality [49]Figure 7(c) shows the results when the radio propagationenvironment was changed to Log-distance-dependent and

3GPP EVA under the same site configurations Under thisradio environment the probability for the end-to-end delayto be less than or equal to 100ms was around 76ndash78 forawareness range between 100m and 500m 70 for 750mand 66 for 1000m Hence the significant effect on the end-to-end delay is because of path loss and frequency selectivefading imposed by the EVA fading environmentThe end-to-enddelay for beacons to and fromcell-edge users significantlyincreases Similarly for inner cell users the end-to-end delayremained close to the Friis results and for up to 500m theend-to-end delay was le50ms with 74 to 71 probability

Wireless Communications and Mobile Computing 13

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 17

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 53

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 20

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 37

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 62

(a)

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 11

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 69

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 21

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 47

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 73

(b)

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 13

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 63

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 24

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 47

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 74

(c)

Figure 8 End-to-end delay for sample networks at average speed 40 kmh with 150 vehicles at beacon frequency (a) 1Hz with MG model(b) 1Hz with GCC model (c) 10Hz with GCC model

65 for 750m and 59 for 1000m Therefore it can bededuced that with the addition of multipath fading alongwith critical radio propagation environment probability ofdelay being le50ms decreases by 19 to 30

Continuing on with multicell and EVA fading radioenvironment by increasing the network size the number ofvehicles in the awareness range increases and the end-to-enddelay increases as shown in Figure 8(a) Nevertheless theend-to-end delay remained below 100ms for up to 250mwith82 probability for 100 vehicles and similar rate was obtainedwith up to 150 vehicles When the awareness range increasesto 500m the probability was 80 for 100 vehicles while for150 vehicles further reduces to 73 For awareness range of750m and 1000m the probability reduced to 75 and 71 for100 vehicles while for 150 vehicles reduces to 69 and 60respectively

Changing the simulation model along with mobilitytraces to Glasgow city center the effect on delay performancewith 150 vehicles at 1 Hz and 10Hz is shown in Figures 8(b)and 8(c) respectively It is interesting to observe the effect ofchanging the mobility model to a European city Where each

city block is 400m in theMGmodel average city block size inGCC is around 200m increasing vehicle density eventuallygrowing the average number of neighbors for GCC modelThis can be observed for an awareness range of 250m theprobability of delay to be le50ms decreases from 77 to 68while for 1000m this deterioration goes from 54 to 37hence an overall decrease of about 9 to 17 for variousawareness ranges Similarly in Figure 8(c) it is observedthat changing the beacon frequency from 1Hz to 10Hz withan awareness range of 500m the probability of delay to bele50ms drastically decreases from 55 to 22 Whereas foran awareness range of 1000m degradation of about 29 isobserved Performance with a high beacon frequency suchas 10Hz is overall seen to be poor That is because packetsthat arrive late are discarded and are mainly sent by cell-edgeusers for which their neighborhood very likely includes usersat the cell edge Hence the delay from the uplink is indirectlyused as a spatial filtering of low quality cell-edge users Thissuggests that if the information is not useful when receivedafter 100ms the better use of resources is to drop the packetat the minimum latency requirement by the server

14 Wireless Communications and Mobile Computing

0010203040506070809

1

0 20 40 60 80 100 120 140 160 180 200

CDF

End-to-end delay (ms)

SAI = 11003816100381610038161003816Fi

1003816100381610038161003816 = 19

SAI = 31003816100381610038161003816Fi

1003816100381610038161003816 = 16

SAI = 51003816100381610038161003816Fi

1003816100381610038161003816 = 41

SAI = 61003816100381610038161003816Fi

1003816100381610038161003816 = 14

SAI = 71003816100381610038161003816Fi

1003816100381610038161003816 = 56

SAI = 91003816100381610038161003816Fi

1003816100381610038161003816 = 54

Without SAI 1003816100381610038161003816Fi1003816100381610038161003816 = 16

(a)

0010203040506070809

1

0 20 40 60 80 100 120 140 160 180 200

CDF

End-to-end delay (ms)

SAI = 11003816100381610038161003816Fi

1003816100381610038161003816 = 26

SAI = 31003816100381610038161003816Fi

1003816100381610038161003816 = 2

SAI = 51003816100381610038161003816Fi

1003816100381610038161003816 = 35

SAI = 61003816100381610038161003816Fi

1003816100381610038161003816 = 22

SAI = 71003816100381610038161003816Fi

1003816100381610038161003816 = 78

SAI = 91003816100381610038161003816Fi

1003816100381610038161003816 = 94

Without SAI 1003816100381610038161003816Fi1003816100381610038161003816 = 21

(b)

Figure 9 LTE end-to-end delay at average speed 40 kmh (GCC) and (a) 100 vehicles and (b) 150 vehicles

Next we observe and compare the results after theproposed algorithm implementation Figure 9 shows theCDF of the end-to-end delay for 100 and 150 vehicles atan average speed of 40 kmh with and without the imple-mentation of SAI algorithm The scenario being comparedhas the awareness range set to 500m and beacon frequencyto 10Hz We observe a significant decrease in the end-to-end delay after SAI algorithm implementation Probabilityfor the end-to-end delay to be less than 100ms is above80 in both scenarios whereas without the algorithm thisprobability is below 70 for 100 vehicles (Figure 9(a)) andbelow 60 for 150 vehicles (Figure 9(b)) The variation inthis delay between various SAIs is because of the differencein transmission parameters Larger awareness range leadingto higher number of neighbors |F119894| and higher beaconfrequency results in more congestion and large queuingtimes The reason for better delay values with SAI algorithmis mainly the result from restricting resources to the requiredtransmission parameters for safety applicationsmentioned inSection 32

Figure 10 shows the aggregate downlink goodput for thesame scenario investigated for end-to-end delay After theimplementation of SAI algorithm the downlink goodput for100 vehicles dropped from 746Mbps to 622Mbps and for150 vehicles it dropped from 2167Mbps to 1376Mbps Thissignificant decrease in the downlink direction is again due tothe restriction of unnecessary data dissemination from theVSA server to vehicles Eventually this decrease of goodputresults in less load on the LTE system The increase in thedownlink goodput for the scenario without SAI explains thehigh end-to-end delay observed in Figure 9 showing a burstof packet in the downlink leading to waiting time in thequeue

Furthermore the impact of regular cellular traffic on ourLTE vehicular network is studied in order to validate ourfindings and algorithm A broad picture of our observa-tions under different scenarios is shown in Figure 11 whereprobability of end-to-end delay being less than or equal to50ms is studied As expected this probability decreases byabout 3-4 when background traffic is added However with

100 150

Number of vehicles

25

20

15

10

5

0

Dow

nlin

k go

odpu

t (M

bps)

Without SAIWith SAI

746622

2167

1376

Figure 10 Downlink goodput for 100 and 150 vehicles at an averagespeed 40 kmh (GCC)

SAI algorithm probability of end-to-end delay being lessthan 50ms stays above 90 even after adding backgroundtraffic

Finally we include the SAI into MAC layer PDU con-trol elements to prove our concept of D2D like unicasttransmission between sender and receivers Modeling of thesystem is carried out in accordance with the description inSection 22 Figure 12 shows end-to-end delay for 100 and150 vehicles with and without the inclusion of SAI in MAClayer It is evident that 150 vehicles experience almost thesame probability for end-to-end delay being less than 50msas experienced by 100 vehicles This shows that with the useof MAC layer control elements around 50 more vehiclescan be accommodated by LTE network increasing the spec-trum efficiency eventually leading to higher capacity of thesystem

Wireless Communications and Mobile Computing 15

Simple case With backgroundtraffic

With SAI With backgroundand SAI

6065707580859095

100

150 vehicles100 vehicles

Prob

abili

ty o

f end

-to-e

ndde

layle

50

ms

Scenarios

Figure 11 Probability of end-to-end delay being le50ms for 100 and150 vehicles at 40 kmh in various scenarios

With SAI in APP layer With SAI in MAC layer80828486889092949698

100

100 vehicles150 vehicles

p

roba

bilit

y fo

r end

-to-e

ndde

layle50

ms

Figure 12 Probability of end-to-end delay being le50ms for 100 and150 vehicles with and without MAC layer inclusion

5 Conclusion

This paper proposes the safety application identifier conceptin the form of an algorithm that is tested and analyzedwithin the impact of vehicular urbanmulticell radio environ-ment employing multipath fading channels and backgroundtraffic The proposed algorithm uses dynamic adaptation oftransmission parameters in order to fulfill stringent vehicu-lar application requirements while minimizing latency andreducing system load

With the help of extensive system-level simulations thecrucial impact of channel modeling and mobility traces isevident with its influence on the received signal qualitydecreasing the probability of end-to-end delay to be le50msfor various awareness ranges by 19 to 30 and 9 to17 respectively Studying the impact of awareness rangeit is observed that with range up to 1000m for 1Hz 500mfor 2Hz and 250m for 10Hz beacon frequency the systemperformed well With the help of these findings the SAIalgorithm is proposed and implemented Probability of expe-riencing an end-to-end delay of less than 100ms increasedby around 20 and the downlink goodput decreased signif-icantly from 2167Mbps to 1376Mbps for 150 vehicles

Furthermore the impact of having background traffic isalso taken into consideration Results show that the systemis slightly affected by having regular background trafficHowever the probability of end-to-end delay being le50ms

still stays satisfactorily above 85 with the use of SAI Thispaper also proposes the inclusion of SAI inMAC layer controlelements producing a D2D type of communication withinvehicles After modeling this system it was found that thesame amount of system resources and requirements are metwith additional 50 vehicles increasing the capacity of ourLTE vehicular network Finally in future works we plan tointegrate SAI incorporated within MAC layer in a hetero-geneous network with DSRC while exploring the networkslicing in 5G cellular networks for vehicular communicationscomplementing each otherrsquos weaknesses and strengths inorder to meet the vehicular network requirements

Conflicts of Interest

The authors declare that they have no conflicts of interest

Acknowledgments

Marvin Sanchez would like to acknowledge the support ofLAMENITEC Erasmus Mundus Program of the EuropeanUnion to be a guest researcher at GCU Results were obtainedusing the EPSRC funded ARCHIE-WeSt High PerformanceComputer (httpswwwarchie-westacuk) EPSRC Grantno EPK0005861

References

[1] S Sesia I Toufik and M Baker LTE the UMTS long termevolution from theory to practice Wiley Publishing 2009

[2] H Hartenstein and K P Laberteaux ldquoA tutorial survey onvehicular ad hoc networksrdquo IEEE Communications Magazinevol 46 no 6 pp 164ndash171 2008

[3] D Jiang and L Delgrossi ldquoTowards an international standardfor wireless access in vehicular environmentsrdquo in Proceedings ofthe 67th IEEE Vehicular Technology Conference (VTC rsquo08) pp2036ndash2040 May 2008

[4] S Al-Sultan M M Al-Doori A H Al-Bayatti and H ZedanldquoA comprehensive survey on vehicular Ad Hoc networkrdquoJournal of Network and Computer Applications vol 37 no 1 pp380ndash392 2014

[5] K Tanuja T Sushma M Bharathi and K Arun ldquoA surveyon VANET technologiesrdquo International Journal of ComputerApplications vol 121 no 18 2015

[6] H Y Kim D M Kang J H Lee and T M Chung ldquoA per-formance evaluation of cellular network suitability for VANETrdquoWorld Academy of Science Engineering and Technology Interna-tional Science Index vol 64 no 6 pp 124ndash127 2012

[7] A Fasbender M Gerdes and S Smets ldquoCellular NetworkingTechnologies in ITS Solutions Opportunities and Challengesrdquopp 1ndash13 Springer Berlin Germany 2012

[8] ldquoCellular coverage and tower map for ee network in glasgowrdquohttpswwwcellmappernetmap

[9] GAraniti C CampoloMCondoluci A Iera andAMolinaroldquoLTE for vehicular networking a surveyrdquo IEEECommunicationsMagazine vol 51 no 5 pp 148ndash157 2013

[10] A Vinel ldquo3GPP LTE versus IEEE 80211pWAVE which tech-nology is able to support cooperative vehicular safety applica-tionsrdquo IEEE Wireless Communications Letters vol 1 no 2 pp125ndash128 2012

Wireless Communications and Mobile Computing 3

Broadcast and unicast transmissions with LTE have beeninvestigated in [10ndash12] Vinel [10] used a simplified model inorder to compute analytically the packet delivery ratio (PDR)assuming fixed number of neighbors (50 vehicles) in a timedivision duplex (TDD) LTE single cell system Due to theoverall system complexity of LTE system-level simulationshave become an indispensable tool for predicting the per-formance over more realistic scenarios In [11] simulationswere used to compare IEEE 80211p with frequency divisionduplex (FDD) LTE single cell system performance usingFriis radio propagation environment on a Manhattan girdmobility model Their performance evaluation showed thatLTE is suitable for VANETs however the use of single cellsimplified higher layers and Friis radio environment does nottest the systemunder extreme propagation scenariosWhen itcomes to system-level simulations choice of mobility modelscarries much importance Grzybek et al in [13] studied andcompared various mobility generators for VANETs propos-ing a mobility generator for realistic traces called vehiluxThis generates traces considering the spatial temporal andbehavioral aspects of traffic distribution Since the scope ofthis paper is within the transmission requirements of safetyapplications the trace generator adopted is routes mobilitymodel [14]This generates traces with geographic restrictionson top of spatial temporal and behavioral aspects allowinga fixed number of vehicles in the area throughout thesimulation time

Kato et al [15] defined the concept of data freshnessand used this parameter to evaluate the performance of LTEnetworks with VANETs If a vehicle application transmitsa packet at a frequency of 10Hz that is a message istransmitted every 100ms the freshness requirement for thatspecific application would be 100ms They observed thatit is possible to achieve a freshness of 100ms by havingvehicular application server at the edge of LTE networkhence discarding the message if it arrives later than thebeacon interval time Authors evaluated the impact of serverlocation by calculating the amount of network usage Theyconcluded that by reducing the round trip time (RTT) ofthe system network usage can be reduced to half hence byplacing the server close to the eNodeB or even the LTE core itcan have a significant impact on the scalability of the system

In order to utilize broadcast functionality on LTEoperators must implement multimedia broadcast multicastservices (MBMS) [1 Chapter 14] Without MBMS currentlyonly low rate public warning system (PWS) messages can besent through the broadcast capability of LTE [16 17] Hencein some situations unicast is the only available schemeFor instance the experimental study in [6] used TCPIPping application to measure the end-to-end delay in orderto evaluate the suitability of LTE for VANET Also unicastfor cooperative awareness message (CAM) transmissions inmulticell environment with LTE has been studied in [12] Inthat study it was found that the downlink was the bottleneckdue to high traffic load However the study was oriented todetermine the capacity limit therefore the downlink trafficwas modeled by fixed size periodic packet transmission thehandover effect was neglected and did not include the end-to-end system Previous performance evaluations have also

suggested that the use of LTE for vehicular communications issuitable but without any centralization it can put enormousload on the network [18] In the pursuit of centralizing vehic-ular communications in LTE group formation and MBMSimplementation have been proposed in [18 19] Groupformation or clustering has shown promising performanceHowever clustering relies on relaying transmissions whichcan pose a privacy and security issue [20]

3GPP release-12 [21] includes device-to-device (D2D)communication under LTE-Advanced which can be a poten-tial candidate technology for V2V communications Bazzi[22] investigated the use of LTE direct communications(D2D) using full duplex radios in order to enable vehiclesto receive and transmit at the same time Results from theirdeveloped analytical framework showed promising reductionof LTE uplink resources as compared with using unicastand broadcast However the proposed framework exhibitsminimum interference only when the awareness range is setto 100m In [23] authors have proposed a scheme that utilizesan unused 16-bit long MAC Control Element inside thebuffer status report (BSR) request for the network to identifythe receiver and transmitter This element uses the spacereserved for future use and is indexed in the MAC ProtocolData Unit (PDU) subheader by the logical channel identifier(LCID) By doing so packets sent from the transmitter getto the receiver via eNodeB instead of traversing throughthe evolved packet core (EPC) saving the round trip timefrom the LTE core network (CN) to the access network Asimilar concept has been adopted in our work exploringthe possibility of using MAC control elements for messagedissemination within vehicular networks in order to reducelatency

The remainder of this paper is organized as follows inSection 12 the vehicular safety requirements and awarenessrange are introduced followed by the proposed algorithmin Section 2 the system model used for LTE along withthe performance measures is then included in Section 3finally the simulation results and conclusions are discussedin Sections 4 and 5 respectively

12 Vehicular Safety Applications In vehicular safety applica-tions the consequence of failure in message delivery withina minimum delay and awareness range may result in a fatalaccident [24] In Table 1 we show a survey of vital active roadsafety applications along with their requirements for criticallatency beacon frequency and transmission range [25ndash29]

Overall these applications require a critical latency lessthan or equal to 100ms and they differ in the message typeminimum frequency and transmission range requirementsWithout any broadcast routing algorithms the packet deliv-ery rate for VANETs is generally low ranging between 60and 80 [30] However the standards do not define anacceptable packet delivery rate

Awareness range previously elaborated in [9 22 31]is the geographical area around the vehicle where all theneighbors are to bemade cognizant of the vehicle Dependingon the applicability of the message this range can vary Thereare two types of messages involved in VANETs classifiedas CAMs [26] and Decentralized Environment Notification

4 Wireless Communications and Mobile Computing

Table 2 Safety application requirements and SAIs

Safety application Message type Transmitted data Criticallatency

Transmissionfrequency BF

Transmissionrange 119877 SAI

Intersection collisionwarning CAM Vehicle type position heading velocity

acceleration yaw rate sim100ms 10Hz 150m

1Lane change assistance CAM Position heading velocity accelerationturn signal status sim100ms 10Hz 150m

Cooperative forwardcollision warning CAM Vehicle type position heading velocity

acceleration yaw rate sim100ms 10Hz 150m

Slow vehicle indication CAMDENM Vehicle type position headingacceleration velocity sim100ms 2Hz 200m 2

Traffic light speedadvisoryviolation CAMDENM Signal phase timing position direction

road geometry sim100ms 2Hz 150m3

Motorcycle approachingindication CAM Vehicle type position heading velocity sim100ms 2Hz 150m

Overtaking vehiclewarning CAM Position velocity yaw rate acceleration sim100ms 10Hz 300m 4

Head on collisionwarning CAM Vehicle type position heading velocity

acceleration yaw rate sim100ms 10Hz 200m 5

Collision risk warning CAMDENM Vehicle type position heading velocityacceleration yaw rate sim100ms 10Hz 300ndash500m 6

Emergency vehiclewarning CAMDENM Position heading velocity acceleration sim100ms 2Hz 300m 7

Cooperative mergingassistance CAM Curve location curvature slope speed

limit surface sim1000ms 1Hz 250m 8

Speed limits notification CAM Velocity acceleration position speedlimit heading sim100ms 1ndash10Hz 300m 9

Message (DENMs) [27] CAMs are periodic messages thatprovide information like the presence position and sensorinformation and are expected to be received by all thevehicles within the awareness range DENMs are used forcooperative road hazard warnings that are broadcasted intheir relevance area whenever a hazardous event occurs [32]The work in [33] proposes the use of dynamic adaptation ofcommunication radius in order to maximize sum rate weadopt the same concept and apply it to dynamically adaptingcommunication range in order tominimize end-to-end delayFigure 1 illustrates the awareness range119877 for vehicle 119894 Noticethat vehicle 119896 is within the awareness range of 119894 since thedistance 119889119894119896 le 1198772 Safety Application Identifier Algorithm

Message delivery and reliability are a major concern in vehic-ular networks due to which a differentiated QoS mechanismis proposedThismechanismworks on the principle of index-ing various applications according to their requirementsmotivated from QoS class identifier (QCI) implemented inLTE networks [34] In Table 2 each application fromTable 1 iscategorized and assigned an SAIThis SAI which is a numberranging between 1 and 9 is included in the transmitted packetappended before the IP header by the application layer asshown in Figure 2 The categorization and assignment of SAIfor ourmodel are further discussed in Section 3This conceptis implemented in the form of an algorithm proposed in thefollowing subsection

Input VSMs Vehicles rarr ServerOutput VSMs Server rarr Vehicles

System Setup(1) while Server larr VSM119894 do(2) Server locates all vehicles(3) VSM119894 rarr (SAI119894Position (119889119896))(4) SAI119894 rarr (119877119894BF119894)(5) (119877119894 119889119896) rarr Distance (119889119894119896)(6) F119894 = forall119896 119889119894119896 lt 119877119894 119894 = 119896(7) Server rArr Vehicles isin F119894 at BF119894(8) end while(9) return F119894

Algorithm 1 SAI Algorithm

21 SAI Algorithm With the concept of safety applicationindexing we propose an algorithm that does not includethe use of group formation or IEEE 80211p The proposedalgorithm is a process that is carried out on the vehicularsafety application (VSA) server elaborated in Algorithm 1When the vehicles start transmissions the packet is carriedfrom the eNodeB to the VSA server via the CN This packetcontains vehicle location (119889119896) and corresponding vehicularsafety message (VSM) As soon as the server starts receivingthese messages it locates all the vehicles being served bythe network forming a virtual geographical map of theserved area The VSA server while receiving the packets

Wireless Communications and Mobile Computing 5

S-GWP-GWVSA

Server

Awareness Range

VSM

S-GW serving gatewayP-GW PDN gatewayVSA vehicular safety application

idik

k

R

Position dkSAIi (Ri BFi)

Figure 1 LTE multicell system and awareness range

will also extract the SAI information from the receivedpacket This SAI information will then be checked againstthe database stored in the server (Table 2) With this SAIthe server retrieves the transmission beacon frequency BF119894and awareness range 119877119894 requirement of the application thatvehicle 119894 is utilizing Using the required awareness range theserver determines the forwarding set of vehicles (neighboringvehicles) by

F119894 = forall119896 119889119894119896 lt 119877119894 119894 = 119896 (1)

where 119889119894119896 is the distance from vehicle 119894 to the neighboringvehicle 119896 Recalling the concept of data freshness defined by[15] the VSA server will discard the packet if it arrives laterthan the beacon inter arrival time (1BF119894)22 MAC Layer Inclusion This work proposes the inclusionof SAI from application layer to MAC Protocol Data Unit(PDU) control elements This inclusion in MAC layer bringsthe processing towards the base station decreasing latencyand reducing the vehicular traffic load on themobile network

In cellular networks eNB uses the cell radio networktemporary identifier (C-RNTI) to uniquely identify UEs Asshown in Figure 3 a unique C-RNTI is assigned to everyuser by the eNB during the initial random access procedurewhich is used for identifying the radio resource control(RRC) connection and for scheduling purposes Coding anddecoding of physical downlink control channel (PDCCH)for a specific UE is based on its C-RNTI UEs initiatecontention based access to the network by transmitting apreamble sequence on the physical random access channel

(PRACH) to which the eNB responds with a C-RNTI onthe physical downlink shared channel (PDSCH) In vehicularcommunications since the awareness range is a function ofvehiclersquos geographical location if the transmitterrsquos location isunknown to the network an additional message containingtransmitterrsquos location is required

So for our proposed system model signaling shown inFigure 3 the transmitter registers with the network in thesame way as discussed above and it includes its location inthe RRC connection request message in its establishmentcause as mo-Data transmitted via the physical uplink sharedchannel (PUSCH) This location is introduced in the RRCconnection request message as a new information elementin line with the 3GPP specifications for RRC connectionrequest [21] At the same time all the UEs already registeredto the network update the eNB of their location using thenetwork assisted global navigation satellite system (GNSS)location update message as specified in [35] Once thevehicle has a beacon ready for transmission it sends outa scheduling request (SR) in the physical uplink controlchannel (PUCCH) The eNB replies back in the downlinkwith the grant for sending BSR After receiving this grantthe vehicle will send out the BSR in PUSCH containing theMAC PDU SAI will be in the unused 16-bit long MACControl Element inside the BSR This element utilizes spacecurrently reserved for future use and is indexed in the MACPDU subheader by the logical channel ID (LCID) value equalto 01011 [36] This new element now referred to as SAIis appended to the existing LCID values such as commoncontrol channel (CCCH) C-RNTI and the padding as shownin Figure 4 The eNB will then use the SAI to determine the

6 Wireless Communications and Mobile Computing

PayloadSAIIP Header

PDCP SDUPDCP header

RLC SDURLC header

MAC SDUMAC header

Transport block CRC

Location coordinates + safety message

PayloadIP SAI

Vehicular safety message

Inclusion of SAI

Header compression

PDCP

RLC

Ciphering

Segmentation

MAC

Multiplexing

PHYOne TB per TTI (1 ms)

Figure 2 SAI inclusion in LTE message structure

transmission parameters allotting transmission resources tothe sender and receiver vehicles

23 Scalability of the System Kato et al [15] emphasizedthe significance of placing the VSA server close to the edgeof LTE network It can also be argued if a single serverwould be able to serve other locations that is multipleeNodeBs or more servers would be required giving rise tothe question of system scalability For the implementationof SAI algorithm at the VSA server the location would playa vital role Considering the data freshness concept ideallocation of the server would be at the EPCTherefore havingmultiple EPCs serving their respective geographical locationswould have their own VSA servers with a similar approachas is for mobility management entity (MME) reducing theRTT while increasing the system capacity and eventuallymeeting the strict transmission requirements for vehicularsafety applications

In the case of having SAI appended in MAC layer theimplementation of SAI algorithm is brought towards theeNodeB reducing the RTT and eventually increasing thesystem capacity For vehicles being served by neighboringeNodeBs the message would be forwarded by the eNodeBserving the source vehicle to the respective neighboring

eNodeB via the X2 interface However if the neighboringeNodeB does not have an X2 interface or is residing in adifferent EPC then the VSA server would come into play Forsuch a scenario the centralization of the server will again besignificant

3 System Model

The network is assumed to be composed of 119873 vehiclesuniquely identified by their number 119894 (119894 = 1 sdot sdot sdot 119873) Vehiclesare assumed to use FDD LTE transceivers with 2 times 10MHzbandwidth uplink carrier frequency 1715MHz and downlinkcarrier frequency 2115MHz (band 4) [34 Table 55-1] Werefer to vehicles as LTE UEs

The work in this paper builds on to the performanceevaluation from [11] adding multicell and multipath alongwith Extended Vehicular A (EVA) fading environment Thenetwork modeled is a 2 times 2 km2 area of Manhattan Grid(MG) and Glasgow city center (GCC) shown in Figures5 and 6 respectively implemented in ns-3 [37] Mobilityof the vehicles in the network is generated using routesmobility model which assigns each vehicle with a routegenerated using Google maps API [14] Along with themobility model the site configuration is also changed for

Wireless Communications and Mobile Computing 7

Vehicle 1 eNB Vehicle 2 Vehicle N

Initial networkaccess

Registered to network

Registered to network

PRACH RA preamble

PDSCH RA responseAssigns C-RNTI 1

PUSCH RRC connection request

PDSCH connection setup

PUSCH RRC connection setup complete

Contains eUE 1location

Assuming norandom access

contention

Network assisted GNSS location update

Network assisted GNSS location update

Beacon for transmission

PUCCH (SR)

PDCCH UL grant for BSR

PUSCH (BSR containing MAC PDU)Contains SAI

Resource allocation

PDCCH TX grant PDCCH RX grant PDCCH RX grant

PUCCH (ACKNACK)PHICH (ACKNACK)

PUSCH beacon transmitted to nUE 2 middot middot middot nUE N

Figure 3 LTE signaling with MAC control elements in BSR

Header Controlelements Payload (SDU) Padding

Subheader1

Subheader2

SubheaderK

MAC Ctrlelement 1

MAC Ctrlelement 2

MAC Ctrlelement N

R ER LCID = 01100 Safety application identifier

MAC PDU

Figure 4 SAI in MAC control elements

8 Wireless Communications and Mobile Computing

500m

S-GWP-GWVSA server

400m

500m1 km

500m

Figure 5 MG model covered by 4 sites and 3 cellssite with 1000mintersite distance (ISD)

GCC This change in the network design considers a morerealistic network model employing the site configurationused by UKrsquos mobile operator EE in Glasgow [8] The LTEfunctionality is implemented through the LTE EPC networksimulator (LENA) module [38] that includes highly detailedeNB and UE functionalities as specified by 3GPP standardsTo manage handover and Internet connections eNBs areinterconnected through their X2 interfaces and connected tothe EPC through S1-U interface Interconnection from the P-GW to the VSA server residing in the EPC is modeled usingan error free 10 Gbps point-to-point link and TCPIP version4

The granularity of the LENA module is up to resourceblock (RB) level including the user plane control planeand reference signals used for coherent reception The safetymessages are UDP packets and often might be segmentedat the radio link control (RLC) and MAC layers to betransmitted on the transport shared channel (S-CH) passedto the physical (PHY) layer The size of transport blocks(TB) used at the PHY layer is decided by the MAC sublayerdepending on the channel quality and scheduling decisionsand may be transmitted using several RBs utilizing a mod-ulation and coding scheme (MCS) [38] Transmissions arescheduled every 1ms corresponding to one subframe usingProportional Fair Algorithm which is one of the approachessuitable for applications that include latency constraints [1Chapter 12] For downlink the algorithm utilizes userrsquoschannel quality indicator (CQI isin [0 sdot sdot sdot 15]) measurementreports sent from the vehicles For uplink channel qualitymeasurements are done by the eNB through scheduling

periodic vehicle transmissions of sounding reference signals(SRS)The radio resource control (RRC)module in LENAns-3 assigns the periodicity of SRS as a function of the numberof UEs attached to an eNB according to 3GPP UE-specificprocedure [37]

Each site modeled by an eNB uses 10W (40 dBm) totaltransmission power and a cosine antenna sector model with65∘ half power beamwidth (HPBW) [39] and azimuthaldirection 0∘ 120∘ or 240∘ with respect to north and 1mintersector space We assume utilization of 2 times 10MHzbandwidth whichmeans there are119872 = 50RBs for the uplinkand for the downlink Therefore assuming that the radiopropagation channel is constant in subframe 119905 if vehicle 119894 isscheduled to receive from its serving cell 119895 on resource block119898 isin [0 sdot sdot sdot 49] the received signal power 119875119895119894 can be modeledas follows

119875119895119894 (119898 119905) = 119875119895 (119898 119905) 119866119895 (120579119895119894)119866119894 (120579119894119895)119871119895119894 (119898 119905) (2)

where 119875119895(119898 119905) is the transmitted power for resource block119898119866119895(120579119895119894) is the antenna gain of cell 119895 in the direction of user 119894119866119894(120579119894119895) is the antenna gain of user 119894 in the direction of cell 119895and 119871119895119894 is the path loss from cell 119895 to user 119894

In the single cell scenario there is no intercell interferenceand mutually exclusive RBs are assigned to users Howeverin a more general deployed urban environment the availablespectrum is utilized by multiple cells in the service area andthe intercell interference is one important limiting aspectespecially for cell-edge users [1 Chapter 12]

In order to allocate usersrsquo transmissions while at the sametime maximizing the signal to interference plus noise ratio(SINR) for each RB frequency reuse and handover are usedFrequency allocation has an impact on the achievable systemcapacity and delay since the transport block size MCS andthe transmission schedule are dependent on the SINR [37Chapter 9] In ns-3 if several RBs are used to transmit a TBthe vector composed of the SINRs on each RB is used toevaluate the TB error probability The SINR in RB 119898 whenuser 119894 is scheduled to receive from serving cell 119895 can bewrittenas follows

SINR119894 (119898 119905)= 119875119895 (119898 119905) 119866119895 (120579119895119894)119866119894 (120579119894119895) 119871119895119894 (119898 119905)1198730 + sumforall119897 =119895 (119875119897 (119898 119905) 119866119897 (120579119897119894) 119866119894 (120579119894119897) 119871 119897119894 (119898 119905))

(3)

where1198730 is the noise powerUEs report the SINR mapped to the CQI value corre-

sponding to the MCS (isin [0 sdot sdot sdot 28]) that ensures TB error ratele 01 according to the standard [40 Section 723] In ns-3 thedownlink SINR is used to generate periodic wideband CQIsfeedback (ie an average value of all RBs) and inband CQIs(ie a value for each RB) We set the CQI to be calculated bycombining the received signal power of the reference signalsto the interference power from the physical downlink sharedchannel as an approximation to (3)

For dynamic frequency allocation we use the distributedfractional frequency reuse algorithm that utilizes intercell

Wireless Communications and Mobile Computing 9

S-GWP-GWVSA

TCC

Figure 6 GCC model covered by 6 sites and 3 cellssite where eNB location is determined using UK operator EE mast data [8]

interference coordination (ICIC) [41] The eNB adaptivelyselects a set of RBs referred to as cell-edge subband (set to10RBs) based on information exchanged with its neighborsTherefore according to ICIC a vehicle should be allocatedin the cell-edge subband if its serving cell reference signalreceived quality (RSRQ) gets lower than minus10 dB

Handover decisions are also performed by eNBs based onconfigurable event-triggeredmeasurement reports includingthe RSRQ and the reference signal received power (RSRP)The measurements are performed by UEs usually in orthog-onal frequency division multiplexing (OFDM) symbols car-rying reference signals (RS) for antenna port 0 In ns-3 since the channel is assumed flat over an RB that iscomposed of 12 subcarriers all resource elements (RE) in anRB have the same power Hence assuming orthogonal RSreception the RSRP for cell 119895 measured by vehicle 119894 is givenby

RSRP119895119894 (119905) = sum119872minus1119898=0 sum119870minus1119896=0 (119875119895119894 (119898 119896 119905) 119870)119872= sum119872minus1119898=0 (119875119895119894 (119898 119905) 12)119872

(4)

where 119875119895119894(119898 119896 119905) is the received power of RS 119896 in RB 119898 119872is the number of RBs and 119870 is the number of RS measuredin the RB The average value is passed to higher layers every200ms therefore the index 119905 is omitted for simplicity

The RSRQ for serving cell 119895measured by vehicle 119894 can becomputed by the following [42]

RSRQ119895119894 = 119872 times RSRP119895119894RSSI119895119894

(5)

where RSSI119895119894 is the received signal strength indicator (RSSI)measured by UE 119894when served by cell 119895 and according to the

10 Wireless Communications and Mobile Computing

Table 3 LTE simulation parameters

Parameter ValueSimulation time 100 seconds

Road model Manhattan grid model (MG) (two way roads)Glasgow City Center (GCC) (2 km times 2 km)

LTE network 4 sites with 3 cellssite 1000m ISD for MG6 sites with 3 cellssite UK Operator EE mast data [8]

Transmission power eNB 40 dBm UE 23 dBmCarrier frequency DLUL 2115MHz1715MHzChannel bandwidth 2 times 10MHz (2 times 50RBs)Noise Figure eNB 5 dB UE 9 dBUE antenna model Isotropic (0 dBi)eNB antenna model 15 dB Cosine model 65∘ HPBWScheduling algorithm Proportional Fair

Handover algorithm A2A4RSRQ RSRQ threshold minus5 dBand NeighbourCellOffset = 2 (1 dB)

Frequency reuse Distributed Fractional Freq Reuse

Path loss model LogDistance (120572 = 3) and3GPP Extended Vehicular A model

Safety message format 256 bytes UDPNumber of vehicles 75 100 125 150Average vehiclersquos speed 20 kmh 40 kmhBeacon frequency 1Hz 10HzAwareness range (119877) 100m 250m 500m 750m 1000m

standard comprises the linear average of the total receivedpower observed by the UE from all sources includingcochannel serving and nonserving cells adjacent channelinterference thermal noise and so forth [42] The RSSImeasurement model in ns-3 is computed considering two REper RB (the one that carries the RS) [37] that is

RSSI119895119894 (119905)= 119872minus1sum119898=0

2 times (119873012 +sumforall119897

119875119897 (119898 119905) 119866119897 (120579119897119894) 119866119894 (120579119894119897)12 times 119871 119897119894 (119898 119905) ) (6)

Furthermore in order to perform handover decisions thereported RSRQ for neighboring cell 119897 when vehicle 119894 is con-nected through serving cell 119895 is computed by the following

RSRQ119897119894 = 119872 times RSRP119897119894RSSI119895119894

(7)

The RSRQs for serving and neighboring cells are usedwith theA2A4RsrqHandoverAlgorithm to trigger LTE eventsA2 and A4 [43 Section 554] Event A2 is set to be triggeredwhen the RSRQ of the serving cell becomes worse than minus5 dB(ServingCellThreshold parameter) While event A2 is trueevent A4 is triggered if a neighboring cell becomes betterthan a threshold which is set by the algorithm to a verylow value for the trigger criteria to be true Thus the mea-surement reports received by the eNB are used to consider aneighboring cell as a target cell for handover only if its RSRQ

is 1 dB higher than the serving cell (NeighbourCellOffsetparameter) Table 3 summarizes the simulation parametersfor LTE

31 Vehicular Radio Urban Environment As mentioned inSection 11 previous evaluations lack the use of multipathand multicell environments Use of proper channel modelingis essential in evaluating system models as it poses close toreality system degradations The system model presented inthis paper consists of 6 sites with 3 cells per site Users areassumed to have isotropic antenna (119866119894(120579119894119895) = 0 dBi) and thepath loss (119871119895119894) ismodeled using the Log-distance propagationwith shadow exponent 119899 = 3 [44] plus 3GPP EVA multipathfading modelL119895119894 [34] and is computed by the following

119871119895119894 (119898 119905) = 1198710 + 10119899 log10 (119889119895119894 (119905)1198890 ) +L119895119894 (119898 119905) dB (8)

where 1198710 = 20 log10(120582(41205871198890)) 120582 is the wavelength and1198890 is the reference distance assumed to be 10m in oursimulations Traces for EVA were generated using MATLABwhile modifying the classical Doppler spectrum for each tapat vehicular speed (V) of 40 kmh and 60 kmh as

119878 (119891) prop 1(1 minus (119891119891119863)2)05 (9)

where 119891119863 = V120582 These EVA traces specified by 3GPP in[34] improvise the NLOS conditions relative to the speed

Wireless Communications and Mobile Computing 11

of UE and the transmission bandwidth Simulation of themultipath fading component in ns-3 uses the trace startingat a random point along the time domain for a given user andserving cell The channel object created with the rayleighchanfunction is used for filtering a discrete-time impulse signal inorder to obtain the channel impulse response The filteringis repeated for different Transmission Time Interval (TTI)thus yielding subsequent time-correlated channel responses(one per TTI) This channel response is then processedwith the pwelch function for obtaining its power spectraldensity values which are then saved in a file with the formatcompatible with the simulator model

32 SAI Implementation Model Archer and Vogel in [45]carried out a survey on traffic safety problems in urbanareas It was concluded that overtaking tends to occur lessfrequently within urban areas where the speed limit isless than 50 kmh Lane changing however occurs quitefrequently within urban areas due to low speed limits Thenumber of rear-end accidents is greater within urban areasthan in rural areas and similarly due to higher level ofcongestion and higher number of traffic junctions there isa greater number of opportunities for turning and crossingaccidents within urban area According to another survey of4500 crashes by the Insurance Institute of Highway Safety[46] 13 of accidents are during lane change maneuvers12 are intersection collisions 22 are due to traffic lightviolation 9 are head on collisions 18 include left turningcrashes 12 are due to emergency vehicles parked on blindturns and 14 are due to over speeding In the light ofthese statistics for our adopted urban environment the SAIs(Table 2) we utilize in our simulations are 1 (25) 3 (22) 5(9) 6 (18) 7 (12) and 9 (14)

33 Background Traffic In order to model close to realscenarios we added background traffic into our simulationsModeling of such traffic is done bymaking 50of the existingvehicles use voice application and 25 stream videos Voiceapplication is modeled by implementing constant bit rate(CBR) traffic generator on G711 codec using onoffapplicationon ns-3 [37] Voice payload is 172 bytes and the data ratechosen is 688 Kbps The reason for using a high data rate isto cater worst case scenario where an operator would preferhaving a high quality voice application At the same timevideo streaming is being carried out using evalvid simulatormodule on H263 codec with a payload of 1460 bytes at a datarate of 122Mbps [47] Both these voice and video applicationsare assigned their respective QCIs

34 Performance Measures The primary performance mea-sures used are the end-to-end packet delay the packet deliveryratio and the downlink goodputThe goodput is defined as thenumber of useful information bits received at the applicationlayer per unit of time

In addition we use the notation |F119894| to represent theaverage number of neighbors calculated over the number ofreceivers whenever a beacon is disseminated The end-to-enddelay of a beacon (119863119894119896) is defined as the time measured fromthe start of its transmission at vehicle 119894rsquos application layer until

its successful reception at vehicle 119896rsquos application layer PDR isdefined as the beacon success rate within the awareness range119877 [48] that is for a beacon sent fromvehicle 119894 its PDR is givenby

119875119894 (119877) = sum119896isinF119894

1198681198941198961003816100381610038161003816F1198941003816100381610038161003816 1003816100381610038161003816F1198941003816100381610038161003816 gt 0 (10)

where | sdot | indicates the cardinality of the set and 119868119894119896 (isin 0 1)is set to 1 if the beacon is received by vehicle 1198964 Simulation Results

The success of cellular networks relies on the scalability ofthe system capacity achieved by reusing the spectrum Thecommon approach implemented by operators is to caterthe capacity needs as the traffic demand grows Thereforefirst we analyze the capacity limit which results from thesingle cell case compared to multiple cells using the Friisradio propagation environment and afterwards we analyzethe impact of the urban radio environment as a function ofthe awareness range

Figure 7(a) shows the cumulative distribution function(CDF) of the end-to-end delay when the network is com-posed of 100 vehicles moving at an average speed of 40 kmhThe service area is covered using a single cell with an isotropicantenna located at the upper left corner of Manhattan Gridmodel as utilized by [11] in a Friis radio environment wherevehicles transmit at 1 Hz beacon frequency We observe thatfor awareness range of up to 500m the end-to-end delay isle100ms with around 95 probability Probability of higherend-to-end delay increases more between 250m to 500mas compared to the increase from 500m to 750m That ismainly the result of the increase in traffic but also as theawareness range increases cars located on roads near the edgeof the Manhattan model will have neighbors towards the celldirection with better signal quality which is further exploitedby the scheduling algorithm Figure 7(a) also shows that thereis no significant increase in end-to-end delay from 750mto 1000m This is due to the border effect incurred by thecar located at the edge of the model having less number ofneighbors when increasing the awareness range as comparedto the cars located at the center of the cell Nevertheless theseresults show that up to 100 vehicles were possible for thesingle cell case at a beacon frequency of 1Hz under Friis radioenvironment In order to analyze the intercell interferenceimpact on the delay while increasing the number of vehiclesand beacon frequency we increase the network capacity bydeploying 4 sites with 3 sectorssite (the average number ofvehicles per sector is reduced to 125) Figure 7(b) shows thatthe end-to-enddelay is then further improved to below 50mswith 91 success rate when awareness range of 1000m at thebeacon frequency of 1Hz was utilized We can also noticethat in this case there are some beacons that experiencedlonger delays compared to the single cell scenario causedby low quality conditions of cell-edge users due to intercellinterference and handover However with 1000m 934 ofthe forwarded beacons were received with end-to-end delayle 100ms and 982 for 250m

12 Wireless Communications and Mobile Computing

0 50 100 150 200 250 3000

01

02

03

04

05

06

07

08

09

1CD

F

End-to-end delay (ms)

R = 100

R = 250

R = 500

R = 750

R = 1000

(a)

0

01

02

03

04

05

06

07

08

09

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 1

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 35

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 14

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 25

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 41

(b)

0

01

02

03

04

05

06

07

08

09

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 1

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 35

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 14

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 25

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 41

(c)

Figure 7 LTE end-to-end delay for 100 vehicles at a beacon frequency 1Hz and average speed 40 kmh with Friis radio environment (a)single cell (b) multicell (4 sites 3 sectorssite) (c) multicell with EVA fading radio environment

The results in Figure 7(b) show that under the Friis radiopropagation environment the capacity increases and thesystem performance is not significantly affected in terms ofthe end-to-end delay in presence of intercell interference andhandover since the scheduler effectively managed interfer-ence by time and frequency allocations However in an urbanenvironment it is expected that fading caused by buildingsand Doppler shift due to the relative velocity of vehicles havea very important impact on the received signal quality [49]Figure 7(c) shows the results when the radio propagationenvironment was changed to Log-distance-dependent and

3GPP EVA under the same site configurations Under thisradio environment the probability for the end-to-end delayto be less than or equal to 100ms was around 76ndash78 forawareness range between 100m and 500m 70 for 750mand 66 for 1000m Hence the significant effect on the end-to-end delay is because of path loss and frequency selectivefading imposed by the EVA fading environmentThe end-to-enddelay for beacons to and fromcell-edge users significantlyincreases Similarly for inner cell users the end-to-end delayremained close to the Friis results and for up to 500m theend-to-end delay was le50ms with 74 to 71 probability

Wireless Communications and Mobile Computing 13

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 17

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 53

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 20

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 37

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 62

(a)

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 11

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 69

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 21

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 47

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 73

(b)

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 13

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 63

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 24

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 47

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 74

(c)

Figure 8 End-to-end delay for sample networks at average speed 40 kmh with 150 vehicles at beacon frequency (a) 1Hz with MG model(b) 1Hz with GCC model (c) 10Hz with GCC model

65 for 750m and 59 for 1000m Therefore it can bededuced that with the addition of multipath fading alongwith critical radio propagation environment probability ofdelay being le50ms decreases by 19 to 30

Continuing on with multicell and EVA fading radioenvironment by increasing the network size the number ofvehicles in the awareness range increases and the end-to-enddelay increases as shown in Figure 8(a) Nevertheless theend-to-end delay remained below 100ms for up to 250mwith82 probability for 100 vehicles and similar rate was obtainedwith up to 150 vehicles When the awareness range increasesto 500m the probability was 80 for 100 vehicles while for150 vehicles further reduces to 73 For awareness range of750m and 1000m the probability reduced to 75 and 71 for100 vehicles while for 150 vehicles reduces to 69 and 60respectively

Changing the simulation model along with mobilitytraces to Glasgow city center the effect on delay performancewith 150 vehicles at 1 Hz and 10Hz is shown in Figures 8(b)and 8(c) respectively It is interesting to observe the effect ofchanging the mobility model to a European city Where each

city block is 400m in theMGmodel average city block size inGCC is around 200m increasing vehicle density eventuallygrowing the average number of neighbors for GCC modelThis can be observed for an awareness range of 250m theprobability of delay to be le50ms decreases from 77 to 68while for 1000m this deterioration goes from 54 to 37hence an overall decrease of about 9 to 17 for variousawareness ranges Similarly in Figure 8(c) it is observedthat changing the beacon frequency from 1Hz to 10Hz withan awareness range of 500m the probability of delay to bele50ms drastically decreases from 55 to 22 Whereas foran awareness range of 1000m degradation of about 29 isobserved Performance with a high beacon frequency suchas 10Hz is overall seen to be poor That is because packetsthat arrive late are discarded and are mainly sent by cell-edgeusers for which their neighborhood very likely includes usersat the cell edge Hence the delay from the uplink is indirectlyused as a spatial filtering of low quality cell-edge users Thissuggests that if the information is not useful when receivedafter 100ms the better use of resources is to drop the packetat the minimum latency requirement by the server

14 Wireless Communications and Mobile Computing

0010203040506070809

1

0 20 40 60 80 100 120 140 160 180 200

CDF

End-to-end delay (ms)

SAI = 11003816100381610038161003816Fi

1003816100381610038161003816 = 19

SAI = 31003816100381610038161003816Fi

1003816100381610038161003816 = 16

SAI = 51003816100381610038161003816Fi

1003816100381610038161003816 = 41

SAI = 61003816100381610038161003816Fi

1003816100381610038161003816 = 14

SAI = 71003816100381610038161003816Fi

1003816100381610038161003816 = 56

SAI = 91003816100381610038161003816Fi

1003816100381610038161003816 = 54

Without SAI 1003816100381610038161003816Fi1003816100381610038161003816 = 16

(a)

0010203040506070809

1

0 20 40 60 80 100 120 140 160 180 200

CDF

End-to-end delay (ms)

SAI = 11003816100381610038161003816Fi

1003816100381610038161003816 = 26

SAI = 31003816100381610038161003816Fi

1003816100381610038161003816 = 2

SAI = 51003816100381610038161003816Fi

1003816100381610038161003816 = 35

SAI = 61003816100381610038161003816Fi

1003816100381610038161003816 = 22

SAI = 71003816100381610038161003816Fi

1003816100381610038161003816 = 78

SAI = 91003816100381610038161003816Fi

1003816100381610038161003816 = 94

Without SAI 1003816100381610038161003816Fi1003816100381610038161003816 = 21

(b)

Figure 9 LTE end-to-end delay at average speed 40 kmh (GCC) and (a) 100 vehicles and (b) 150 vehicles

Next we observe and compare the results after theproposed algorithm implementation Figure 9 shows theCDF of the end-to-end delay for 100 and 150 vehicles atan average speed of 40 kmh with and without the imple-mentation of SAI algorithm The scenario being comparedhas the awareness range set to 500m and beacon frequencyto 10Hz We observe a significant decrease in the end-to-end delay after SAI algorithm implementation Probabilityfor the end-to-end delay to be less than 100ms is above80 in both scenarios whereas without the algorithm thisprobability is below 70 for 100 vehicles (Figure 9(a)) andbelow 60 for 150 vehicles (Figure 9(b)) The variation inthis delay between various SAIs is because of the differencein transmission parameters Larger awareness range leadingto higher number of neighbors |F119894| and higher beaconfrequency results in more congestion and large queuingtimes The reason for better delay values with SAI algorithmis mainly the result from restricting resources to the requiredtransmission parameters for safety applicationsmentioned inSection 32

Figure 10 shows the aggregate downlink goodput for thesame scenario investigated for end-to-end delay After theimplementation of SAI algorithm the downlink goodput for100 vehicles dropped from 746Mbps to 622Mbps and for150 vehicles it dropped from 2167Mbps to 1376Mbps Thissignificant decrease in the downlink direction is again due tothe restriction of unnecessary data dissemination from theVSA server to vehicles Eventually this decrease of goodputresults in less load on the LTE system The increase in thedownlink goodput for the scenario without SAI explains thehigh end-to-end delay observed in Figure 9 showing a burstof packet in the downlink leading to waiting time in thequeue

Furthermore the impact of regular cellular traffic on ourLTE vehicular network is studied in order to validate ourfindings and algorithm A broad picture of our observa-tions under different scenarios is shown in Figure 11 whereprobability of end-to-end delay being less than or equal to50ms is studied As expected this probability decreases byabout 3-4 when background traffic is added However with

100 150

Number of vehicles

25

20

15

10

5

0

Dow

nlin

k go

odpu

t (M

bps)

Without SAIWith SAI

746622

2167

1376

Figure 10 Downlink goodput for 100 and 150 vehicles at an averagespeed 40 kmh (GCC)

SAI algorithm probability of end-to-end delay being lessthan 50ms stays above 90 even after adding backgroundtraffic

Finally we include the SAI into MAC layer PDU con-trol elements to prove our concept of D2D like unicasttransmission between sender and receivers Modeling of thesystem is carried out in accordance with the description inSection 22 Figure 12 shows end-to-end delay for 100 and150 vehicles with and without the inclusion of SAI in MAClayer It is evident that 150 vehicles experience almost thesame probability for end-to-end delay being less than 50msas experienced by 100 vehicles This shows that with the useof MAC layer control elements around 50 more vehiclescan be accommodated by LTE network increasing the spec-trum efficiency eventually leading to higher capacity of thesystem

Wireless Communications and Mobile Computing 15

Simple case With backgroundtraffic

With SAI With backgroundand SAI

6065707580859095

100

150 vehicles100 vehicles

Prob

abili

ty o

f end

-to-e

ndde

layle

50

ms

Scenarios

Figure 11 Probability of end-to-end delay being le50ms for 100 and150 vehicles at 40 kmh in various scenarios

With SAI in APP layer With SAI in MAC layer80828486889092949698

100

100 vehicles150 vehicles

p

roba

bilit

y fo

r end

-to-e

ndde

layle50

ms

Figure 12 Probability of end-to-end delay being le50ms for 100 and150 vehicles with and without MAC layer inclusion

5 Conclusion

This paper proposes the safety application identifier conceptin the form of an algorithm that is tested and analyzedwithin the impact of vehicular urbanmulticell radio environ-ment employing multipath fading channels and backgroundtraffic The proposed algorithm uses dynamic adaptation oftransmission parameters in order to fulfill stringent vehicu-lar application requirements while minimizing latency andreducing system load

With the help of extensive system-level simulations thecrucial impact of channel modeling and mobility traces isevident with its influence on the received signal qualitydecreasing the probability of end-to-end delay to be le50msfor various awareness ranges by 19 to 30 and 9 to17 respectively Studying the impact of awareness rangeit is observed that with range up to 1000m for 1Hz 500mfor 2Hz and 250m for 10Hz beacon frequency the systemperformed well With the help of these findings the SAIalgorithm is proposed and implemented Probability of expe-riencing an end-to-end delay of less than 100ms increasedby around 20 and the downlink goodput decreased signif-icantly from 2167Mbps to 1376Mbps for 150 vehicles

Furthermore the impact of having background traffic isalso taken into consideration Results show that the systemis slightly affected by having regular background trafficHowever the probability of end-to-end delay being le50ms

still stays satisfactorily above 85 with the use of SAI Thispaper also proposes the inclusion of SAI inMAC layer controlelements producing a D2D type of communication withinvehicles After modeling this system it was found that thesame amount of system resources and requirements are metwith additional 50 vehicles increasing the capacity of ourLTE vehicular network Finally in future works we plan tointegrate SAI incorporated within MAC layer in a hetero-geneous network with DSRC while exploring the networkslicing in 5G cellular networks for vehicular communicationscomplementing each otherrsquos weaknesses and strengths inorder to meet the vehicular network requirements

Conflicts of Interest

The authors declare that they have no conflicts of interest

Acknowledgments

Marvin Sanchez would like to acknowledge the support ofLAMENITEC Erasmus Mundus Program of the EuropeanUnion to be a guest researcher at GCU Results were obtainedusing the EPSRC funded ARCHIE-WeSt High PerformanceComputer (httpswwwarchie-westacuk) EPSRC Grantno EPK0005861

References

[1] S Sesia I Toufik and M Baker LTE the UMTS long termevolution from theory to practice Wiley Publishing 2009

[2] H Hartenstein and K P Laberteaux ldquoA tutorial survey onvehicular ad hoc networksrdquo IEEE Communications Magazinevol 46 no 6 pp 164ndash171 2008

[3] D Jiang and L Delgrossi ldquoTowards an international standardfor wireless access in vehicular environmentsrdquo in Proceedings ofthe 67th IEEE Vehicular Technology Conference (VTC rsquo08) pp2036ndash2040 May 2008

[4] S Al-Sultan M M Al-Doori A H Al-Bayatti and H ZedanldquoA comprehensive survey on vehicular Ad Hoc networkrdquoJournal of Network and Computer Applications vol 37 no 1 pp380ndash392 2014

[5] K Tanuja T Sushma M Bharathi and K Arun ldquoA surveyon VANET technologiesrdquo International Journal of ComputerApplications vol 121 no 18 2015

[6] H Y Kim D M Kang J H Lee and T M Chung ldquoA per-formance evaluation of cellular network suitability for VANETrdquoWorld Academy of Science Engineering and Technology Interna-tional Science Index vol 64 no 6 pp 124ndash127 2012

[7] A Fasbender M Gerdes and S Smets ldquoCellular NetworkingTechnologies in ITS Solutions Opportunities and Challengesrdquopp 1ndash13 Springer Berlin Germany 2012

[8] ldquoCellular coverage and tower map for ee network in glasgowrdquohttpswwwcellmappernetmap

[9] GAraniti C CampoloMCondoluci A Iera andAMolinaroldquoLTE for vehicular networking a surveyrdquo IEEECommunicationsMagazine vol 51 no 5 pp 148ndash157 2013

[10] A Vinel ldquo3GPP LTE versus IEEE 80211pWAVE which tech-nology is able to support cooperative vehicular safety applica-tionsrdquo IEEE Wireless Communications Letters vol 1 no 2 pp125ndash128 2012

4 Wireless Communications and Mobile Computing

Table 2 Safety application requirements and SAIs

Safety application Message type Transmitted data Criticallatency

Transmissionfrequency BF

Transmissionrange 119877 SAI

Intersection collisionwarning CAM Vehicle type position heading velocity

acceleration yaw rate sim100ms 10Hz 150m

1Lane change assistance CAM Position heading velocity accelerationturn signal status sim100ms 10Hz 150m

Cooperative forwardcollision warning CAM Vehicle type position heading velocity

acceleration yaw rate sim100ms 10Hz 150m

Slow vehicle indication CAMDENM Vehicle type position headingacceleration velocity sim100ms 2Hz 200m 2

Traffic light speedadvisoryviolation CAMDENM Signal phase timing position direction

road geometry sim100ms 2Hz 150m3

Motorcycle approachingindication CAM Vehicle type position heading velocity sim100ms 2Hz 150m

Overtaking vehiclewarning CAM Position velocity yaw rate acceleration sim100ms 10Hz 300m 4

Head on collisionwarning CAM Vehicle type position heading velocity

acceleration yaw rate sim100ms 10Hz 200m 5

Collision risk warning CAMDENM Vehicle type position heading velocityacceleration yaw rate sim100ms 10Hz 300ndash500m 6

Emergency vehiclewarning CAMDENM Position heading velocity acceleration sim100ms 2Hz 300m 7

Cooperative mergingassistance CAM Curve location curvature slope speed

limit surface sim1000ms 1Hz 250m 8

Speed limits notification CAM Velocity acceleration position speedlimit heading sim100ms 1ndash10Hz 300m 9

Message (DENMs) [27] CAMs are periodic messages thatprovide information like the presence position and sensorinformation and are expected to be received by all thevehicles within the awareness range DENMs are used forcooperative road hazard warnings that are broadcasted intheir relevance area whenever a hazardous event occurs [32]The work in [33] proposes the use of dynamic adaptation ofcommunication radius in order to maximize sum rate weadopt the same concept and apply it to dynamically adaptingcommunication range in order tominimize end-to-end delayFigure 1 illustrates the awareness range119877 for vehicle 119894 Noticethat vehicle 119896 is within the awareness range of 119894 since thedistance 119889119894119896 le 1198772 Safety Application Identifier Algorithm

Message delivery and reliability are a major concern in vehic-ular networks due to which a differentiated QoS mechanismis proposedThismechanismworks on the principle of index-ing various applications according to their requirementsmotivated from QoS class identifier (QCI) implemented inLTE networks [34] In Table 2 each application fromTable 1 iscategorized and assigned an SAIThis SAI which is a numberranging between 1 and 9 is included in the transmitted packetappended before the IP header by the application layer asshown in Figure 2 The categorization and assignment of SAIfor ourmodel are further discussed in Section 3This conceptis implemented in the form of an algorithm proposed in thefollowing subsection

Input VSMs Vehicles rarr ServerOutput VSMs Server rarr Vehicles

System Setup(1) while Server larr VSM119894 do(2) Server locates all vehicles(3) VSM119894 rarr (SAI119894Position (119889119896))(4) SAI119894 rarr (119877119894BF119894)(5) (119877119894 119889119896) rarr Distance (119889119894119896)(6) F119894 = forall119896 119889119894119896 lt 119877119894 119894 = 119896(7) Server rArr Vehicles isin F119894 at BF119894(8) end while(9) return F119894

Algorithm 1 SAI Algorithm

21 SAI Algorithm With the concept of safety applicationindexing we propose an algorithm that does not includethe use of group formation or IEEE 80211p The proposedalgorithm is a process that is carried out on the vehicularsafety application (VSA) server elaborated in Algorithm 1When the vehicles start transmissions the packet is carriedfrom the eNodeB to the VSA server via the CN This packetcontains vehicle location (119889119896) and corresponding vehicularsafety message (VSM) As soon as the server starts receivingthese messages it locates all the vehicles being served bythe network forming a virtual geographical map of theserved area The VSA server while receiving the packets

Wireless Communications and Mobile Computing 5

S-GWP-GWVSA

Server

Awareness Range

VSM

S-GW serving gatewayP-GW PDN gatewayVSA vehicular safety application

idik

k

R

Position dkSAIi (Ri BFi)

Figure 1 LTE multicell system and awareness range

will also extract the SAI information from the receivedpacket This SAI information will then be checked againstthe database stored in the server (Table 2) With this SAIthe server retrieves the transmission beacon frequency BF119894and awareness range 119877119894 requirement of the application thatvehicle 119894 is utilizing Using the required awareness range theserver determines the forwarding set of vehicles (neighboringvehicles) by

F119894 = forall119896 119889119894119896 lt 119877119894 119894 = 119896 (1)

where 119889119894119896 is the distance from vehicle 119894 to the neighboringvehicle 119896 Recalling the concept of data freshness defined by[15] the VSA server will discard the packet if it arrives laterthan the beacon inter arrival time (1BF119894)22 MAC Layer Inclusion This work proposes the inclusionof SAI from application layer to MAC Protocol Data Unit(PDU) control elements This inclusion in MAC layer bringsthe processing towards the base station decreasing latencyand reducing the vehicular traffic load on themobile network

In cellular networks eNB uses the cell radio networktemporary identifier (C-RNTI) to uniquely identify UEs Asshown in Figure 3 a unique C-RNTI is assigned to everyuser by the eNB during the initial random access procedurewhich is used for identifying the radio resource control(RRC) connection and for scheduling purposes Coding anddecoding of physical downlink control channel (PDCCH)for a specific UE is based on its C-RNTI UEs initiatecontention based access to the network by transmitting apreamble sequence on the physical random access channel

(PRACH) to which the eNB responds with a C-RNTI onthe physical downlink shared channel (PDSCH) In vehicularcommunications since the awareness range is a function ofvehiclersquos geographical location if the transmitterrsquos location isunknown to the network an additional message containingtransmitterrsquos location is required

So for our proposed system model signaling shown inFigure 3 the transmitter registers with the network in thesame way as discussed above and it includes its location inthe RRC connection request message in its establishmentcause as mo-Data transmitted via the physical uplink sharedchannel (PUSCH) This location is introduced in the RRCconnection request message as a new information elementin line with the 3GPP specifications for RRC connectionrequest [21] At the same time all the UEs already registeredto the network update the eNB of their location using thenetwork assisted global navigation satellite system (GNSS)location update message as specified in [35] Once thevehicle has a beacon ready for transmission it sends outa scheduling request (SR) in the physical uplink controlchannel (PUCCH) The eNB replies back in the downlinkwith the grant for sending BSR After receiving this grantthe vehicle will send out the BSR in PUSCH containing theMAC PDU SAI will be in the unused 16-bit long MACControl Element inside the BSR This element utilizes spacecurrently reserved for future use and is indexed in the MACPDU subheader by the logical channel ID (LCID) value equalto 01011 [36] This new element now referred to as SAIis appended to the existing LCID values such as commoncontrol channel (CCCH) C-RNTI and the padding as shownin Figure 4 The eNB will then use the SAI to determine the

6 Wireless Communications and Mobile Computing

PayloadSAIIP Header

PDCP SDUPDCP header

RLC SDURLC header

MAC SDUMAC header

Transport block CRC

Location coordinates + safety message

PayloadIP SAI

Vehicular safety message

Inclusion of SAI

Header compression

PDCP

RLC

Ciphering

Segmentation

MAC

Multiplexing

PHYOne TB per TTI (1 ms)

Figure 2 SAI inclusion in LTE message structure

transmission parameters allotting transmission resources tothe sender and receiver vehicles

23 Scalability of the System Kato et al [15] emphasizedthe significance of placing the VSA server close to the edgeof LTE network It can also be argued if a single serverwould be able to serve other locations that is multipleeNodeBs or more servers would be required giving rise tothe question of system scalability For the implementationof SAI algorithm at the VSA server the location would playa vital role Considering the data freshness concept ideallocation of the server would be at the EPCTherefore havingmultiple EPCs serving their respective geographical locationswould have their own VSA servers with a similar approachas is for mobility management entity (MME) reducing theRTT while increasing the system capacity and eventuallymeeting the strict transmission requirements for vehicularsafety applications

In the case of having SAI appended in MAC layer theimplementation of SAI algorithm is brought towards theeNodeB reducing the RTT and eventually increasing thesystem capacity For vehicles being served by neighboringeNodeBs the message would be forwarded by the eNodeBserving the source vehicle to the respective neighboring

eNodeB via the X2 interface However if the neighboringeNodeB does not have an X2 interface or is residing in adifferent EPC then the VSA server would come into play Forsuch a scenario the centralization of the server will again besignificant

3 System Model

The network is assumed to be composed of 119873 vehiclesuniquely identified by their number 119894 (119894 = 1 sdot sdot sdot 119873) Vehiclesare assumed to use FDD LTE transceivers with 2 times 10MHzbandwidth uplink carrier frequency 1715MHz and downlinkcarrier frequency 2115MHz (band 4) [34 Table 55-1] Werefer to vehicles as LTE UEs

The work in this paper builds on to the performanceevaluation from [11] adding multicell and multipath alongwith Extended Vehicular A (EVA) fading environment Thenetwork modeled is a 2 times 2 km2 area of Manhattan Grid(MG) and Glasgow city center (GCC) shown in Figures5 and 6 respectively implemented in ns-3 [37] Mobilityof the vehicles in the network is generated using routesmobility model which assigns each vehicle with a routegenerated using Google maps API [14] Along with themobility model the site configuration is also changed for

Wireless Communications and Mobile Computing 7

Vehicle 1 eNB Vehicle 2 Vehicle N

Initial networkaccess

Registered to network

Registered to network

PRACH RA preamble

PDSCH RA responseAssigns C-RNTI 1

PUSCH RRC connection request

PDSCH connection setup

PUSCH RRC connection setup complete

Contains eUE 1location

Assuming norandom access

contention

Network assisted GNSS location update

Network assisted GNSS location update

Beacon for transmission

PUCCH (SR)

PDCCH UL grant for BSR

PUSCH (BSR containing MAC PDU)Contains SAI

Resource allocation

PDCCH TX grant PDCCH RX grant PDCCH RX grant

PUCCH (ACKNACK)PHICH (ACKNACK)

PUSCH beacon transmitted to nUE 2 middot middot middot nUE N

Figure 3 LTE signaling with MAC control elements in BSR

Header Controlelements Payload (SDU) Padding

Subheader1

Subheader2

SubheaderK

MAC Ctrlelement 1

MAC Ctrlelement 2

MAC Ctrlelement N

R ER LCID = 01100 Safety application identifier

MAC PDU

Figure 4 SAI in MAC control elements

8 Wireless Communications and Mobile Computing

500m

S-GWP-GWVSA server

400m

500m1 km

500m

Figure 5 MG model covered by 4 sites and 3 cellssite with 1000mintersite distance (ISD)

GCC This change in the network design considers a morerealistic network model employing the site configurationused by UKrsquos mobile operator EE in Glasgow [8] The LTEfunctionality is implemented through the LTE EPC networksimulator (LENA) module [38] that includes highly detailedeNB and UE functionalities as specified by 3GPP standardsTo manage handover and Internet connections eNBs areinterconnected through their X2 interfaces and connected tothe EPC through S1-U interface Interconnection from the P-GW to the VSA server residing in the EPC is modeled usingan error free 10 Gbps point-to-point link and TCPIP version4

The granularity of the LENA module is up to resourceblock (RB) level including the user plane control planeand reference signals used for coherent reception The safetymessages are UDP packets and often might be segmentedat the radio link control (RLC) and MAC layers to betransmitted on the transport shared channel (S-CH) passedto the physical (PHY) layer The size of transport blocks(TB) used at the PHY layer is decided by the MAC sublayerdepending on the channel quality and scheduling decisionsand may be transmitted using several RBs utilizing a mod-ulation and coding scheme (MCS) [38] Transmissions arescheduled every 1ms corresponding to one subframe usingProportional Fair Algorithm which is one of the approachessuitable for applications that include latency constraints [1Chapter 12] For downlink the algorithm utilizes userrsquoschannel quality indicator (CQI isin [0 sdot sdot sdot 15]) measurementreports sent from the vehicles For uplink channel qualitymeasurements are done by the eNB through scheduling

periodic vehicle transmissions of sounding reference signals(SRS)The radio resource control (RRC)module in LENAns-3 assigns the periodicity of SRS as a function of the numberof UEs attached to an eNB according to 3GPP UE-specificprocedure [37]

Each site modeled by an eNB uses 10W (40 dBm) totaltransmission power and a cosine antenna sector model with65∘ half power beamwidth (HPBW) [39] and azimuthaldirection 0∘ 120∘ or 240∘ with respect to north and 1mintersector space We assume utilization of 2 times 10MHzbandwidth whichmeans there are119872 = 50RBs for the uplinkand for the downlink Therefore assuming that the radiopropagation channel is constant in subframe 119905 if vehicle 119894 isscheduled to receive from its serving cell 119895 on resource block119898 isin [0 sdot sdot sdot 49] the received signal power 119875119895119894 can be modeledas follows

119875119895119894 (119898 119905) = 119875119895 (119898 119905) 119866119895 (120579119895119894)119866119894 (120579119894119895)119871119895119894 (119898 119905) (2)

where 119875119895(119898 119905) is the transmitted power for resource block119898119866119895(120579119895119894) is the antenna gain of cell 119895 in the direction of user 119894119866119894(120579119894119895) is the antenna gain of user 119894 in the direction of cell 119895and 119871119895119894 is the path loss from cell 119895 to user 119894

In the single cell scenario there is no intercell interferenceand mutually exclusive RBs are assigned to users Howeverin a more general deployed urban environment the availablespectrum is utilized by multiple cells in the service area andthe intercell interference is one important limiting aspectespecially for cell-edge users [1 Chapter 12]

In order to allocate usersrsquo transmissions while at the sametime maximizing the signal to interference plus noise ratio(SINR) for each RB frequency reuse and handover are usedFrequency allocation has an impact on the achievable systemcapacity and delay since the transport block size MCS andthe transmission schedule are dependent on the SINR [37Chapter 9] In ns-3 if several RBs are used to transmit a TBthe vector composed of the SINRs on each RB is used toevaluate the TB error probability The SINR in RB 119898 whenuser 119894 is scheduled to receive from serving cell 119895 can bewrittenas follows

SINR119894 (119898 119905)= 119875119895 (119898 119905) 119866119895 (120579119895119894)119866119894 (120579119894119895) 119871119895119894 (119898 119905)1198730 + sumforall119897 =119895 (119875119897 (119898 119905) 119866119897 (120579119897119894) 119866119894 (120579119894119897) 119871 119897119894 (119898 119905))

(3)

where1198730 is the noise powerUEs report the SINR mapped to the CQI value corre-

sponding to the MCS (isin [0 sdot sdot sdot 28]) that ensures TB error ratele 01 according to the standard [40 Section 723] In ns-3 thedownlink SINR is used to generate periodic wideband CQIsfeedback (ie an average value of all RBs) and inband CQIs(ie a value for each RB) We set the CQI to be calculated bycombining the received signal power of the reference signalsto the interference power from the physical downlink sharedchannel as an approximation to (3)

For dynamic frequency allocation we use the distributedfractional frequency reuse algorithm that utilizes intercell

Wireless Communications and Mobile Computing 9

S-GWP-GWVSA

TCC

Figure 6 GCC model covered by 6 sites and 3 cellssite where eNB location is determined using UK operator EE mast data [8]

interference coordination (ICIC) [41] The eNB adaptivelyselects a set of RBs referred to as cell-edge subband (set to10RBs) based on information exchanged with its neighborsTherefore according to ICIC a vehicle should be allocatedin the cell-edge subband if its serving cell reference signalreceived quality (RSRQ) gets lower than minus10 dB

Handover decisions are also performed by eNBs based onconfigurable event-triggeredmeasurement reports includingthe RSRQ and the reference signal received power (RSRP)The measurements are performed by UEs usually in orthog-onal frequency division multiplexing (OFDM) symbols car-rying reference signals (RS) for antenna port 0 In ns-3 since the channel is assumed flat over an RB that iscomposed of 12 subcarriers all resource elements (RE) in anRB have the same power Hence assuming orthogonal RSreception the RSRP for cell 119895 measured by vehicle 119894 is givenby

RSRP119895119894 (119905) = sum119872minus1119898=0 sum119870minus1119896=0 (119875119895119894 (119898 119896 119905) 119870)119872= sum119872minus1119898=0 (119875119895119894 (119898 119905) 12)119872

(4)

where 119875119895119894(119898 119896 119905) is the received power of RS 119896 in RB 119898 119872is the number of RBs and 119870 is the number of RS measuredin the RB The average value is passed to higher layers every200ms therefore the index 119905 is omitted for simplicity

The RSRQ for serving cell 119895measured by vehicle 119894 can becomputed by the following [42]

RSRQ119895119894 = 119872 times RSRP119895119894RSSI119895119894

(5)

where RSSI119895119894 is the received signal strength indicator (RSSI)measured by UE 119894when served by cell 119895 and according to the

10 Wireless Communications and Mobile Computing

Table 3 LTE simulation parameters

Parameter ValueSimulation time 100 seconds

Road model Manhattan grid model (MG) (two way roads)Glasgow City Center (GCC) (2 km times 2 km)

LTE network 4 sites with 3 cellssite 1000m ISD for MG6 sites with 3 cellssite UK Operator EE mast data [8]

Transmission power eNB 40 dBm UE 23 dBmCarrier frequency DLUL 2115MHz1715MHzChannel bandwidth 2 times 10MHz (2 times 50RBs)Noise Figure eNB 5 dB UE 9 dBUE antenna model Isotropic (0 dBi)eNB antenna model 15 dB Cosine model 65∘ HPBWScheduling algorithm Proportional Fair

Handover algorithm A2A4RSRQ RSRQ threshold minus5 dBand NeighbourCellOffset = 2 (1 dB)

Frequency reuse Distributed Fractional Freq Reuse

Path loss model LogDistance (120572 = 3) and3GPP Extended Vehicular A model

Safety message format 256 bytes UDPNumber of vehicles 75 100 125 150Average vehiclersquos speed 20 kmh 40 kmhBeacon frequency 1Hz 10HzAwareness range (119877) 100m 250m 500m 750m 1000m

standard comprises the linear average of the total receivedpower observed by the UE from all sources includingcochannel serving and nonserving cells adjacent channelinterference thermal noise and so forth [42] The RSSImeasurement model in ns-3 is computed considering two REper RB (the one that carries the RS) [37] that is

RSSI119895119894 (119905)= 119872minus1sum119898=0

2 times (119873012 +sumforall119897

119875119897 (119898 119905) 119866119897 (120579119897119894) 119866119894 (120579119894119897)12 times 119871 119897119894 (119898 119905) ) (6)

Furthermore in order to perform handover decisions thereported RSRQ for neighboring cell 119897 when vehicle 119894 is con-nected through serving cell 119895 is computed by the following

RSRQ119897119894 = 119872 times RSRP119897119894RSSI119895119894

(7)

The RSRQs for serving and neighboring cells are usedwith theA2A4RsrqHandoverAlgorithm to trigger LTE eventsA2 and A4 [43 Section 554] Event A2 is set to be triggeredwhen the RSRQ of the serving cell becomes worse than minus5 dB(ServingCellThreshold parameter) While event A2 is trueevent A4 is triggered if a neighboring cell becomes betterthan a threshold which is set by the algorithm to a verylow value for the trigger criteria to be true Thus the mea-surement reports received by the eNB are used to consider aneighboring cell as a target cell for handover only if its RSRQ

is 1 dB higher than the serving cell (NeighbourCellOffsetparameter) Table 3 summarizes the simulation parametersfor LTE

31 Vehicular Radio Urban Environment As mentioned inSection 11 previous evaluations lack the use of multipathand multicell environments Use of proper channel modelingis essential in evaluating system models as it poses close toreality system degradations The system model presented inthis paper consists of 6 sites with 3 cells per site Users areassumed to have isotropic antenna (119866119894(120579119894119895) = 0 dBi) and thepath loss (119871119895119894) ismodeled using the Log-distance propagationwith shadow exponent 119899 = 3 [44] plus 3GPP EVA multipathfading modelL119895119894 [34] and is computed by the following

119871119895119894 (119898 119905) = 1198710 + 10119899 log10 (119889119895119894 (119905)1198890 ) +L119895119894 (119898 119905) dB (8)

where 1198710 = 20 log10(120582(41205871198890)) 120582 is the wavelength and1198890 is the reference distance assumed to be 10m in oursimulations Traces for EVA were generated using MATLABwhile modifying the classical Doppler spectrum for each tapat vehicular speed (V) of 40 kmh and 60 kmh as

119878 (119891) prop 1(1 minus (119891119891119863)2)05 (9)

where 119891119863 = V120582 These EVA traces specified by 3GPP in[34] improvise the NLOS conditions relative to the speed

Wireless Communications and Mobile Computing 11

of UE and the transmission bandwidth Simulation of themultipath fading component in ns-3 uses the trace startingat a random point along the time domain for a given user andserving cell The channel object created with the rayleighchanfunction is used for filtering a discrete-time impulse signal inorder to obtain the channel impulse response The filteringis repeated for different Transmission Time Interval (TTI)thus yielding subsequent time-correlated channel responses(one per TTI) This channel response is then processedwith the pwelch function for obtaining its power spectraldensity values which are then saved in a file with the formatcompatible with the simulator model

32 SAI Implementation Model Archer and Vogel in [45]carried out a survey on traffic safety problems in urbanareas It was concluded that overtaking tends to occur lessfrequently within urban areas where the speed limit isless than 50 kmh Lane changing however occurs quitefrequently within urban areas due to low speed limits Thenumber of rear-end accidents is greater within urban areasthan in rural areas and similarly due to higher level ofcongestion and higher number of traffic junctions there isa greater number of opportunities for turning and crossingaccidents within urban area According to another survey of4500 crashes by the Insurance Institute of Highway Safety[46] 13 of accidents are during lane change maneuvers12 are intersection collisions 22 are due to traffic lightviolation 9 are head on collisions 18 include left turningcrashes 12 are due to emergency vehicles parked on blindturns and 14 are due to over speeding In the light ofthese statistics for our adopted urban environment the SAIs(Table 2) we utilize in our simulations are 1 (25) 3 (22) 5(9) 6 (18) 7 (12) and 9 (14)

33 Background Traffic In order to model close to realscenarios we added background traffic into our simulationsModeling of such traffic is done bymaking 50of the existingvehicles use voice application and 25 stream videos Voiceapplication is modeled by implementing constant bit rate(CBR) traffic generator on G711 codec using onoffapplicationon ns-3 [37] Voice payload is 172 bytes and the data ratechosen is 688 Kbps The reason for using a high data rate isto cater worst case scenario where an operator would preferhaving a high quality voice application At the same timevideo streaming is being carried out using evalvid simulatormodule on H263 codec with a payload of 1460 bytes at a datarate of 122Mbps [47] Both these voice and video applicationsare assigned their respective QCIs

34 Performance Measures The primary performance mea-sures used are the end-to-end packet delay the packet deliveryratio and the downlink goodputThe goodput is defined as thenumber of useful information bits received at the applicationlayer per unit of time

In addition we use the notation |F119894| to represent theaverage number of neighbors calculated over the number ofreceivers whenever a beacon is disseminated The end-to-enddelay of a beacon (119863119894119896) is defined as the time measured fromthe start of its transmission at vehicle 119894rsquos application layer until

its successful reception at vehicle 119896rsquos application layer PDR isdefined as the beacon success rate within the awareness range119877 [48] that is for a beacon sent fromvehicle 119894 its PDR is givenby

119875119894 (119877) = sum119896isinF119894

1198681198941198961003816100381610038161003816F1198941003816100381610038161003816 1003816100381610038161003816F1198941003816100381610038161003816 gt 0 (10)

where | sdot | indicates the cardinality of the set and 119868119894119896 (isin 0 1)is set to 1 if the beacon is received by vehicle 1198964 Simulation Results

The success of cellular networks relies on the scalability ofthe system capacity achieved by reusing the spectrum Thecommon approach implemented by operators is to caterthe capacity needs as the traffic demand grows Thereforefirst we analyze the capacity limit which results from thesingle cell case compared to multiple cells using the Friisradio propagation environment and afterwards we analyzethe impact of the urban radio environment as a function ofthe awareness range

Figure 7(a) shows the cumulative distribution function(CDF) of the end-to-end delay when the network is com-posed of 100 vehicles moving at an average speed of 40 kmhThe service area is covered using a single cell with an isotropicantenna located at the upper left corner of Manhattan Gridmodel as utilized by [11] in a Friis radio environment wherevehicles transmit at 1 Hz beacon frequency We observe thatfor awareness range of up to 500m the end-to-end delay isle100ms with around 95 probability Probability of higherend-to-end delay increases more between 250m to 500mas compared to the increase from 500m to 750m That ismainly the result of the increase in traffic but also as theawareness range increases cars located on roads near the edgeof the Manhattan model will have neighbors towards the celldirection with better signal quality which is further exploitedby the scheduling algorithm Figure 7(a) also shows that thereis no significant increase in end-to-end delay from 750mto 1000m This is due to the border effect incurred by thecar located at the edge of the model having less number ofneighbors when increasing the awareness range as comparedto the cars located at the center of the cell Nevertheless theseresults show that up to 100 vehicles were possible for thesingle cell case at a beacon frequency of 1Hz under Friis radioenvironment In order to analyze the intercell interferenceimpact on the delay while increasing the number of vehiclesand beacon frequency we increase the network capacity bydeploying 4 sites with 3 sectorssite (the average number ofvehicles per sector is reduced to 125) Figure 7(b) shows thatthe end-to-enddelay is then further improved to below 50mswith 91 success rate when awareness range of 1000m at thebeacon frequency of 1Hz was utilized We can also noticethat in this case there are some beacons that experiencedlonger delays compared to the single cell scenario causedby low quality conditions of cell-edge users due to intercellinterference and handover However with 1000m 934 ofthe forwarded beacons were received with end-to-end delayle 100ms and 982 for 250m

12 Wireless Communications and Mobile Computing

0 50 100 150 200 250 3000

01

02

03

04

05

06

07

08

09

1CD

F

End-to-end delay (ms)

R = 100

R = 250

R = 500

R = 750

R = 1000

(a)

0

01

02

03

04

05

06

07

08

09

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 1

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 35

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 14

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 25

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 41

(b)

0

01

02

03

04

05

06

07

08

09

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 1

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 35

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 14

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 25

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 41

(c)

Figure 7 LTE end-to-end delay for 100 vehicles at a beacon frequency 1Hz and average speed 40 kmh with Friis radio environment (a)single cell (b) multicell (4 sites 3 sectorssite) (c) multicell with EVA fading radio environment

The results in Figure 7(b) show that under the Friis radiopropagation environment the capacity increases and thesystem performance is not significantly affected in terms ofthe end-to-end delay in presence of intercell interference andhandover since the scheduler effectively managed interfer-ence by time and frequency allocations However in an urbanenvironment it is expected that fading caused by buildingsand Doppler shift due to the relative velocity of vehicles havea very important impact on the received signal quality [49]Figure 7(c) shows the results when the radio propagationenvironment was changed to Log-distance-dependent and

3GPP EVA under the same site configurations Under thisradio environment the probability for the end-to-end delayto be less than or equal to 100ms was around 76ndash78 forawareness range between 100m and 500m 70 for 750mand 66 for 1000m Hence the significant effect on the end-to-end delay is because of path loss and frequency selectivefading imposed by the EVA fading environmentThe end-to-enddelay for beacons to and fromcell-edge users significantlyincreases Similarly for inner cell users the end-to-end delayremained close to the Friis results and for up to 500m theend-to-end delay was le50ms with 74 to 71 probability

Wireless Communications and Mobile Computing 13

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 17

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 53

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 20

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 37

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 62

(a)

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 11

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 69

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 21

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 47

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 73

(b)

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 13

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 63

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 24

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 47

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 74

(c)

Figure 8 End-to-end delay for sample networks at average speed 40 kmh with 150 vehicles at beacon frequency (a) 1Hz with MG model(b) 1Hz with GCC model (c) 10Hz with GCC model

65 for 750m and 59 for 1000m Therefore it can bededuced that with the addition of multipath fading alongwith critical radio propagation environment probability ofdelay being le50ms decreases by 19 to 30

Continuing on with multicell and EVA fading radioenvironment by increasing the network size the number ofvehicles in the awareness range increases and the end-to-enddelay increases as shown in Figure 8(a) Nevertheless theend-to-end delay remained below 100ms for up to 250mwith82 probability for 100 vehicles and similar rate was obtainedwith up to 150 vehicles When the awareness range increasesto 500m the probability was 80 for 100 vehicles while for150 vehicles further reduces to 73 For awareness range of750m and 1000m the probability reduced to 75 and 71 for100 vehicles while for 150 vehicles reduces to 69 and 60respectively

Changing the simulation model along with mobilitytraces to Glasgow city center the effect on delay performancewith 150 vehicles at 1 Hz and 10Hz is shown in Figures 8(b)and 8(c) respectively It is interesting to observe the effect ofchanging the mobility model to a European city Where each

city block is 400m in theMGmodel average city block size inGCC is around 200m increasing vehicle density eventuallygrowing the average number of neighbors for GCC modelThis can be observed for an awareness range of 250m theprobability of delay to be le50ms decreases from 77 to 68while for 1000m this deterioration goes from 54 to 37hence an overall decrease of about 9 to 17 for variousawareness ranges Similarly in Figure 8(c) it is observedthat changing the beacon frequency from 1Hz to 10Hz withan awareness range of 500m the probability of delay to bele50ms drastically decreases from 55 to 22 Whereas foran awareness range of 1000m degradation of about 29 isobserved Performance with a high beacon frequency suchas 10Hz is overall seen to be poor That is because packetsthat arrive late are discarded and are mainly sent by cell-edgeusers for which their neighborhood very likely includes usersat the cell edge Hence the delay from the uplink is indirectlyused as a spatial filtering of low quality cell-edge users Thissuggests that if the information is not useful when receivedafter 100ms the better use of resources is to drop the packetat the minimum latency requirement by the server

14 Wireless Communications and Mobile Computing

0010203040506070809

1

0 20 40 60 80 100 120 140 160 180 200

CDF

End-to-end delay (ms)

SAI = 11003816100381610038161003816Fi

1003816100381610038161003816 = 19

SAI = 31003816100381610038161003816Fi

1003816100381610038161003816 = 16

SAI = 51003816100381610038161003816Fi

1003816100381610038161003816 = 41

SAI = 61003816100381610038161003816Fi

1003816100381610038161003816 = 14

SAI = 71003816100381610038161003816Fi

1003816100381610038161003816 = 56

SAI = 91003816100381610038161003816Fi

1003816100381610038161003816 = 54

Without SAI 1003816100381610038161003816Fi1003816100381610038161003816 = 16

(a)

0010203040506070809

1

0 20 40 60 80 100 120 140 160 180 200

CDF

End-to-end delay (ms)

SAI = 11003816100381610038161003816Fi

1003816100381610038161003816 = 26

SAI = 31003816100381610038161003816Fi

1003816100381610038161003816 = 2

SAI = 51003816100381610038161003816Fi

1003816100381610038161003816 = 35

SAI = 61003816100381610038161003816Fi

1003816100381610038161003816 = 22

SAI = 71003816100381610038161003816Fi

1003816100381610038161003816 = 78

SAI = 91003816100381610038161003816Fi

1003816100381610038161003816 = 94

Without SAI 1003816100381610038161003816Fi1003816100381610038161003816 = 21

(b)

Figure 9 LTE end-to-end delay at average speed 40 kmh (GCC) and (a) 100 vehicles and (b) 150 vehicles

Next we observe and compare the results after theproposed algorithm implementation Figure 9 shows theCDF of the end-to-end delay for 100 and 150 vehicles atan average speed of 40 kmh with and without the imple-mentation of SAI algorithm The scenario being comparedhas the awareness range set to 500m and beacon frequencyto 10Hz We observe a significant decrease in the end-to-end delay after SAI algorithm implementation Probabilityfor the end-to-end delay to be less than 100ms is above80 in both scenarios whereas without the algorithm thisprobability is below 70 for 100 vehicles (Figure 9(a)) andbelow 60 for 150 vehicles (Figure 9(b)) The variation inthis delay between various SAIs is because of the differencein transmission parameters Larger awareness range leadingto higher number of neighbors |F119894| and higher beaconfrequency results in more congestion and large queuingtimes The reason for better delay values with SAI algorithmis mainly the result from restricting resources to the requiredtransmission parameters for safety applicationsmentioned inSection 32

Figure 10 shows the aggregate downlink goodput for thesame scenario investigated for end-to-end delay After theimplementation of SAI algorithm the downlink goodput for100 vehicles dropped from 746Mbps to 622Mbps and for150 vehicles it dropped from 2167Mbps to 1376Mbps Thissignificant decrease in the downlink direction is again due tothe restriction of unnecessary data dissemination from theVSA server to vehicles Eventually this decrease of goodputresults in less load on the LTE system The increase in thedownlink goodput for the scenario without SAI explains thehigh end-to-end delay observed in Figure 9 showing a burstof packet in the downlink leading to waiting time in thequeue

Furthermore the impact of regular cellular traffic on ourLTE vehicular network is studied in order to validate ourfindings and algorithm A broad picture of our observa-tions under different scenarios is shown in Figure 11 whereprobability of end-to-end delay being less than or equal to50ms is studied As expected this probability decreases byabout 3-4 when background traffic is added However with

100 150

Number of vehicles

25

20

15

10

5

0

Dow

nlin

k go

odpu

t (M

bps)

Without SAIWith SAI

746622

2167

1376

Figure 10 Downlink goodput for 100 and 150 vehicles at an averagespeed 40 kmh (GCC)

SAI algorithm probability of end-to-end delay being lessthan 50ms stays above 90 even after adding backgroundtraffic

Finally we include the SAI into MAC layer PDU con-trol elements to prove our concept of D2D like unicasttransmission between sender and receivers Modeling of thesystem is carried out in accordance with the description inSection 22 Figure 12 shows end-to-end delay for 100 and150 vehicles with and without the inclusion of SAI in MAClayer It is evident that 150 vehicles experience almost thesame probability for end-to-end delay being less than 50msas experienced by 100 vehicles This shows that with the useof MAC layer control elements around 50 more vehiclescan be accommodated by LTE network increasing the spec-trum efficiency eventually leading to higher capacity of thesystem

Wireless Communications and Mobile Computing 15

Simple case With backgroundtraffic

With SAI With backgroundand SAI

6065707580859095

100

150 vehicles100 vehicles

Prob

abili

ty o

f end

-to-e

ndde

layle

50

ms

Scenarios

Figure 11 Probability of end-to-end delay being le50ms for 100 and150 vehicles at 40 kmh in various scenarios

With SAI in APP layer With SAI in MAC layer80828486889092949698

100

100 vehicles150 vehicles

p

roba

bilit

y fo

r end

-to-e

ndde

layle50

ms

Figure 12 Probability of end-to-end delay being le50ms for 100 and150 vehicles with and without MAC layer inclusion

5 Conclusion

This paper proposes the safety application identifier conceptin the form of an algorithm that is tested and analyzedwithin the impact of vehicular urbanmulticell radio environ-ment employing multipath fading channels and backgroundtraffic The proposed algorithm uses dynamic adaptation oftransmission parameters in order to fulfill stringent vehicu-lar application requirements while minimizing latency andreducing system load

With the help of extensive system-level simulations thecrucial impact of channel modeling and mobility traces isevident with its influence on the received signal qualitydecreasing the probability of end-to-end delay to be le50msfor various awareness ranges by 19 to 30 and 9 to17 respectively Studying the impact of awareness rangeit is observed that with range up to 1000m for 1Hz 500mfor 2Hz and 250m for 10Hz beacon frequency the systemperformed well With the help of these findings the SAIalgorithm is proposed and implemented Probability of expe-riencing an end-to-end delay of less than 100ms increasedby around 20 and the downlink goodput decreased signif-icantly from 2167Mbps to 1376Mbps for 150 vehicles

Furthermore the impact of having background traffic isalso taken into consideration Results show that the systemis slightly affected by having regular background trafficHowever the probability of end-to-end delay being le50ms

still stays satisfactorily above 85 with the use of SAI Thispaper also proposes the inclusion of SAI inMAC layer controlelements producing a D2D type of communication withinvehicles After modeling this system it was found that thesame amount of system resources and requirements are metwith additional 50 vehicles increasing the capacity of ourLTE vehicular network Finally in future works we plan tointegrate SAI incorporated within MAC layer in a hetero-geneous network with DSRC while exploring the networkslicing in 5G cellular networks for vehicular communicationscomplementing each otherrsquos weaknesses and strengths inorder to meet the vehicular network requirements

Conflicts of Interest

The authors declare that they have no conflicts of interest

Acknowledgments

Marvin Sanchez would like to acknowledge the support ofLAMENITEC Erasmus Mundus Program of the EuropeanUnion to be a guest researcher at GCU Results were obtainedusing the EPSRC funded ARCHIE-WeSt High PerformanceComputer (httpswwwarchie-westacuk) EPSRC Grantno EPK0005861

References

[1] S Sesia I Toufik and M Baker LTE the UMTS long termevolution from theory to practice Wiley Publishing 2009

[2] H Hartenstein and K P Laberteaux ldquoA tutorial survey onvehicular ad hoc networksrdquo IEEE Communications Magazinevol 46 no 6 pp 164ndash171 2008

[3] D Jiang and L Delgrossi ldquoTowards an international standardfor wireless access in vehicular environmentsrdquo in Proceedings ofthe 67th IEEE Vehicular Technology Conference (VTC rsquo08) pp2036ndash2040 May 2008

[4] S Al-Sultan M M Al-Doori A H Al-Bayatti and H ZedanldquoA comprehensive survey on vehicular Ad Hoc networkrdquoJournal of Network and Computer Applications vol 37 no 1 pp380ndash392 2014

[5] K Tanuja T Sushma M Bharathi and K Arun ldquoA surveyon VANET technologiesrdquo International Journal of ComputerApplications vol 121 no 18 2015

[6] H Y Kim D M Kang J H Lee and T M Chung ldquoA per-formance evaluation of cellular network suitability for VANETrdquoWorld Academy of Science Engineering and Technology Interna-tional Science Index vol 64 no 6 pp 124ndash127 2012

[7] A Fasbender M Gerdes and S Smets ldquoCellular NetworkingTechnologies in ITS Solutions Opportunities and Challengesrdquopp 1ndash13 Springer Berlin Germany 2012

[8] ldquoCellular coverage and tower map for ee network in glasgowrdquohttpswwwcellmappernetmap

[9] GAraniti C CampoloMCondoluci A Iera andAMolinaroldquoLTE for vehicular networking a surveyrdquo IEEECommunicationsMagazine vol 51 no 5 pp 148ndash157 2013

[10] A Vinel ldquo3GPP LTE versus IEEE 80211pWAVE which tech-nology is able to support cooperative vehicular safety applica-tionsrdquo IEEE Wireless Communications Letters vol 1 no 2 pp125ndash128 2012

Wireless Communications and Mobile Computing 5

S-GWP-GWVSA

Server

Awareness Range

VSM

S-GW serving gatewayP-GW PDN gatewayVSA vehicular safety application

idik

k

R

Position dkSAIi (Ri BFi)

Figure 1 LTE multicell system and awareness range

will also extract the SAI information from the receivedpacket This SAI information will then be checked againstthe database stored in the server (Table 2) With this SAIthe server retrieves the transmission beacon frequency BF119894and awareness range 119877119894 requirement of the application thatvehicle 119894 is utilizing Using the required awareness range theserver determines the forwarding set of vehicles (neighboringvehicles) by

F119894 = forall119896 119889119894119896 lt 119877119894 119894 = 119896 (1)

where 119889119894119896 is the distance from vehicle 119894 to the neighboringvehicle 119896 Recalling the concept of data freshness defined by[15] the VSA server will discard the packet if it arrives laterthan the beacon inter arrival time (1BF119894)22 MAC Layer Inclusion This work proposes the inclusionof SAI from application layer to MAC Protocol Data Unit(PDU) control elements This inclusion in MAC layer bringsthe processing towards the base station decreasing latencyand reducing the vehicular traffic load on themobile network

In cellular networks eNB uses the cell radio networktemporary identifier (C-RNTI) to uniquely identify UEs Asshown in Figure 3 a unique C-RNTI is assigned to everyuser by the eNB during the initial random access procedurewhich is used for identifying the radio resource control(RRC) connection and for scheduling purposes Coding anddecoding of physical downlink control channel (PDCCH)for a specific UE is based on its C-RNTI UEs initiatecontention based access to the network by transmitting apreamble sequence on the physical random access channel

(PRACH) to which the eNB responds with a C-RNTI onthe physical downlink shared channel (PDSCH) In vehicularcommunications since the awareness range is a function ofvehiclersquos geographical location if the transmitterrsquos location isunknown to the network an additional message containingtransmitterrsquos location is required

So for our proposed system model signaling shown inFigure 3 the transmitter registers with the network in thesame way as discussed above and it includes its location inthe RRC connection request message in its establishmentcause as mo-Data transmitted via the physical uplink sharedchannel (PUSCH) This location is introduced in the RRCconnection request message as a new information elementin line with the 3GPP specifications for RRC connectionrequest [21] At the same time all the UEs already registeredto the network update the eNB of their location using thenetwork assisted global navigation satellite system (GNSS)location update message as specified in [35] Once thevehicle has a beacon ready for transmission it sends outa scheduling request (SR) in the physical uplink controlchannel (PUCCH) The eNB replies back in the downlinkwith the grant for sending BSR After receiving this grantthe vehicle will send out the BSR in PUSCH containing theMAC PDU SAI will be in the unused 16-bit long MACControl Element inside the BSR This element utilizes spacecurrently reserved for future use and is indexed in the MACPDU subheader by the logical channel ID (LCID) value equalto 01011 [36] This new element now referred to as SAIis appended to the existing LCID values such as commoncontrol channel (CCCH) C-RNTI and the padding as shownin Figure 4 The eNB will then use the SAI to determine the

6 Wireless Communications and Mobile Computing

PayloadSAIIP Header

PDCP SDUPDCP header

RLC SDURLC header

MAC SDUMAC header

Transport block CRC

Location coordinates + safety message

PayloadIP SAI

Vehicular safety message

Inclusion of SAI

Header compression

PDCP

RLC

Ciphering

Segmentation

MAC

Multiplexing

PHYOne TB per TTI (1 ms)

Figure 2 SAI inclusion in LTE message structure

transmission parameters allotting transmission resources tothe sender and receiver vehicles

23 Scalability of the System Kato et al [15] emphasizedthe significance of placing the VSA server close to the edgeof LTE network It can also be argued if a single serverwould be able to serve other locations that is multipleeNodeBs or more servers would be required giving rise tothe question of system scalability For the implementationof SAI algorithm at the VSA server the location would playa vital role Considering the data freshness concept ideallocation of the server would be at the EPCTherefore havingmultiple EPCs serving their respective geographical locationswould have their own VSA servers with a similar approachas is for mobility management entity (MME) reducing theRTT while increasing the system capacity and eventuallymeeting the strict transmission requirements for vehicularsafety applications

In the case of having SAI appended in MAC layer theimplementation of SAI algorithm is brought towards theeNodeB reducing the RTT and eventually increasing thesystem capacity For vehicles being served by neighboringeNodeBs the message would be forwarded by the eNodeBserving the source vehicle to the respective neighboring

eNodeB via the X2 interface However if the neighboringeNodeB does not have an X2 interface or is residing in adifferent EPC then the VSA server would come into play Forsuch a scenario the centralization of the server will again besignificant

3 System Model

The network is assumed to be composed of 119873 vehiclesuniquely identified by their number 119894 (119894 = 1 sdot sdot sdot 119873) Vehiclesare assumed to use FDD LTE transceivers with 2 times 10MHzbandwidth uplink carrier frequency 1715MHz and downlinkcarrier frequency 2115MHz (band 4) [34 Table 55-1] Werefer to vehicles as LTE UEs

The work in this paper builds on to the performanceevaluation from [11] adding multicell and multipath alongwith Extended Vehicular A (EVA) fading environment Thenetwork modeled is a 2 times 2 km2 area of Manhattan Grid(MG) and Glasgow city center (GCC) shown in Figures5 and 6 respectively implemented in ns-3 [37] Mobilityof the vehicles in the network is generated using routesmobility model which assigns each vehicle with a routegenerated using Google maps API [14] Along with themobility model the site configuration is also changed for

Wireless Communications and Mobile Computing 7

Vehicle 1 eNB Vehicle 2 Vehicle N

Initial networkaccess

Registered to network

Registered to network

PRACH RA preamble

PDSCH RA responseAssigns C-RNTI 1

PUSCH RRC connection request

PDSCH connection setup

PUSCH RRC connection setup complete

Contains eUE 1location

Assuming norandom access

contention

Network assisted GNSS location update

Network assisted GNSS location update

Beacon for transmission

PUCCH (SR)

PDCCH UL grant for BSR

PUSCH (BSR containing MAC PDU)Contains SAI

Resource allocation

PDCCH TX grant PDCCH RX grant PDCCH RX grant

PUCCH (ACKNACK)PHICH (ACKNACK)

PUSCH beacon transmitted to nUE 2 middot middot middot nUE N

Figure 3 LTE signaling with MAC control elements in BSR

Header Controlelements Payload (SDU) Padding

Subheader1

Subheader2

SubheaderK

MAC Ctrlelement 1

MAC Ctrlelement 2

MAC Ctrlelement N

R ER LCID = 01100 Safety application identifier

MAC PDU

Figure 4 SAI in MAC control elements

8 Wireless Communications and Mobile Computing

500m

S-GWP-GWVSA server

400m

500m1 km

500m

Figure 5 MG model covered by 4 sites and 3 cellssite with 1000mintersite distance (ISD)

GCC This change in the network design considers a morerealistic network model employing the site configurationused by UKrsquos mobile operator EE in Glasgow [8] The LTEfunctionality is implemented through the LTE EPC networksimulator (LENA) module [38] that includes highly detailedeNB and UE functionalities as specified by 3GPP standardsTo manage handover and Internet connections eNBs areinterconnected through their X2 interfaces and connected tothe EPC through S1-U interface Interconnection from the P-GW to the VSA server residing in the EPC is modeled usingan error free 10 Gbps point-to-point link and TCPIP version4

The granularity of the LENA module is up to resourceblock (RB) level including the user plane control planeand reference signals used for coherent reception The safetymessages are UDP packets and often might be segmentedat the radio link control (RLC) and MAC layers to betransmitted on the transport shared channel (S-CH) passedto the physical (PHY) layer The size of transport blocks(TB) used at the PHY layer is decided by the MAC sublayerdepending on the channel quality and scheduling decisionsand may be transmitted using several RBs utilizing a mod-ulation and coding scheme (MCS) [38] Transmissions arescheduled every 1ms corresponding to one subframe usingProportional Fair Algorithm which is one of the approachessuitable for applications that include latency constraints [1Chapter 12] For downlink the algorithm utilizes userrsquoschannel quality indicator (CQI isin [0 sdot sdot sdot 15]) measurementreports sent from the vehicles For uplink channel qualitymeasurements are done by the eNB through scheduling

periodic vehicle transmissions of sounding reference signals(SRS)The radio resource control (RRC)module in LENAns-3 assigns the periodicity of SRS as a function of the numberof UEs attached to an eNB according to 3GPP UE-specificprocedure [37]

Each site modeled by an eNB uses 10W (40 dBm) totaltransmission power and a cosine antenna sector model with65∘ half power beamwidth (HPBW) [39] and azimuthaldirection 0∘ 120∘ or 240∘ with respect to north and 1mintersector space We assume utilization of 2 times 10MHzbandwidth whichmeans there are119872 = 50RBs for the uplinkand for the downlink Therefore assuming that the radiopropagation channel is constant in subframe 119905 if vehicle 119894 isscheduled to receive from its serving cell 119895 on resource block119898 isin [0 sdot sdot sdot 49] the received signal power 119875119895119894 can be modeledas follows

119875119895119894 (119898 119905) = 119875119895 (119898 119905) 119866119895 (120579119895119894)119866119894 (120579119894119895)119871119895119894 (119898 119905) (2)

where 119875119895(119898 119905) is the transmitted power for resource block119898119866119895(120579119895119894) is the antenna gain of cell 119895 in the direction of user 119894119866119894(120579119894119895) is the antenna gain of user 119894 in the direction of cell 119895and 119871119895119894 is the path loss from cell 119895 to user 119894

In the single cell scenario there is no intercell interferenceand mutually exclusive RBs are assigned to users Howeverin a more general deployed urban environment the availablespectrum is utilized by multiple cells in the service area andthe intercell interference is one important limiting aspectespecially for cell-edge users [1 Chapter 12]

In order to allocate usersrsquo transmissions while at the sametime maximizing the signal to interference plus noise ratio(SINR) for each RB frequency reuse and handover are usedFrequency allocation has an impact on the achievable systemcapacity and delay since the transport block size MCS andthe transmission schedule are dependent on the SINR [37Chapter 9] In ns-3 if several RBs are used to transmit a TBthe vector composed of the SINRs on each RB is used toevaluate the TB error probability The SINR in RB 119898 whenuser 119894 is scheduled to receive from serving cell 119895 can bewrittenas follows

SINR119894 (119898 119905)= 119875119895 (119898 119905) 119866119895 (120579119895119894)119866119894 (120579119894119895) 119871119895119894 (119898 119905)1198730 + sumforall119897 =119895 (119875119897 (119898 119905) 119866119897 (120579119897119894) 119866119894 (120579119894119897) 119871 119897119894 (119898 119905))

(3)

where1198730 is the noise powerUEs report the SINR mapped to the CQI value corre-

sponding to the MCS (isin [0 sdot sdot sdot 28]) that ensures TB error ratele 01 according to the standard [40 Section 723] In ns-3 thedownlink SINR is used to generate periodic wideband CQIsfeedback (ie an average value of all RBs) and inband CQIs(ie a value for each RB) We set the CQI to be calculated bycombining the received signal power of the reference signalsto the interference power from the physical downlink sharedchannel as an approximation to (3)

For dynamic frequency allocation we use the distributedfractional frequency reuse algorithm that utilizes intercell

Wireless Communications and Mobile Computing 9

S-GWP-GWVSA

TCC

Figure 6 GCC model covered by 6 sites and 3 cellssite where eNB location is determined using UK operator EE mast data [8]

interference coordination (ICIC) [41] The eNB adaptivelyselects a set of RBs referred to as cell-edge subband (set to10RBs) based on information exchanged with its neighborsTherefore according to ICIC a vehicle should be allocatedin the cell-edge subband if its serving cell reference signalreceived quality (RSRQ) gets lower than minus10 dB

Handover decisions are also performed by eNBs based onconfigurable event-triggeredmeasurement reports includingthe RSRQ and the reference signal received power (RSRP)The measurements are performed by UEs usually in orthog-onal frequency division multiplexing (OFDM) symbols car-rying reference signals (RS) for antenna port 0 In ns-3 since the channel is assumed flat over an RB that iscomposed of 12 subcarriers all resource elements (RE) in anRB have the same power Hence assuming orthogonal RSreception the RSRP for cell 119895 measured by vehicle 119894 is givenby

RSRP119895119894 (119905) = sum119872minus1119898=0 sum119870minus1119896=0 (119875119895119894 (119898 119896 119905) 119870)119872= sum119872minus1119898=0 (119875119895119894 (119898 119905) 12)119872

(4)

where 119875119895119894(119898 119896 119905) is the received power of RS 119896 in RB 119898 119872is the number of RBs and 119870 is the number of RS measuredin the RB The average value is passed to higher layers every200ms therefore the index 119905 is omitted for simplicity

The RSRQ for serving cell 119895measured by vehicle 119894 can becomputed by the following [42]

RSRQ119895119894 = 119872 times RSRP119895119894RSSI119895119894

(5)

where RSSI119895119894 is the received signal strength indicator (RSSI)measured by UE 119894when served by cell 119895 and according to the

10 Wireless Communications and Mobile Computing

Table 3 LTE simulation parameters

Parameter ValueSimulation time 100 seconds

Road model Manhattan grid model (MG) (two way roads)Glasgow City Center (GCC) (2 km times 2 km)

LTE network 4 sites with 3 cellssite 1000m ISD for MG6 sites with 3 cellssite UK Operator EE mast data [8]

Transmission power eNB 40 dBm UE 23 dBmCarrier frequency DLUL 2115MHz1715MHzChannel bandwidth 2 times 10MHz (2 times 50RBs)Noise Figure eNB 5 dB UE 9 dBUE antenna model Isotropic (0 dBi)eNB antenna model 15 dB Cosine model 65∘ HPBWScheduling algorithm Proportional Fair

Handover algorithm A2A4RSRQ RSRQ threshold minus5 dBand NeighbourCellOffset = 2 (1 dB)

Frequency reuse Distributed Fractional Freq Reuse

Path loss model LogDistance (120572 = 3) and3GPP Extended Vehicular A model

Safety message format 256 bytes UDPNumber of vehicles 75 100 125 150Average vehiclersquos speed 20 kmh 40 kmhBeacon frequency 1Hz 10HzAwareness range (119877) 100m 250m 500m 750m 1000m

standard comprises the linear average of the total receivedpower observed by the UE from all sources includingcochannel serving and nonserving cells adjacent channelinterference thermal noise and so forth [42] The RSSImeasurement model in ns-3 is computed considering two REper RB (the one that carries the RS) [37] that is

RSSI119895119894 (119905)= 119872minus1sum119898=0

2 times (119873012 +sumforall119897

119875119897 (119898 119905) 119866119897 (120579119897119894) 119866119894 (120579119894119897)12 times 119871 119897119894 (119898 119905) ) (6)

Furthermore in order to perform handover decisions thereported RSRQ for neighboring cell 119897 when vehicle 119894 is con-nected through serving cell 119895 is computed by the following

RSRQ119897119894 = 119872 times RSRP119897119894RSSI119895119894

(7)

The RSRQs for serving and neighboring cells are usedwith theA2A4RsrqHandoverAlgorithm to trigger LTE eventsA2 and A4 [43 Section 554] Event A2 is set to be triggeredwhen the RSRQ of the serving cell becomes worse than minus5 dB(ServingCellThreshold parameter) While event A2 is trueevent A4 is triggered if a neighboring cell becomes betterthan a threshold which is set by the algorithm to a verylow value for the trigger criteria to be true Thus the mea-surement reports received by the eNB are used to consider aneighboring cell as a target cell for handover only if its RSRQ

is 1 dB higher than the serving cell (NeighbourCellOffsetparameter) Table 3 summarizes the simulation parametersfor LTE

31 Vehicular Radio Urban Environment As mentioned inSection 11 previous evaluations lack the use of multipathand multicell environments Use of proper channel modelingis essential in evaluating system models as it poses close toreality system degradations The system model presented inthis paper consists of 6 sites with 3 cells per site Users areassumed to have isotropic antenna (119866119894(120579119894119895) = 0 dBi) and thepath loss (119871119895119894) ismodeled using the Log-distance propagationwith shadow exponent 119899 = 3 [44] plus 3GPP EVA multipathfading modelL119895119894 [34] and is computed by the following

119871119895119894 (119898 119905) = 1198710 + 10119899 log10 (119889119895119894 (119905)1198890 ) +L119895119894 (119898 119905) dB (8)

where 1198710 = 20 log10(120582(41205871198890)) 120582 is the wavelength and1198890 is the reference distance assumed to be 10m in oursimulations Traces for EVA were generated using MATLABwhile modifying the classical Doppler spectrum for each tapat vehicular speed (V) of 40 kmh and 60 kmh as

119878 (119891) prop 1(1 minus (119891119891119863)2)05 (9)

where 119891119863 = V120582 These EVA traces specified by 3GPP in[34] improvise the NLOS conditions relative to the speed

Wireless Communications and Mobile Computing 11

of UE and the transmission bandwidth Simulation of themultipath fading component in ns-3 uses the trace startingat a random point along the time domain for a given user andserving cell The channel object created with the rayleighchanfunction is used for filtering a discrete-time impulse signal inorder to obtain the channel impulse response The filteringis repeated for different Transmission Time Interval (TTI)thus yielding subsequent time-correlated channel responses(one per TTI) This channel response is then processedwith the pwelch function for obtaining its power spectraldensity values which are then saved in a file with the formatcompatible with the simulator model

32 SAI Implementation Model Archer and Vogel in [45]carried out a survey on traffic safety problems in urbanareas It was concluded that overtaking tends to occur lessfrequently within urban areas where the speed limit isless than 50 kmh Lane changing however occurs quitefrequently within urban areas due to low speed limits Thenumber of rear-end accidents is greater within urban areasthan in rural areas and similarly due to higher level ofcongestion and higher number of traffic junctions there isa greater number of opportunities for turning and crossingaccidents within urban area According to another survey of4500 crashes by the Insurance Institute of Highway Safety[46] 13 of accidents are during lane change maneuvers12 are intersection collisions 22 are due to traffic lightviolation 9 are head on collisions 18 include left turningcrashes 12 are due to emergency vehicles parked on blindturns and 14 are due to over speeding In the light ofthese statistics for our adopted urban environment the SAIs(Table 2) we utilize in our simulations are 1 (25) 3 (22) 5(9) 6 (18) 7 (12) and 9 (14)

33 Background Traffic In order to model close to realscenarios we added background traffic into our simulationsModeling of such traffic is done bymaking 50of the existingvehicles use voice application and 25 stream videos Voiceapplication is modeled by implementing constant bit rate(CBR) traffic generator on G711 codec using onoffapplicationon ns-3 [37] Voice payload is 172 bytes and the data ratechosen is 688 Kbps The reason for using a high data rate isto cater worst case scenario where an operator would preferhaving a high quality voice application At the same timevideo streaming is being carried out using evalvid simulatormodule on H263 codec with a payload of 1460 bytes at a datarate of 122Mbps [47] Both these voice and video applicationsare assigned their respective QCIs

34 Performance Measures The primary performance mea-sures used are the end-to-end packet delay the packet deliveryratio and the downlink goodputThe goodput is defined as thenumber of useful information bits received at the applicationlayer per unit of time

In addition we use the notation |F119894| to represent theaverage number of neighbors calculated over the number ofreceivers whenever a beacon is disseminated The end-to-enddelay of a beacon (119863119894119896) is defined as the time measured fromthe start of its transmission at vehicle 119894rsquos application layer until

its successful reception at vehicle 119896rsquos application layer PDR isdefined as the beacon success rate within the awareness range119877 [48] that is for a beacon sent fromvehicle 119894 its PDR is givenby

119875119894 (119877) = sum119896isinF119894

1198681198941198961003816100381610038161003816F1198941003816100381610038161003816 1003816100381610038161003816F1198941003816100381610038161003816 gt 0 (10)

where | sdot | indicates the cardinality of the set and 119868119894119896 (isin 0 1)is set to 1 if the beacon is received by vehicle 1198964 Simulation Results

The success of cellular networks relies on the scalability ofthe system capacity achieved by reusing the spectrum Thecommon approach implemented by operators is to caterthe capacity needs as the traffic demand grows Thereforefirst we analyze the capacity limit which results from thesingle cell case compared to multiple cells using the Friisradio propagation environment and afterwards we analyzethe impact of the urban radio environment as a function ofthe awareness range

Figure 7(a) shows the cumulative distribution function(CDF) of the end-to-end delay when the network is com-posed of 100 vehicles moving at an average speed of 40 kmhThe service area is covered using a single cell with an isotropicantenna located at the upper left corner of Manhattan Gridmodel as utilized by [11] in a Friis radio environment wherevehicles transmit at 1 Hz beacon frequency We observe thatfor awareness range of up to 500m the end-to-end delay isle100ms with around 95 probability Probability of higherend-to-end delay increases more between 250m to 500mas compared to the increase from 500m to 750m That ismainly the result of the increase in traffic but also as theawareness range increases cars located on roads near the edgeof the Manhattan model will have neighbors towards the celldirection with better signal quality which is further exploitedby the scheduling algorithm Figure 7(a) also shows that thereis no significant increase in end-to-end delay from 750mto 1000m This is due to the border effect incurred by thecar located at the edge of the model having less number ofneighbors when increasing the awareness range as comparedto the cars located at the center of the cell Nevertheless theseresults show that up to 100 vehicles were possible for thesingle cell case at a beacon frequency of 1Hz under Friis radioenvironment In order to analyze the intercell interferenceimpact on the delay while increasing the number of vehiclesand beacon frequency we increase the network capacity bydeploying 4 sites with 3 sectorssite (the average number ofvehicles per sector is reduced to 125) Figure 7(b) shows thatthe end-to-enddelay is then further improved to below 50mswith 91 success rate when awareness range of 1000m at thebeacon frequency of 1Hz was utilized We can also noticethat in this case there are some beacons that experiencedlonger delays compared to the single cell scenario causedby low quality conditions of cell-edge users due to intercellinterference and handover However with 1000m 934 ofthe forwarded beacons were received with end-to-end delayle 100ms and 982 for 250m

12 Wireless Communications and Mobile Computing

0 50 100 150 200 250 3000

01

02

03

04

05

06

07

08

09

1CD

F

End-to-end delay (ms)

R = 100

R = 250

R = 500

R = 750

R = 1000

(a)

0

01

02

03

04

05

06

07

08

09

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 1

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 35

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 14

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 25

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 41

(b)

0

01

02

03

04

05

06

07

08

09

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 1

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 35

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 14

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 25

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 41

(c)

Figure 7 LTE end-to-end delay for 100 vehicles at a beacon frequency 1Hz and average speed 40 kmh with Friis radio environment (a)single cell (b) multicell (4 sites 3 sectorssite) (c) multicell with EVA fading radio environment

The results in Figure 7(b) show that under the Friis radiopropagation environment the capacity increases and thesystem performance is not significantly affected in terms ofthe end-to-end delay in presence of intercell interference andhandover since the scheduler effectively managed interfer-ence by time and frequency allocations However in an urbanenvironment it is expected that fading caused by buildingsand Doppler shift due to the relative velocity of vehicles havea very important impact on the received signal quality [49]Figure 7(c) shows the results when the radio propagationenvironment was changed to Log-distance-dependent and

3GPP EVA under the same site configurations Under thisradio environment the probability for the end-to-end delayto be less than or equal to 100ms was around 76ndash78 forawareness range between 100m and 500m 70 for 750mand 66 for 1000m Hence the significant effect on the end-to-end delay is because of path loss and frequency selectivefading imposed by the EVA fading environmentThe end-to-enddelay for beacons to and fromcell-edge users significantlyincreases Similarly for inner cell users the end-to-end delayremained close to the Friis results and for up to 500m theend-to-end delay was le50ms with 74 to 71 probability

Wireless Communications and Mobile Computing 13

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 17

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 53

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 20

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 37

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 62

(a)

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 11

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 69

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 21

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 47

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 73

(b)

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 13

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 63

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 24

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 47

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 74

(c)

Figure 8 End-to-end delay for sample networks at average speed 40 kmh with 150 vehicles at beacon frequency (a) 1Hz with MG model(b) 1Hz with GCC model (c) 10Hz with GCC model

65 for 750m and 59 for 1000m Therefore it can bededuced that with the addition of multipath fading alongwith critical radio propagation environment probability ofdelay being le50ms decreases by 19 to 30

Continuing on with multicell and EVA fading radioenvironment by increasing the network size the number ofvehicles in the awareness range increases and the end-to-enddelay increases as shown in Figure 8(a) Nevertheless theend-to-end delay remained below 100ms for up to 250mwith82 probability for 100 vehicles and similar rate was obtainedwith up to 150 vehicles When the awareness range increasesto 500m the probability was 80 for 100 vehicles while for150 vehicles further reduces to 73 For awareness range of750m and 1000m the probability reduced to 75 and 71 for100 vehicles while for 150 vehicles reduces to 69 and 60respectively

Changing the simulation model along with mobilitytraces to Glasgow city center the effect on delay performancewith 150 vehicles at 1 Hz and 10Hz is shown in Figures 8(b)and 8(c) respectively It is interesting to observe the effect ofchanging the mobility model to a European city Where each

city block is 400m in theMGmodel average city block size inGCC is around 200m increasing vehicle density eventuallygrowing the average number of neighbors for GCC modelThis can be observed for an awareness range of 250m theprobability of delay to be le50ms decreases from 77 to 68while for 1000m this deterioration goes from 54 to 37hence an overall decrease of about 9 to 17 for variousawareness ranges Similarly in Figure 8(c) it is observedthat changing the beacon frequency from 1Hz to 10Hz withan awareness range of 500m the probability of delay to bele50ms drastically decreases from 55 to 22 Whereas foran awareness range of 1000m degradation of about 29 isobserved Performance with a high beacon frequency suchas 10Hz is overall seen to be poor That is because packetsthat arrive late are discarded and are mainly sent by cell-edgeusers for which their neighborhood very likely includes usersat the cell edge Hence the delay from the uplink is indirectlyused as a spatial filtering of low quality cell-edge users Thissuggests that if the information is not useful when receivedafter 100ms the better use of resources is to drop the packetat the minimum latency requirement by the server

14 Wireless Communications and Mobile Computing

0010203040506070809

1

0 20 40 60 80 100 120 140 160 180 200

CDF

End-to-end delay (ms)

SAI = 11003816100381610038161003816Fi

1003816100381610038161003816 = 19

SAI = 31003816100381610038161003816Fi

1003816100381610038161003816 = 16

SAI = 51003816100381610038161003816Fi

1003816100381610038161003816 = 41

SAI = 61003816100381610038161003816Fi

1003816100381610038161003816 = 14

SAI = 71003816100381610038161003816Fi

1003816100381610038161003816 = 56

SAI = 91003816100381610038161003816Fi

1003816100381610038161003816 = 54

Without SAI 1003816100381610038161003816Fi1003816100381610038161003816 = 16

(a)

0010203040506070809

1

0 20 40 60 80 100 120 140 160 180 200

CDF

End-to-end delay (ms)

SAI = 11003816100381610038161003816Fi

1003816100381610038161003816 = 26

SAI = 31003816100381610038161003816Fi

1003816100381610038161003816 = 2

SAI = 51003816100381610038161003816Fi

1003816100381610038161003816 = 35

SAI = 61003816100381610038161003816Fi

1003816100381610038161003816 = 22

SAI = 71003816100381610038161003816Fi

1003816100381610038161003816 = 78

SAI = 91003816100381610038161003816Fi

1003816100381610038161003816 = 94

Without SAI 1003816100381610038161003816Fi1003816100381610038161003816 = 21

(b)

Figure 9 LTE end-to-end delay at average speed 40 kmh (GCC) and (a) 100 vehicles and (b) 150 vehicles

Next we observe and compare the results after theproposed algorithm implementation Figure 9 shows theCDF of the end-to-end delay for 100 and 150 vehicles atan average speed of 40 kmh with and without the imple-mentation of SAI algorithm The scenario being comparedhas the awareness range set to 500m and beacon frequencyto 10Hz We observe a significant decrease in the end-to-end delay after SAI algorithm implementation Probabilityfor the end-to-end delay to be less than 100ms is above80 in both scenarios whereas without the algorithm thisprobability is below 70 for 100 vehicles (Figure 9(a)) andbelow 60 for 150 vehicles (Figure 9(b)) The variation inthis delay between various SAIs is because of the differencein transmission parameters Larger awareness range leadingto higher number of neighbors |F119894| and higher beaconfrequency results in more congestion and large queuingtimes The reason for better delay values with SAI algorithmis mainly the result from restricting resources to the requiredtransmission parameters for safety applicationsmentioned inSection 32

Figure 10 shows the aggregate downlink goodput for thesame scenario investigated for end-to-end delay After theimplementation of SAI algorithm the downlink goodput for100 vehicles dropped from 746Mbps to 622Mbps and for150 vehicles it dropped from 2167Mbps to 1376Mbps Thissignificant decrease in the downlink direction is again due tothe restriction of unnecessary data dissemination from theVSA server to vehicles Eventually this decrease of goodputresults in less load on the LTE system The increase in thedownlink goodput for the scenario without SAI explains thehigh end-to-end delay observed in Figure 9 showing a burstof packet in the downlink leading to waiting time in thequeue

Furthermore the impact of regular cellular traffic on ourLTE vehicular network is studied in order to validate ourfindings and algorithm A broad picture of our observa-tions under different scenarios is shown in Figure 11 whereprobability of end-to-end delay being less than or equal to50ms is studied As expected this probability decreases byabout 3-4 when background traffic is added However with

100 150

Number of vehicles

25

20

15

10

5

0

Dow

nlin

k go

odpu

t (M

bps)

Without SAIWith SAI

746622

2167

1376

Figure 10 Downlink goodput for 100 and 150 vehicles at an averagespeed 40 kmh (GCC)

SAI algorithm probability of end-to-end delay being lessthan 50ms stays above 90 even after adding backgroundtraffic

Finally we include the SAI into MAC layer PDU con-trol elements to prove our concept of D2D like unicasttransmission between sender and receivers Modeling of thesystem is carried out in accordance with the description inSection 22 Figure 12 shows end-to-end delay for 100 and150 vehicles with and without the inclusion of SAI in MAClayer It is evident that 150 vehicles experience almost thesame probability for end-to-end delay being less than 50msas experienced by 100 vehicles This shows that with the useof MAC layer control elements around 50 more vehiclescan be accommodated by LTE network increasing the spec-trum efficiency eventually leading to higher capacity of thesystem

Wireless Communications and Mobile Computing 15

Simple case With backgroundtraffic

With SAI With backgroundand SAI

6065707580859095

100

150 vehicles100 vehicles

Prob

abili

ty o

f end

-to-e

ndde

layle

50

ms

Scenarios

Figure 11 Probability of end-to-end delay being le50ms for 100 and150 vehicles at 40 kmh in various scenarios

With SAI in APP layer With SAI in MAC layer80828486889092949698

100

100 vehicles150 vehicles

p

roba

bilit

y fo

r end

-to-e

ndde

layle50

ms

Figure 12 Probability of end-to-end delay being le50ms for 100 and150 vehicles with and without MAC layer inclusion

5 Conclusion

This paper proposes the safety application identifier conceptin the form of an algorithm that is tested and analyzedwithin the impact of vehicular urbanmulticell radio environ-ment employing multipath fading channels and backgroundtraffic The proposed algorithm uses dynamic adaptation oftransmission parameters in order to fulfill stringent vehicu-lar application requirements while minimizing latency andreducing system load

With the help of extensive system-level simulations thecrucial impact of channel modeling and mobility traces isevident with its influence on the received signal qualitydecreasing the probability of end-to-end delay to be le50msfor various awareness ranges by 19 to 30 and 9 to17 respectively Studying the impact of awareness rangeit is observed that with range up to 1000m for 1Hz 500mfor 2Hz and 250m for 10Hz beacon frequency the systemperformed well With the help of these findings the SAIalgorithm is proposed and implemented Probability of expe-riencing an end-to-end delay of less than 100ms increasedby around 20 and the downlink goodput decreased signif-icantly from 2167Mbps to 1376Mbps for 150 vehicles

Furthermore the impact of having background traffic isalso taken into consideration Results show that the systemis slightly affected by having regular background trafficHowever the probability of end-to-end delay being le50ms

still stays satisfactorily above 85 with the use of SAI Thispaper also proposes the inclusion of SAI inMAC layer controlelements producing a D2D type of communication withinvehicles After modeling this system it was found that thesame amount of system resources and requirements are metwith additional 50 vehicles increasing the capacity of ourLTE vehicular network Finally in future works we plan tointegrate SAI incorporated within MAC layer in a hetero-geneous network with DSRC while exploring the networkslicing in 5G cellular networks for vehicular communicationscomplementing each otherrsquos weaknesses and strengths inorder to meet the vehicular network requirements

Conflicts of Interest

The authors declare that they have no conflicts of interest

Acknowledgments

Marvin Sanchez would like to acknowledge the support ofLAMENITEC Erasmus Mundus Program of the EuropeanUnion to be a guest researcher at GCU Results were obtainedusing the EPSRC funded ARCHIE-WeSt High PerformanceComputer (httpswwwarchie-westacuk) EPSRC Grantno EPK0005861

References

[1] S Sesia I Toufik and M Baker LTE the UMTS long termevolution from theory to practice Wiley Publishing 2009

[2] H Hartenstein and K P Laberteaux ldquoA tutorial survey onvehicular ad hoc networksrdquo IEEE Communications Magazinevol 46 no 6 pp 164ndash171 2008

[3] D Jiang and L Delgrossi ldquoTowards an international standardfor wireless access in vehicular environmentsrdquo in Proceedings ofthe 67th IEEE Vehicular Technology Conference (VTC rsquo08) pp2036ndash2040 May 2008

[4] S Al-Sultan M M Al-Doori A H Al-Bayatti and H ZedanldquoA comprehensive survey on vehicular Ad Hoc networkrdquoJournal of Network and Computer Applications vol 37 no 1 pp380ndash392 2014

[5] K Tanuja T Sushma M Bharathi and K Arun ldquoA surveyon VANET technologiesrdquo International Journal of ComputerApplications vol 121 no 18 2015

[6] H Y Kim D M Kang J H Lee and T M Chung ldquoA per-formance evaluation of cellular network suitability for VANETrdquoWorld Academy of Science Engineering and Technology Interna-tional Science Index vol 64 no 6 pp 124ndash127 2012

[7] A Fasbender M Gerdes and S Smets ldquoCellular NetworkingTechnologies in ITS Solutions Opportunities and Challengesrdquopp 1ndash13 Springer Berlin Germany 2012

[8] ldquoCellular coverage and tower map for ee network in glasgowrdquohttpswwwcellmappernetmap

[9] GAraniti C CampoloMCondoluci A Iera andAMolinaroldquoLTE for vehicular networking a surveyrdquo IEEECommunicationsMagazine vol 51 no 5 pp 148ndash157 2013

[10] A Vinel ldquo3GPP LTE versus IEEE 80211pWAVE which tech-nology is able to support cooperative vehicular safety applica-tionsrdquo IEEE Wireless Communications Letters vol 1 no 2 pp125ndash128 2012

6 Wireless Communications and Mobile Computing

PayloadSAIIP Header

PDCP SDUPDCP header

RLC SDURLC header

MAC SDUMAC header

Transport block CRC

Location coordinates + safety message

PayloadIP SAI

Vehicular safety message

Inclusion of SAI

Header compression

PDCP

RLC

Ciphering

Segmentation

MAC

Multiplexing

PHYOne TB per TTI (1 ms)

Figure 2 SAI inclusion in LTE message structure

transmission parameters allotting transmission resources tothe sender and receiver vehicles

23 Scalability of the System Kato et al [15] emphasizedthe significance of placing the VSA server close to the edgeof LTE network It can also be argued if a single serverwould be able to serve other locations that is multipleeNodeBs or more servers would be required giving rise tothe question of system scalability For the implementationof SAI algorithm at the VSA server the location would playa vital role Considering the data freshness concept ideallocation of the server would be at the EPCTherefore havingmultiple EPCs serving their respective geographical locationswould have their own VSA servers with a similar approachas is for mobility management entity (MME) reducing theRTT while increasing the system capacity and eventuallymeeting the strict transmission requirements for vehicularsafety applications

In the case of having SAI appended in MAC layer theimplementation of SAI algorithm is brought towards theeNodeB reducing the RTT and eventually increasing thesystem capacity For vehicles being served by neighboringeNodeBs the message would be forwarded by the eNodeBserving the source vehicle to the respective neighboring

eNodeB via the X2 interface However if the neighboringeNodeB does not have an X2 interface or is residing in adifferent EPC then the VSA server would come into play Forsuch a scenario the centralization of the server will again besignificant

3 System Model

The network is assumed to be composed of 119873 vehiclesuniquely identified by their number 119894 (119894 = 1 sdot sdot sdot 119873) Vehiclesare assumed to use FDD LTE transceivers with 2 times 10MHzbandwidth uplink carrier frequency 1715MHz and downlinkcarrier frequency 2115MHz (band 4) [34 Table 55-1] Werefer to vehicles as LTE UEs

The work in this paper builds on to the performanceevaluation from [11] adding multicell and multipath alongwith Extended Vehicular A (EVA) fading environment Thenetwork modeled is a 2 times 2 km2 area of Manhattan Grid(MG) and Glasgow city center (GCC) shown in Figures5 and 6 respectively implemented in ns-3 [37] Mobilityof the vehicles in the network is generated using routesmobility model which assigns each vehicle with a routegenerated using Google maps API [14] Along with themobility model the site configuration is also changed for

Wireless Communications and Mobile Computing 7

Vehicle 1 eNB Vehicle 2 Vehicle N

Initial networkaccess

Registered to network

Registered to network

PRACH RA preamble

PDSCH RA responseAssigns C-RNTI 1

PUSCH RRC connection request

PDSCH connection setup

PUSCH RRC connection setup complete

Contains eUE 1location

Assuming norandom access

contention

Network assisted GNSS location update

Network assisted GNSS location update

Beacon for transmission

PUCCH (SR)

PDCCH UL grant for BSR

PUSCH (BSR containing MAC PDU)Contains SAI

Resource allocation

PDCCH TX grant PDCCH RX grant PDCCH RX grant

PUCCH (ACKNACK)PHICH (ACKNACK)

PUSCH beacon transmitted to nUE 2 middot middot middot nUE N

Figure 3 LTE signaling with MAC control elements in BSR

Header Controlelements Payload (SDU) Padding

Subheader1

Subheader2

SubheaderK

MAC Ctrlelement 1

MAC Ctrlelement 2

MAC Ctrlelement N

R ER LCID = 01100 Safety application identifier

MAC PDU

Figure 4 SAI in MAC control elements

8 Wireless Communications and Mobile Computing

500m

S-GWP-GWVSA server

400m

500m1 km

500m

Figure 5 MG model covered by 4 sites and 3 cellssite with 1000mintersite distance (ISD)

GCC This change in the network design considers a morerealistic network model employing the site configurationused by UKrsquos mobile operator EE in Glasgow [8] The LTEfunctionality is implemented through the LTE EPC networksimulator (LENA) module [38] that includes highly detailedeNB and UE functionalities as specified by 3GPP standardsTo manage handover and Internet connections eNBs areinterconnected through their X2 interfaces and connected tothe EPC through S1-U interface Interconnection from the P-GW to the VSA server residing in the EPC is modeled usingan error free 10 Gbps point-to-point link and TCPIP version4

The granularity of the LENA module is up to resourceblock (RB) level including the user plane control planeand reference signals used for coherent reception The safetymessages are UDP packets and often might be segmentedat the radio link control (RLC) and MAC layers to betransmitted on the transport shared channel (S-CH) passedto the physical (PHY) layer The size of transport blocks(TB) used at the PHY layer is decided by the MAC sublayerdepending on the channel quality and scheduling decisionsand may be transmitted using several RBs utilizing a mod-ulation and coding scheme (MCS) [38] Transmissions arescheduled every 1ms corresponding to one subframe usingProportional Fair Algorithm which is one of the approachessuitable for applications that include latency constraints [1Chapter 12] For downlink the algorithm utilizes userrsquoschannel quality indicator (CQI isin [0 sdot sdot sdot 15]) measurementreports sent from the vehicles For uplink channel qualitymeasurements are done by the eNB through scheduling

periodic vehicle transmissions of sounding reference signals(SRS)The radio resource control (RRC)module in LENAns-3 assigns the periodicity of SRS as a function of the numberof UEs attached to an eNB according to 3GPP UE-specificprocedure [37]

Each site modeled by an eNB uses 10W (40 dBm) totaltransmission power and a cosine antenna sector model with65∘ half power beamwidth (HPBW) [39] and azimuthaldirection 0∘ 120∘ or 240∘ with respect to north and 1mintersector space We assume utilization of 2 times 10MHzbandwidth whichmeans there are119872 = 50RBs for the uplinkand for the downlink Therefore assuming that the radiopropagation channel is constant in subframe 119905 if vehicle 119894 isscheduled to receive from its serving cell 119895 on resource block119898 isin [0 sdot sdot sdot 49] the received signal power 119875119895119894 can be modeledas follows

119875119895119894 (119898 119905) = 119875119895 (119898 119905) 119866119895 (120579119895119894)119866119894 (120579119894119895)119871119895119894 (119898 119905) (2)

where 119875119895(119898 119905) is the transmitted power for resource block119898119866119895(120579119895119894) is the antenna gain of cell 119895 in the direction of user 119894119866119894(120579119894119895) is the antenna gain of user 119894 in the direction of cell 119895and 119871119895119894 is the path loss from cell 119895 to user 119894

In the single cell scenario there is no intercell interferenceand mutually exclusive RBs are assigned to users Howeverin a more general deployed urban environment the availablespectrum is utilized by multiple cells in the service area andthe intercell interference is one important limiting aspectespecially for cell-edge users [1 Chapter 12]

In order to allocate usersrsquo transmissions while at the sametime maximizing the signal to interference plus noise ratio(SINR) for each RB frequency reuse and handover are usedFrequency allocation has an impact on the achievable systemcapacity and delay since the transport block size MCS andthe transmission schedule are dependent on the SINR [37Chapter 9] In ns-3 if several RBs are used to transmit a TBthe vector composed of the SINRs on each RB is used toevaluate the TB error probability The SINR in RB 119898 whenuser 119894 is scheduled to receive from serving cell 119895 can bewrittenas follows

SINR119894 (119898 119905)= 119875119895 (119898 119905) 119866119895 (120579119895119894)119866119894 (120579119894119895) 119871119895119894 (119898 119905)1198730 + sumforall119897 =119895 (119875119897 (119898 119905) 119866119897 (120579119897119894) 119866119894 (120579119894119897) 119871 119897119894 (119898 119905))

(3)

where1198730 is the noise powerUEs report the SINR mapped to the CQI value corre-

sponding to the MCS (isin [0 sdot sdot sdot 28]) that ensures TB error ratele 01 according to the standard [40 Section 723] In ns-3 thedownlink SINR is used to generate periodic wideband CQIsfeedback (ie an average value of all RBs) and inband CQIs(ie a value for each RB) We set the CQI to be calculated bycombining the received signal power of the reference signalsto the interference power from the physical downlink sharedchannel as an approximation to (3)

For dynamic frequency allocation we use the distributedfractional frequency reuse algorithm that utilizes intercell

Wireless Communications and Mobile Computing 9

S-GWP-GWVSA

TCC

Figure 6 GCC model covered by 6 sites and 3 cellssite where eNB location is determined using UK operator EE mast data [8]

interference coordination (ICIC) [41] The eNB adaptivelyselects a set of RBs referred to as cell-edge subband (set to10RBs) based on information exchanged with its neighborsTherefore according to ICIC a vehicle should be allocatedin the cell-edge subband if its serving cell reference signalreceived quality (RSRQ) gets lower than minus10 dB

Handover decisions are also performed by eNBs based onconfigurable event-triggeredmeasurement reports includingthe RSRQ and the reference signal received power (RSRP)The measurements are performed by UEs usually in orthog-onal frequency division multiplexing (OFDM) symbols car-rying reference signals (RS) for antenna port 0 In ns-3 since the channel is assumed flat over an RB that iscomposed of 12 subcarriers all resource elements (RE) in anRB have the same power Hence assuming orthogonal RSreception the RSRP for cell 119895 measured by vehicle 119894 is givenby

RSRP119895119894 (119905) = sum119872minus1119898=0 sum119870minus1119896=0 (119875119895119894 (119898 119896 119905) 119870)119872= sum119872minus1119898=0 (119875119895119894 (119898 119905) 12)119872

(4)

where 119875119895119894(119898 119896 119905) is the received power of RS 119896 in RB 119898 119872is the number of RBs and 119870 is the number of RS measuredin the RB The average value is passed to higher layers every200ms therefore the index 119905 is omitted for simplicity

The RSRQ for serving cell 119895measured by vehicle 119894 can becomputed by the following [42]

RSRQ119895119894 = 119872 times RSRP119895119894RSSI119895119894

(5)

where RSSI119895119894 is the received signal strength indicator (RSSI)measured by UE 119894when served by cell 119895 and according to the

10 Wireless Communications and Mobile Computing

Table 3 LTE simulation parameters

Parameter ValueSimulation time 100 seconds

Road model Manhattan grid model (MG) (two way roads)Glasgow City Center (GCC) (2 km times 2 km)

LTE network 4 sites with 3 cellssite 1000m ISD for MG6 sites with 3 cellssite UK Operator EE mast data [8]

Transmission power eNB 40 dBm UE 23 dBmCarrier frequency DLUL 2115MHz1715MHzChannel bandwidth 2 times 10MHz (2 times 50RBs)Noise Figure eNB 5 dB UE 9 dBUE antenna model Isotropic (0 dBi)eNB antenna model 15 dB Cosine model 65∘ HPBWScheduling algorithm Proportional Fair

Handover algorithm A2A4RSRQ RSRQ threshold minus5 dBand NeighbourCellOffset = 2 (1 dB)

Frequency reuse Distributed Fractional Freq Reuse

Path loss model LogDistance (120572 = 3) and3GPP Extended Vehicular A model

Safety message format 256 bytes UDPNumber of vehicles 75 100 125 150Average vehiclersquos speed 20 kmh 40 kmhBeacon frequency 1Hz 10HzAwareness range (119877) 100m 250m 500m 750m 1000m

standard comprises the linear average of the total receivedpower observed by the UE from all sources includingcochannel serving and nonserving cells adjacent channelinterference thermal noise and so forth [42] The RSSImeasurement model in ns-3 is computed considering two REper RB (the one that carries the RS) [37] that is

RSSI119895119894 (119905)= 119872minus1sum119898=0

2 times (119873012 +sumforall119897

119875119897 (119898 119905) 119866119897 (120579119897119894) 119866119894 (120579119894119897)12 times 119871 119897119894 (119898 119905) ) (6)

Furthermore in order to perform handover decisions thereported RSRQ for neighboring cell 119897 when vehicle 119894 is con-nected through serving cell 119895 is computed by the following

RSRQ119897119894 = 119872 times RSRP119897119894RSSI119895119894

(7)

The RSRQs for serving and neighboring cells are usedwith theA2A4RsrqHandoverAlgorithm to trigger LTE eventsA2 and A4 [43 Section 554] Event A2 is set to be triggeredwhen the RSRQ of the serving cell becomes worse than minus5 dB(ServingCellThreshold parameter) While event A2 is trueevent A4 is triggered if a neighboring cell becomes betterthan a threshold which is set by the algorithm to a verylow value for the trigger criteria to be true Thus the mea-surement reports received by the eNB are used to consider aneighboring cell as a target cell for handover only if its RSRQ

is 1 dB higher than the serving cell (NeighbourCellOffsetparameter) Table 3 summarizes the simulation parametersfor LTE

31 Vehicular Radio Urban Environment As mentioned inSection 11 previous evaluations lack the use of multipathand multicell environments Use of proper channel modelingis essential in evaluating system models as it poses close toreality system degradations The system model presented inthis paper consists of 6 sites with 3 cells per site Users areassumed to have isotropic antenna (119866119894(120579119894119895) = 0 dBi) and thepath loss (119871119895119894) ismodeled using the Log-distance propagationwith shadow exponent 119899 = 3 [44] plus 3GPP EVA multipathfading modelL119895119894 [34] and is computed by the following

119871119895119894 (119898 119905) = 1198710 + 10119899 log10 (119889119895119894 (119905)1198890 ) +L119895119894 (119898 119905) dB (8)

where 1198710 = 20 log10(120582(41205871198890)) 120582 is the wavelength and1198890 is the reference distance assumed to be 10m in oursimulations Traces for EVA were generated using MATLABwhile modifying the classical Doppler spectrum for each tapat vehicular speed (V) of 40 kmh and 60 kmh as

119878 (119891) prop 1(1 minus (119891119891119863)2)05 (9)

where 119891119863 = V120582 These EVA traces specified by 3GPP in[34] improvise the NLOS conditions relative to the speed

Wireless Communications and Mobile Computing 11

of UE and the transmission bandwidth Simulation of themultipath fading component in ns-3 uses the trace startingat a random point along the time domain for a given user andserving cell The channel object created with the rayleighchanfunction is used for filtering a discrete-time impulse signal inorder to obtain the channel impulse response The filteringis repeated for different Transmission Time Interval (TTI)thus yielding subsequent time-correlated channel responses(one per TTI) This channel response is then processedwith the pwelch function for obtaining its power spectraldensity values which are then saved in a file with the formatcompatible with the simulator model

32 SAI Implementation Model Archer and Vogel in [45]carried out a survey on traffic safety problems in urbanareas It was concluded that overtaking tends to occur lessfrequently within urban areas where the speed limit isless than 50 kmh Lane changing however occurs quitefrequently within urban areas due to low speed limits Thenumber of rear-end accidents is greater within urban areasthan in rural areas and similarly due to higher level ofcongestion and higher number of traffic junctions there isa greater number of opportunities for turning and crossingaccidents within urban area According to another survey of4500 crashes by the Insurance Institute of Highway Safety[46] 13 of accidents are during lane change maneuvers12 are intersection collisions 22 are due to traffic lightviolation 9 are head on collisions 18 include left turningcrashes 12 are due to emergency vehicles parked on blindturns and 14 are due to over speeding In the light ofthese statistics for our adopted urban environment the SAIs(Table 2) we utilize in our simulations are 1 (25) 3 (22) 5(9) 6 (18) 7 (12) and 9 (14)

33 Background Traffic In order to model close to realscenarios we added background traffic into our simulationsModeling of such traffic is done bymaking 50of the existingvehicles use voice application and 25 stream videos Voiceapplication is modeled by implementing constant bit rate(CBR) traffic generator on G711 codec using onoffapplicationon ns-3 [37] Voice payload is 172 bytes and the data ratechosen is 688 Kbps The reason for using a high data rate isto cater worst case scenario where an operator would preferhaving a high quality voice application At the same timevideo streaming is being carried out using evalvid simulatormodule on H263 codec with a payload of 1460 bytes at a datarate of 122Mbps [47] Both these voice and video applicationsare assigned their respective QCIs

34 Performance Measures The primary performance mea-sures used are the end-to-end packet delay the packet deliveryratio and the downlink goodputThe goodput is defined as thenumber of useful information bits received at the applicationlayer per unit of time

In addition we use the notation |F119894| to represent theaverage number of neighbors calculated over the number ofreceivers whenever a beacon is disseminated The end-to-enddelay of a beacon (119863119894119896) is defined as the time measured fromthe start of its transmission at vehicle 119894rsquos application layer until

its successful reception at vehicle 119896rsquos application layer PDR isdefined as the beacon success rate within the awareness range119877 [48] that is for a beacon sent fromvehicle 119894 its PDR is givenby

119875119894 (119877) = sum119896isinF119894

1198681198941198961003816100381610038161003816F1198941003816100381610038161003816 1003816100381610038161003816F1198941003816100381610038161003816 gt 0 (10)

where | sdot | indicates the cardinality of the set and 119868119894119896 (isin 0 1)is set to 1 if the beacon is received by vehicle 1198964 Simulation Results

The success of cellular networks relies on the scalability ofthe system capacity achieved by reusing the spectrum Thecommon approach implemented by operators is to caterthe capacity needs as the traffic demand grows Thereforefirst we analyze the capacity limit which results from thesingle cell case compared to multiple cells using the Friisradio propagation environment and afterwards we analyzethe impact of the urban radio environment as a function ofthe awareness range

Figure 7(a) shows the cumulative distribution function(CDF) of the end-to-end delay when the network is com-posed of 100 vehicles moving at an average speed of 40 kmhThe service area is covered using a single cell with an isotropicantenna located at the upper left corner of Manhattan Gridmodel as utilized by [11] in a Friis radio environment wherevehicles transmit at 1 Hz beacon frequency We observe thatfor awareness range of up to 500m the end-to-end delay isle100ms with around 95 probability Probability of higherend-to-end delay increases more between 250m to 500mas compared to the increase from 500m to 750m That ismainly the result of the increase in traffic but also as theawareness range increases cars located on roads near the edgeof the Manhattan model will have neighbors towards the celldirection with better signal quality which is further exploitedby the scheduling algorithm Figure 7(a) also shows that thereis no significant increase in end-to-end delay from 750mto 1000m This is due to the border effect incurred by thecar located at the edge of the model having less number ofneighbors when increasing the awareness range as comparedto the cars located at the center of the cell Nevertheless theseresults show that up to 100 vehicles were possible for thesingle cell case at a beacon frequency of 1Hz under Friis radioenvironment In order to analyze the intercell interferenceimpact on the delay while increasing the number of vehiclesand beacon frequency we increase the network capacity bydeploying 4 sites with 3 sectorssite (the average number ofvehicles per sector is reduced to 125) Figure 7(b) shows thatthe end-to-enddelay is then further improved to below 50mswith 91 success rate when awareness range of 1000m at thebeacon frequency of 1Hz was utilized We can also noticethat in this case there are some beacons that experiencedlonger delays compared to the single cell scenario causedby low quality conditions of cell-edge users due to intercellinterference and handover However with 1000m 934 ofthe forwarded beacons were received with end-to-end delayle 100ms and 982 for 250m

12 Wireless Communications and Mobile Computing

0 50 100 150 200 250 3000

01

02

03

04

05

06

07

08

09

1CD

F

End-to-end delay (ms)

R = 100

R = 250

R = 500

R = 750

R = 1000

(a)

0

01

02

03

04

05

06

07

08

09

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 1

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 35

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 14

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 25

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 41

(b)

0

01

02

03

04

05

06

07

08

09

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 1

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 35

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 14

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 25

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 41

(c)

Figure 7 LTE end-to-end delay for 100 vehicles at a beacon frequency 1Hz and average speed 40 kmh with Friis radio environment (a)single cell (b) multicell (4 sites 3 sectorssite) (c) multicell with EVA fading radio environment

The results in Figure 7(b) show that under the Friis radiopropagation environment the capacity increases and thesystem performance is not significantly affected in terms ofthe end-to-end delay in presence of intercell interference andhandover since the scheduler effectively managed interfer-ence by time and frequency allocations However in an urbanenvironment it is expected that fading caused by buildingsand Doppler shift due to the relative velocity of vehicles havea very important impact on the received signal quality [49]Figure 7(c) shows the results when the radio propagationenvironment was changed to Log-distance-dependent and

3GPP EVA under the same site configurations Under thisradio environment the probability for the end-to-end delayto be less than or equal to 100ms was around 76ndash78 forawareness range between 100m and 500m 70 for 750mand 66 for 1000m Hence the significant effect on the end-to-end delay is because of path loss and frequency selectivefading imposed by the EVA fading environmentThe end-to-enddelay for beacons to and fromcell-edge users significantlyincreases Similarly for inner cell users the end-to-end delayremained close to the Friis results and for up to 500m theend-to-end delay was le50ms with 74 to 71 probability

Wireless Communications and Mobile Computing 13

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 17

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 53

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 20

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 37

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 62

(a)

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 11

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 69

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 21

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 47

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 73

(b)

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 13

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 63

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 24

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 47

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 74

(c)

Figure 8 End-to-end delay for sample networks at average speed 40 kmh with 150 vehicles at beacon frequency (a) 1Hz with MG model(b) 1Hz with GCC model (c) 10Hz with GCC model

65 for 750m and 59 for 1000m Therefore it can bededuced that with the addition of multipath fading alongwith critical radio propagation environment probability ofdelay being le50ms decreases by 19 to 30

Continuing on with multicell and EVA fading radioenvironment by increasing the network size the number ofvehicles in the awareness range increases and the end-to-enddelay increases as shown in Figure 8(a) Nevertheless theend-to-end delay remained below 100ms for up to 250mwith82 probability for 100 vehicles and similar rate was obtainedwith up to 150 vehicles When the awareness range increasesto 500m the probability was 80 for 100 vehicles while for150 vehicles further reduces to 73 For awareness range of750m and 1000m the probability reduced to 75 and 71 for100 vehicles while for 150 vehicles reduces to 69 and 60respectively

Changing the simulation model along with mobilitytraces to Glasgow city center the effect on delay performancewith 150 vehicles at 1 Hz and 10Hz is shown in Figures 8(b)and 8(c) respectively It is interesting to observe the effect ofchanging the mobility model to a European city Where each

city block is 400m in theMGmodel average city block size inGCC is around 200m increasing vehicle density eventuallygrowing the average number of neighbors for GCC modelThis can be observed for an awareness range of 250m theprobability of delay to be le50ms decreases from 77 to 68while for 1000m this deterioration goes from 54 to 37hence an overall decrease of about 9 to 17 for variousawareness ranges Similarly in Figure 8(c) it is observedthat changing the beacon frequency from 1Hz to 10Hz withan awareness range of 500m the probability of delay to bele50ms drastically decreases from 55 to 22 Whereas foran awareness range of 1000m degradation of about 29 isobserved Performance with a high beacon frequency suchas 10Hz is overall seen to be poor That is because packetsthat arrive late are discarded and are mainly sent by cell-edgeusers for which their neighborhood very likely includes usersat the cell edge Hence the delay from the uplink is indirectlyused as a spatial filtering of low quality cell-edge users Thissuggests that if the information is not useful when receivedafter 100ms the better use of resources is to drop the packetat the minimum latency requirement by the server

14 Wireless Communications and Mobile Computing

0010203040506070809

1

0 20 40 60 80 100 120 140 160 180 200

CDF

End-to-end delay (ms)

SAI = 11003816100381610038161003816Fi

1003816100381610038161003816 = 19

SAI = 31003816100381610038161003816Fi

1003816100381610038161003816 = 16

SAI = 51003816100381610038161003816Fi

1003816100381610038161003816 = 41

SAI = 61003816100381610038161003816Fi

1003816100381610038161003816 = 14

SAI = 71003816100381610038161003816Fi

1003816100381610038161003816 = 56

SAI = 91003816100381610038161003816Fi

1003816100381610038161003816 = 54

Without SAI 1003816100381610038161003816Fi1003816100381610038161003816 = 16

(a)

0010203040506070809

1

0 20 40 60 80 100 120 140 160 180 200

CDF

End-to-end delay (ms)

SAI = 11003816100381610038161003816Fi

1003816100381610038161003816 = 26

SAI = 31003816100381610038161003816Fi

1003816100381610038161003816 = 2

SAI = 51003816100381610038161003816Fi

1003816100381610038161003816 = 35

SAI = 61003816100381610038161003816Fi

1003816100381610038161003816 = 22

SAI = 71003816100381610038161003816Fi

1003816100381610038161003816 = 78

SAI = 91003816100381610038161003816Fi

1003816100381610038161003816 = 94

Without SAI 1003816100381610038161003816Fi1003816100381610038161003816 = 21

(b)

Figure 9 LTE end-to-end delay at average speed 40 kmh (GCC) and (a) 100 vehicles and (b) 150 vehicles

Next we observe and compare the results after theproposed algorithm implementation Figure 9 shows theCDF of the end-to-end delay for 100 and 150 vehicles atan average speed of 40 kmh with and without the imple-mentation of SAI algorithm The scenario being comparedhas the awareness range set to 500m and beacon frequencyto 10Hz We observe a significant decrease in the end-to-end delay after SAI algorithm implementation Probabilityfor the end-to-end delay to be less than 100ms is above80 in both scenarios whereas without the algorithm thisprobability is below 70 for 100 vehicles (Figure 9(a)) andbelow 60 for 150 vehicles (Figure 9(b)) The variation inthis delay between various SAIs is because of the differencein transmission parameters Larger awareness range leadingto higher number of neighbors |F119894| and higher beaconfrequency results in more congestion and large queuingtimes The reason for better delay values with SAI algorithmis mainly the result from restricting resources to the requiredtransmission parameters for safety applicationsmentioned inSection 32

Figure 10 shows the aggregate downlink goodput for thesame scenario investigated for end-to-end delay After theimplementation of SAI algorithm the downlink goodput for100 vehicles dropped from 746Mbps to 622Mbps and for150 vehicles it dropped from 2167Mbps to 1376Mbps Thissignificant decrease in the downlink direction is again due tothe restriction of unnecessary data dissemination from theVSA server to vehicles Eventually this decrease of goodputresults in less load on the LTE system The increase in thedownlink goodput for the scenario without SAI explains thehigh end-to-end delay observed in Figure 9 showing a burstof packet in the downlink leading to waiting time in thequeue

Furthermore the impact of regular cellular traffic on ourLTE vehicular network is studied in order to validate ourfindings and algorithm A broad picture of our observa-tions under different scenarios is shown in Figure 11 whereprobability of end-to-end delay being less than or equal to50ms is studied As expected this probability decreases byabout 3-4 when background traffic is added However with

100 150

Number of vehicles

25

20

15

10

5

0

Dow

nlin

k go

odpu

t (M

bps)

Without SAIWith SAI

746622

2167

1376

Figure 10 Downlink goodput for 100 and 150 vehicles at an averagespeed 40 kmh (GCC)

SAI algorithm probability of end-to-end delay being lessthan 50ms stays above 90 even after adding backgroundtraffic

Finally we include the SAI into MAC layer PDU con-trol elements to prove our concept of D2D like unicasttransmission between sender and receivers Modeling of thesystem is carried out in accordance with the description inSection 22 Figure 12 shows end-to-end delay for 100 and150 vehicles with and without the inclusion of SAI in MAClayer It is evident that 150 vehicles experience almost thesame probability for end-to-end delay being less than 50msas experienced by 100 vehicles This shows that with the useof MAC layer control elements around 50 more vehiclescan be accommodated by LTE network increasing the spec-trum efficiency eventually leading to higher capacity of thesystem

Wireless Communications and Mobile Computing 15

Simple case With backgroundtraffic

With SAI With backgroundand SAI

6065707580859095

100

150 vehicles100 vehicles

Prob

abili

ty o

f end

-to-e

ndde

layle

50

ms

Scenarios

Figure 11 Probability of end-to-end delay being le50ms for 100 and150 vehicles at 40 kmh in various scenarios

With SAI in APP layer With SAI in MAC layer80828486889092949698

100

100 vehicles150 vehicles

p

roba

bilit

y fo

r end

-to-e

ndde

layle50

ms

Figure 12 Probability of end-to-end delay being le50ms for 100 and150 vehicles with and without MAC layer inclusion

5 Conclusion

This paper proposes the safety application identifier conceptin the form of an algorithm that is tested and analyzedwithin the impact of vehicular urbanmulticell radio environ-ment employing multipath fading channels and backgroundtraffic The proposed algorithm uses dynamic adaptation oftransmission parameters in order to fulfill stringent vehicu-lar application requirements while minimizing latency andreducing system load

With the help of extensive system-level simulations thecrucial impact of channel modeling and mobility traces isevident with its influence on the received signal qualitydecreasing the probability of end-to-end delay to be le50msfor various awareness ranges by 19 to 30 and 9 to17 respectively Studying the impact of awareness rangeit is observed that with range up to 1000m for 1Hz 500mfor 2Hz and 250m for 10Hz beacon frequency the systemperformed well With the help of these findings the SAIalgorithm is proposed and implemented Probability of expe-riencing an end-to-end delay of less than 100ms increasedby around 20 and the downlink goodput decreased signif-icantly from 2167Mbps to 1376Mbps for 150 vehicles

Furthermore the impact of having background traffic isalso taken into consideration Results show that the systemis slightly affected by having regular background trafficHowever the probability of end-to-end delay being le50ms

still stays satisfactorily above 85 with the use of SAI Thispaper also proposes the inclusion of SAI inMAC layer controlelements producing a D2D type of communication withinvehicles After modeling this system it was found that thesame amount of system resources and requirements are metwith additional 50 vehicles increasing the capacity of ourLTE vehicular network Finally in future works we plan tointegrate SAI incorporated within MAC layer in a hetero-geneous network with DSRC while exploring the networkslicing in 5G cellular networks for vehicular communicationscomplementing each otherrsquos weaknesses and strengths inorder to meet the vehicular network requirements

Conflicts of Interest

The authors declare that they have no conflicts of interest

Acknowledgments

Marvin Sanchez would like to acknowledge the support ofLAMENITEC Erasmus Mundus Program of the EuropeanUnion to be a guest researcher at GCU Results were obtainedusing the EPSRC funded ARCHIE-WeSt High PerformanceComputer (httpswwwarchie-westacuk) EPSRC Grantno EPK0005861

References

[1] S Sesia I Toufik and M Baker LTE the UMTS long termevolution from theory to practice Wiley Publishing 2009

[2] H Hartenstein and K P Laberteaux ldquoA tutorial survey onvehicular ad hoc networksrdquo IEEE Communications Magazinevol 46 no 6 pp 164ndash171 2008

[3] D Jiang and L Delgrossi ldquoTowards an international standardfor wireless access in vehicular environmentsrdquo in Proceedings ofthe 67th IEEE Vehicular Technology Conference (VTC rsquo08) pp2036ndash2040 May 2008

[4] S Al-Sultan M M Al-Doori A H Al-Bayatti and H ZedanldquoA comprehensive survey on vehicular Ad Hoc networkrdquoJournal of Network and Computer Applications vol 37 no 1 pp380ndash392 2014

[5] K Tanuja T Sushma M Bharathi and K Arun ldquoA surveyon VANET technologiesrdquo International Journal of ComputerApplications vol 121 no 18 2015

[6] H Y Kim D M Kang J H Lee and T M Chung ldquoA per-formance evaluation of cellular network suitability for VANETrdquoWorld Academy of Science Engineering and Technology Interna-tional Science Index vol 64 no 6 pp 124ndash127 2012

[7] A Fasbender M Gerdes and S Smets ldquoCellular NetworkingTechnologies in ITS Solutions Opportunities and Challengesrdquopp 1ndash13 Springer Berlin Germany 2012

[8] ldquoCellular coverage and tower map for ee network in glasgowrdquohttpswwwcellmappernetmap

[9] GAraniti C CampoloMCondoluci A Iera andAMolinaroldquoLTE for vehicular networking a surveyrdquo IEEECommunicationsMagazine vol 51 no 5 pp 148ndash157 2013

[10] A Vinel ldquo3GPP LTE versus IEEE 80211pWAVE which tech-nology is able to support cooperative vehicular safety applica-tionsrdquo IEEE Wireless Communications Letters vol 1 no 2 pp125ndash128 2012

Wireless Communications and Mobile Computing 7

Vehicle 1 eNB Vehicle 2 Vehicle N

Initial networkaccess

Registered to network

Registered to network

PRACH RA preamble

PDSCH RA responseAssigns C-RNTI 1

PUSCH RRC connection request

PDSCH connection setup

PUSCH RRC connection setup complete

Contains eUE 1location

Assuming norandom access

contention

Network assisted GNSS location update

Network assisted GNSS location update

Beacon for transmission

PUCCH (SR)

PDCCH UL grant for BSR

PUSCH (BSR containing MAC PDU)Contains SAI

Resource allocation

PDCCH TX grant PDCCH RX grant PDCCH RX grant

PUCCH (ACKNACK)PHICH (ACKNACK)

PUSCH beacon transmitted to nUE 2 middot middot middot nUE N

Figure 3 LTE signaling with MAC control elements in BSR

Header Controlelements Payload (SDU) Padding

Subheader1

Subheader2

SubheaderK

MAC Ctrlelement 1

MAC Ctrlelement 2

MAC Ctrlelement N

R ER LCID = 01100 Safety application identifier

MAC PDU

Figure 4 SAI in MAC control elements

8 Wireless Communications and Mobile Computing

500m

S-GWP-GWVSA server

400m

500m1 km

500m

Figure 5 MG model covered by 4 sites and 3 cellssite with 1000mintersite distance (ISD)

GCC This change in the network design considers a morerealistic network model employing the site configurationused by UKrsquos mobile operator EE in Glasgow [8] The LTEfunctionality is implemented through the LTE EPC networksimulator (LENA) module [38] that includes highly detailedeNB and UE functionalities as specified by 3GPP standardsTo manage handover and Internet connections eNBs areinterconnected through their X2 interfaces and connected tothe EPC through S1-U interface Interconnection from the P-GW to the VSA server residing in the EPC is modeled usingan error free 10 Gbps point-to-point link and TCPIP version4

The granularity of the LENA module is up to resourceblock (RB) level including the user plane control planeand reference signals used for coherent reception The safetymessages are UDP packets and often might be segmentedat the radio link control (RLC) and MAC layers to betransmitted on the transport shared channel (S-CH) passedto the physical (PHY) layer The size of transport blocks(TB) used at the PHY layer is decided by the MAC sublayerdepending on the channel quality and scheduling decisionsand may be transmitted using several RBs utilizing a mod-ulation and coding scheme (MCS) [38] Transmissions arescheduled every 1ms corresponding to one subframe usingProportional Fair Algorithm which is one of the approachessuitable for applications that include latency constraints [1Chapter 12] For downlink the algorithm utilizes userrsquoschannel quality indicator (CQI isin [0 sdot sdot sdot 15]) measurementreports sent from the vehicles For uplink channel qualitymeasurements are done by the eNB through scheduling

periodic vehicle transmissions of sounding reference signals(SRS)The radio resource control (RRC)module in LENAns-3 assigns the periodicity of SRS as a function of the numberof UEs attached to an eNB according to 3GPP UE-specificprocedure [37]

Each site modeled by an eNB uses 10W (40 dBm) totaltransmission power and a cosine antenna sector model with65∘ half power beamwidth (HPBW) [39] and azimuthaldirection 0∘ 120∘ or 240∘ with respect to north and 1mintersector space We assume utilization of 2 times 10MHzbandwidth whichmeans there are119872 = 50RBs for the uplinkand for the downlink Therefore assuming that the radiopropagation channel is constant in subframe 119905 if vehicle 119894 isscheduled to receive from its serving cell 119895 on resource block119898 isin [0 sdot sdot sdot 49] the received signal power 119875119895119894 can be modeledas follows

119875119895119894 (119898 119905) = 119875119895 (119898 119905) 119866119895 (120579119895119894)119866119894 (120579119894119895)119871119895119894 (119898 119905) (2)

where 119875119895(119898 119905) is the transmitted power for resource block119898119866119895(120579119895119894) is the antenna gain of cell 119895 in the direction of user 119894119866119894(120579119894119895) is the antenna gain of user 119894 in the direction of cell 119895and 119871119895119894 is the path loss from cell 119895 to user 119894

In the single cell scenario there is no intercell interferenceand mutually exclusive RBs are assigned to users Howeverin a more general deployed urban environment the availablespectrum is utilized by multiple cells in the service area andthe intercell interference is one important limiting aspectespecially for cell-edge users [1 Chapter 12]

In order to allocate usersrsquo transmissions while at the sametime maximizing the signal to interference plus noise ratio(SINR) for each RB frequency reuse and handover are usedFrequency allocation has an impact on the achievable systemcapacity and delay since the transport block size MCS andthe transmission schedule are dependent on the SINR [37Chapter 9] In ns-3 if several RBs are used to transmit a TBthe vector composed of the SINRs on each RB is used toevaluate the TB error probability The SINR in RB 119898 whenuser 119894 is scheduled to receive from serving cell 119895 can bewrittenas follows

SINR119894 (119898 119905)= 119875119895 (119898 119905) 119866119895 (120579119895119894)119866119894 (120579119894119895) 119871119895119894 (119898 119905)1198730 + sumforall119897 =119895 (119875119897 (119898 119905) 119866119897 (120579119897119894) 119866119894 (120579119894119897) 119871 119897119894 (119898 119905))

(3)

where1198730 is the noise powerUEs report the SINR mapped to the CQI value corre-

sponding to the MCS (isin [0 sdot sdot sdot 28]) that ensures TB error ratele 01 according to the standard [40 Section 723] In ns-3 thedownlink SINR is used to generate periodic wideband CQIsfeedback (ie an average value of all RBs) and inband CQIs(ie a value for each RB) We set the CQI to be calculated bycombining the received signal power of the reference signalsto the interference power from the physical downlink sharedchannel as an approximation to (3)

For dynamic frequency allocation we use the distributedfractional frequency reuse algorithm that utilizes intercell

Wireless Communications and Mobile Computing 9

S-GWP-GWVSA

TCC

Figure 6 GCC model covered by 6 sites and 3 cellssite where eNB location is determined using UK operator EE mast data [8]

interference coordination (ICIC) [41] The eNB adaptivelyselects a set of RBs referred to as cell-edge subband (set to10RBs) based on information exchanged with its neighborsTherefore according to ICIC a vehicle should be allocatedin the cell-edge subband if its serving cell reference signalreceived quality (RSRQ) gets lower than minus10 dB

Handover decisions are also performed by eNBs based onconfigurable event-triggeredmeasurement reports includingthe RSRQ and the reference signal received power (RSRP)The measurements are performed by UEs usually in orthog-onal frequency division multiplexing (OFDM) symbols car-rying reference signals (RS) for antenna port 0 In ns-3 since the channel is assumed flat over an RB that iscomposed of 12 subcarriers all resource elements (RE) in anRB have the same power Hence assuming orthogonal RSreception the RSRP for cell 119895 measured by vehicle 119894 is givenby

RSRP119895119894 (119905) = sum119872minus1119898=0 sum119870minus1119896=0 (119875119895119894 (119898 119896 119905) 119870)119872= sum119872minus1119898=0 (119875119895119894 (119898 119905) 12)119872

(4)

where 119875119895119894(119898 119896 119905) is the received power of RS 119896 in RB 119898 119872is the number of RBs and 119870 is the number of RS measuredin the RB The average value is passed to higher layers every200ms therefore the index 119905 is omitted for simplicity

The RSRQ for serving cell 119895measured by vehicle 119894 can becomputed by the following [42]

RSRQ119895119894 = 119872 times RSRP119895119894RSSI119895119894

(5)

where RSSI119895119894 is the received signal strength indicator (RSSI)measured by UE 119894when served by cell 119895 and according to the

10 Wireless Communications and Mobile Computing

Table 3 LTE simulation parameters

Parameter ValueSimulation time 100 seconds

Road model Manhattan grid model (MG) (two way roads)Glasgow City Center (GCC) (2 km times 2 km)

LTE network 4 sites with 3 cellssite 1000m ISD for MG6 sites with 3 cellssite UK Operator EE mast data [8]

Transmission power eNB 40 dBm UE 23 dBmCarrier frequency DLUL 2115MHz1715MHzChannel bandwidth 2 times 10MHz (2 times 50RBs)Noise Figure eNB 5 dB UE 9 dBUE antenna model Isotropic (0 dBi)eNB antenna model 15 dB Cosine model 65∘ HPBWScheduling algorithm Proportional Fair

Handover algorithm A2A4RSRQ RSRQ threshold minus5 dBand NeighbourCellOffset = 2 (1 dB)

Frequency reuse Distributed Fractional Freq Reuse

Path loss model LogDistance (120572 = 3) and3GPP Extended Vehicular A model

Safety message format 256 bytes UDPNumber of vehicles 75 100 125 150Average vehiclersquos speed 20 kmh 40 kmhBeacon frequency 1Hz 10HzAwareness range (119877) 100m 250m 500m 750m 1000m

standard comprises the linear average of the total receivedpower observed by the UE from all sources includingcochannel serving and nonserving cells adjacent channelinterference thermal noise and so forth [42] The RSSImeasurement model in ns-3 is computed considering two REper RB (the one that carries the RS) [37] that is

RSSI119895119894 (119905)= 119872minus1sum119898=0

2 times (119873012 +sumforall119897

119875119897 (119898 119905) 119866119897 (120579119897119894) 119866119894 (120579119894119897)12 times 119871 119897119894 (119898 119905) ) (6)

Furthermore in order to perform handover decisions thereported RSRQ for neighboring cell 119897 when vehicle 119894 is con-nected through serving cell 119895 is computed by the following

RSRQ119897119894 = 119872 times RSRP119897119894RSSI119895119894

(7)

The RSRQs for serving and neighboring cells are usedwith theA2A4RsrqHandoverAlgorithm to trigger LTE eventsA2 and A4 [43 Section 554] Event A2 is set to be triggeredwhen the RSRQ of the serving cell becomes worse than minus5 dB(ServingCellThreshold parameter) While event A2 is trueevent A4 is triggered if a neighboring cell becomes betterthan a threshold which is set by the algorithm to a verylow value for the trigger criteria to be true Thus the mea-surement reports received by the eNB are used to consider aneighboring cell as a target cell for handover only if its RSRQ

is 1 dB higher than the serving cell (NeighbourCellOffsetparameter) Table 3 summarizes the simulation parametersfor LTE

31 Vehicular Radio Urban Environment As mentioned inSection 11 previous evaluations lack the use of multipathand multicell environments Use of proper channel modelingis essential in evaluating system models as it poses close toreality system degradations The system model presented inthis paper consists of 6 sites with 3 cells per site Users areassumed to have isotropic antenna (119866119894(120579119894119895) = 0 dBi) and thepath loss (119871119895119894) ismodeled using the Log-distance propagationwith shadow exponent 119899 = 3 [44] plus 3GPP EVA multipathfading modelL119895119894 [34] and is computed by the following

119871119895119894 (119898 119905) = 1198710 + 10119899 log10 (119889119895119894 (119905)1198890 ) +L119895119894 (119898 119905) dB (8)

where 1198710 = 20 log10(120582(41205871198890)) 120582 is the wavelength and1198890 is the reference distance assumed to be 10m in oursimulations Traces for EVA were generated using MATLABwhile modifying the classical Doppler spectrum for each tapat vehicular speed (V) of 40 kmh and 60 kmh as

119878 (119891) prop 1(1 minus (119891119891119863)2)05 (9)

where 119891119863 = V120582 These EVA traces specified by 3GPP in[34] improvise the NLOS conditions relative to the speed

Wireless Communications and Mobile Computing 11

of UE and the transmission bandwidth Simulation of themultipath fading component in ns-3 uses the trace startingat a random point along the time domain for a given user andserving cell The channel object created with the rayleighchanfunction is used for filtering a discrete-time impulse signal inorder to obtain the channel impulse response The filteringis repeated for different Transmission Time Interval (TTI)thus yielding subsequent time-correlated channel responses(one per TTI) This channel response is then processedwith the pwelch function for obtaining its power spectraldensity values which are then saved in a file with the formatcompatible with the simulator model

32 SAI Implementation Model Archer and Vogel in [45]carried out a survey on traffic safety problems in urbanareas It was concluded that overtaking tends to occur lessfrequently within urban areas where the speed limit isless than 50 kmh Lane changing however occurs quitefrequently within urban areas due to low speed limits Thenumber of rear-end accidents is greater within urban areasthan in rural areas and similarly due to higher level ofcongestion and higher number of traffic junctions there isa greater number of opportunities for turning and crossingaccidents within urban area According to another survey of4500 crashes by the Insurance Institute of Highway Safety[46] 13 of accidents are during lane change maneuvers12 are intersection collisions 22 are due to traffic lightviolation 9 are head on collisions 18 include left turningcrashes 12 are due to emergency vehicles parked on blindturns and 14 are due to over speeding In the light ofthese statistics for our adopted urban environment the SAIs(Table 2) we utilize in our simulations are 1 (25) 3 (22) 5(9) 6 (18) 7 (12) and 9 (14)

33 Background Traffic In order to model close to realscenarios we added background traffic into our simulationsModeling of such traffic is done bymaking 50of the existingvehicles use voice application and 25 stream videos Voiceapplication is modeled by implementing constant bit rate(CBR) traffic generator on G711 codec using onoffapplicationon ns-3 [37] Voice payload is 172 bytes and the data ratechosen is 688 Kbps The reason for using a high data rate isto cater worst case scenario where an operator would preferhaving a high quality voice application At the same timevideo streaming is being carried out using evalvid simulatormodule on H263 codec with a payload of 1460 bytes at a datarate of 122Mbps [47] Both these voice and video applicationsare assigned their respective QCIs

34 Performance Measures The primary performance mea-sures used are the end-to-end packet delay the packet deliveryratio and the downlink goodputThe goodput is defined as thenumber of useful information bits received at the applicationlayer per unit of time

In addition we use the notation |F119894| to represent theaverage number of neighbors calculated over the number ofreceivers whenever a beacon is disseminated The end-to-enddelay of a beacon (119863119894119896) is defined as the time measured fromthe start of its transmission at vehicle 119894rsquos application layer until

its successful reception at vehicle 119896rsquos application layer PDR isdefined as the beacon success rate within the awareness range119877 [48] that is for a beacon sent fromvehicle 119894 its PDR is givenby

119875119894 (119877) = sum119896isinF119894

1198681198941198961003816100381610038161003816F1198941003816100381610038161003816 1003816100381610038161003816F1198941003816100381610038161003816 gt 0 (10)

where | sdot | indicates the cardinality of the set and 119868119894119896 (isin 0 1)is set to 1 if the beacon is received by vehicle 1198964 Simulation Results

The success of cellular networks relies on the scalability ofthe system capacity achieved by reusing the spectrum Thecommon approach implemented by operators is to caterthe capacity needs as the traffic demand grows Thereforefirst we analyze the capacity limit which results from thesingle cell case compared to multiple cells using the Friisradio propagation environment and afterwards we analyzethe impact of the urban radio environment as a function ofthe awareness range

Figure 7(a) shows the cumulative distribution function(CDF) of the end-to-end delay when the network is com-posed of 100 vehicles moving at an average speed of 40 kmhThe service area is covered using a single cell with an isotropicantenna located at the upper left corner of Manhattan Gridmodel as utilized by [11] in a Friis radio environment wherevehicles transmit at 1 Hz beacon frequency We observe thatfor awareness range of up to 500m the end-to-end delay isle100ms with around 95 probability Probability of higherend-to-end delay increases more between 250m to 500mas compared to the increase from 500m to 750m That ismainly the result of the increase in traffic but also as theawareness range increases cars located on roads near the edgeof the Manhattan model will have neighbors towards the celldirection with better signal quality which is further exploitedby the scheduling algorithm Figure 7(a) also shows that thereis no significant increase in end-to-end delay from 750mto 1000m This is due to the border effect incurred by thecar located at the edge of the model having less number ofneighbors when increasing the awareness range as comparedto the cars located at the center of the cell Nevertheless theseresults show that up to 100 vehicles were possible for thesingle cell case at a beacon frequency of 1Hz under Friis radioenvironment In order to analyze the intercell interferenceimpact on the delay while increasing the number of vehiclesand beacon frequency we increase the network capacity bydeploying 4 sites with 3 sectorssite (the average number ofvehicles per sector is reduced to 125) Figure 7(b) shows thatthe end-to-enddelay is then further improved to below 50mswith 91 success rate when awareness range of 1000m at thebeacon frequency of 1Hz was utilized We can also noticethat in this case there are some beacons that experiencedlonger delays compared to the single cell scenario causedby low quality conditions of cell-edge users due to intercellinterference and handover However with 1000m 934 ofthe forwarded beacons were received with end-to-end delayle 100ms and 982 for 250m

12 Wireless Communications and Mobile Computing

0 50 100 150 200 250 3000

01

02

03

04

05

06

07

08

09

1CD

F

End-to-end delay (ms)

R = 100

R = 250

R = 500

R = 750

R = 1000

(a)

0

01

02

03

04

05

06

07

08

09

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 1

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 35

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 14

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 25

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 41

(b)

0

01

02

03

04

05

06

07

08

09

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 1

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 35

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 14

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 25

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 41

(c)

Figure 7 LTE end-to-end delay for 100 vehicles at a beacon frequency 1Hz and average speed 40 kmh with Friis radio environment (a)single cell (b) multicell (4 sites 3 sectorssite) (c) multicell with EVA fading radio environment

The results in Figure 7(b) show that under the Friis radiopropagation environment the capacity increases and thesystem performance is not significantly affected in terms ofthe end-to-end delay in presence of intercell interference andhandover since the scheduler effectively managed interfer-ence by time and frequency allocations However in an urbanenvironment it is expected that fading caused by buildingsand Doppler shift due to the relative velocity of vehicles havea very important impact on the received signal quality [49]Figure 7(c) shows the results when the radio propagationenvironment was changed to Log-distance-dependent and

3GPP EVA under the same site configurations Under thisradio environment the probability for the end-to-end delayto be less than or equal to 100ms was around 76ndash78 forawareness range between 100m and 500m 70 for 750mand 66 for 1000m Hence the significant effect on the end-to-end delay is because of path loss and frequency selectivefading imposed by the EVA fading environmentThe end-to-enddelay for beacons to and fromcell-edge users significantlyincreases Similarly for inner cell users the end-to-end delayremained close to the Friis results and for up to 500m theend-to-end delay was le50ms with 74 to 71 probability

Wireless Communications and Mobile Computing 13

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 17

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 53

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 20

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 37

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 62

(a)

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 11

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 69

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 21

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 47

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 73

(b)

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 13

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 63

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 24

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 47

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 74

(c)

Figure 8 End-to-end delay for sample networks at average speed 40 kmh with 150 vehicles at beacon frequency (a) 1Hz with MG model(b) 1Hz with GCC model (c) 10Hz with GCC model

65 for 750m and 59 for 1000m Therefore it can bededuced that with the addition of multipath fading alongwith critical radio propagation environment probability ofdelay being le50ms decreases by 19 to 30

Continuing on with multicell and EVA fading radioenvironment by increasing the network size the number ofvehicles in the awareness range increases and the end-to-enddelay increases as shown in Figure 8(a) Nevertheless theend-to-end delay remained below 100ms for up to 250mwith82 probability for 100 vehicles and similar rate was obtainedwith up to 150 vehicles When the awareness range increasesto 500m the probability was 80 for 100 vehicles while for150 vehicles further reduces to 73 For awareness range of750m and 1000m the probability reduced to 75 and 71 for100 vehicles while for 150 vehicles reduces to 69 and 60respectively

Changing the simulation model along with mobilitytraces to Glasgow city center the effect on delay performancewith 150 vehicles at 1 Hz and 10Hz is shown in Figures 8(b)and 8(c) respectively It is interesting to observe the effect ofchanging the mobility model to a European city Where each

city block is 400m in theMGmodel average city block size inGCC is around 200m increasing vehicle density eventuallygrowing the average number of neighbors for GCC modelThis can be observed for an awareness range of 250m theprobability of delay to be le50ms decreases from 77 to 68while for 1000m this deterioration goes from 54 to 37hence an overall decrease of about 9 to 17 for variousawareness ranges Similarly in Figure 8(c) it is observedthat changing the beacon frequency from 1Hz to 10Hz withan awareness range of 500m the probability of delay to bele50ms drastically decreases from 55 to 22 Whereas foran awareness range of 1000m degradation of about 29 isobserved Performance with a high beacon frequency suchas 10Hz is overall seen to be poor That is because packetsthat arrive late are discarded and are mainly sent by cell-edgeusers for which their neighborhood very likely includes usersat the cell edge Hence the delay from the uplink is indirectlyused as a spatial filtering of low quality cell-edge users Thissuggests that if the information is not useful when receivedafter 100ms the better use of resources is to drop the packetat the minimum latency requirement by the server

14 Wireless Communications and Mobile Computing

0010203040506070809

1

0 20 40 60 80 100 120 140 160 180 200

CDF

End-to-end delay (ms)

SAI = 11003816100381610038161003816Fi

1003816100381610038161003816 = 19

SAI = 31003816100381610038161003816Fi

1003816100381610038161003816 = 16

SAI = 51003816100381610038161003816Fi

1003816100381610038161003816 = 41

SAI = 61003816100381610038161003816Fi

1003816100381610038161003816 = 14

SAI = 71003816100381610038161003816Fi

1003816100381610038161003816 = 56

SAI = 91003816100381610038161003816Fi

1003816100381610038161003816 = 54

Without SAI 1003816100381610038161003816Fi1003816100381610038161003816 = 16

(a)

0010203040506070809

1

0 20 40 60 80 100 120 140 160 180 200

CDF

End-to-end delay (ms)

SAI = 11003816100381610038161003816Fi

1003816100381610038161003816 = 26

SAI = 31003816100381610038161003816Fi

1003816100381610038161003816 = 2

SAI = 51003816100381610038161003816Fi

1003816100381610038161003816 = 35

SAI = 61003816100381610038161003816Fi

1003816100381610038161003816 = 22

SAI = 71003816100381610038161003816Fi

1003816100381610038161003816 = 78

SAI = 91003816100381610038161003816Fi

1003816100381610038161003816 = 94

Without SAI 1003816100381610038161003816Fi1003816100381610038161003816 = 21

(b)

Figure 9 LTE end-to-end delay at average speed 40 kmh (GCC) and (a) 100 vehicles and (b) 150 vehicles

Next we observe and compare the results after theproposed algorithm implementation Figure 9 shows theCDF of the end-to-end delay for 100 and 150 vehicles atan average speed of 40 kmh with and without the imple-mentation of SAI algorithm The scenario being comparedhas the awareness range set to 500m and beacon frequencyto 10Hz We observe a significant decrease in the end-to-end delay after SAI algorithm implementation Probabilityfor the end-to-end delay to be less than 100ms is above80 in both scenarios whereas without the algorithm thisprobability is below 70 for 100 vehicles (Figure 9(a)) andbelow 60 for 150 vehicles (Figure 9(b)) The variation inthis delay between various SAIs is because of the differencein transmission parameters Larger awareness range leadingto higher number of neighbors |F119894| and higher beaconfrequency results in more congestion and large queuingtimes The reason for better delay values with SAI algorithmis mainly the result from restricting resources to the requiredtransmission parameters for safety applicationsmentioned inSection 32

Figure 10 shows the aggregate downlink goodput for thesame scenario investigated for end-to-end delay After theimplementation of SAI algorithm the downlink goodput for100 vehicles dropped from 746Mbps to 622Mbps and for150 vehicles it dropped from 2167Mbps to 1376Mbps Thissignificant decrease in the downlink direction is again due tothe restriction of unnecessary data dissemination from theVSA server to vehicles Eventually this decrease of goodputresults in less load on the LTE system The increase in thedownlink goodput for the scenario without SAI explains thehigh end-to-end delay observed in Figure 9 showing a burstof packet in the downlink leading to waiting time in thequeue

Furthermore the impact of regular cellular traffic on ourLTE vehicular network is studied in order to validate ourfindings and algorithm A broad picture of our observa-tions under different scenarios is shown in Figure 11 whereprobability of end-to-end delay being less than or equal to50ms is studied As expected this probability decreases byabout 3-4 when background traffic is added However with

100 150

Number of vehicles

25

20

15

10

5

0

Dow

nlin

k go

odpu

t (M

bps)

Without SAIWith SAI

746622

2167

1376

Figure 10 Downlink goodput for 100 and 150 vehicles at an averagespeed 40 kmh (GCC)

SAI algorithm probability of end-to-end delay being lessthan 50ms stays above 90 even after adding backgroundtraffic

Finally we include the SAI into MAC layer PDU con-trol elements to prove our concept of D2D like unicasttransmission between sender and receivers Modeling of thesystem is carried out in accordance with the description inSection 22 Figure 12 shows end-to-end delay for 100 and150 vehicles with and without the inclusion of SAI in MAClayer It is evident that 150 vehicles experience almost thesame probability for end-to-end delay being less than 50msas experienced by 100 vehicles This shows that with the useof MAC layer control elements around 50 more vehiclescan be accommodated by LTE network increasing the spec-trum efficiency eventually leading to higher capacity of thesystem

Wireless Communications and Mobile Computing 15

Simple case With backgroundtraffic

With SAI With backgroundand SAI

6065707580859095

100

150 vehicles100 vehicles

Prob

abili

ty o

f end

-to-e

ndde

layle

50

ms

Scenarios

Figure 11 Probability of end-to-end delay being le50ms for 100 and150 vehicles at 40 kmh in various scenarios

With SAI in APP layer With SAI in MAC layer80828486889092949698

100

100 vehicles150 vehicles

p

roba

bilit

y fo

r end

-to-e

ndde

layle50

ms

Figure 12 Probability of end-to-end delay being le50ms for 100 and150 vehicles with and without MAC layer inclusion

5 Conclusion

This paper proposes the safety application identifier conceptin the form of an algorithm that is tested and analyzedwithin the impact of vehicular urbanmulticell radio environ-ment employing multipath fading channels and backgroundtraffic The proposed algorithm uses dynamic adaptation oftransmission parameters in order to fulfill stringent vehicu-lar application requirements while minimizing latency andreducing system load

With the help of extensive system-level simulations thecrucial impact of channel modeling and mobility traces isevident with its influence on the received signal qualitydecreasing the probability of end-to-end delay to be le50msfor various awareness ranges by 19 to 30 and 9 to17 respectively Studying the impact of awareness rangeit is observed that with range up to 1000m for 1Hz 500mfor 2Hz and 250m for 10Hz beacon frequency the systemperformed well With the help of these findings the SAIalgorithm is proposed and implemented Probability of expe-riencing an end-to-end delay of less than 100ms increasedby around 20 and the downlink goodput decreased signif-icantly from 2167Mbps to 1376Mbps for 150 vehicles

Furthermore the impact of having background traffic isalso taken into consideration Results show that the systemis slightly affected by having regular background trafficHowever the probability of end-to-end delay being le50ms

still stays satisfactorily above 85 with the use of SAI Thispaper also proposes the inclusion of SAI inMAC layer controlelements producing a D2D type of communication withinvehicles After modeling this system it was found that thesame amount of system resources and requirements are metwith additional 50 vehicles increasing the capacity of ourLTE vehicular network Finally in future works we plan tointegrate SAI incorporated within MAC layer in a hetero-geneous network with DSRC while exploring the networkslicing in 5G cellular networks for vehicular communicationscomplementing each otherrsquos weaknesses and strengths inorder to meet the vehicular network requirements

Conflicts of Interest

The authors declare that they have no conflicts of interest

Acknowledgments

Marvin Sanchez would like to acknowledge the support ofLAMENITEC Erasmus Mundus Program of the EuropeanUnion to be a guest researcher at GCU Results were obtainedusing the EPSRC funded ARCHIE-WeSt High PerformanceComputer (httpswwwarchie-westacuk) EPSRC Grantno EPK0005861

References

[1] S Sesia I Toufik and M Baker LTE the UMTS long termevolution from theory to practice Wiley Publishing 2009

[2] H Hartenstein and K P Laberteaux ldquoA tutorial survey onvehicular ad hoc networksrdquo IEEE Communications Magazinevol 46 no 6 pp 164ndash171 2008

[3] D Jiang and L Delgrossi ldquoTowards an international standardfor wireless access in vehicular environmentsrdquo in Proceedings ofthe 67th IEEE Vehicular Technology Conference (VTC rsquo08) pp2036ndash2040 May 2008

[4] S Al-Sultan M M Al-Doori A H Al-Bayatti and H ZedanldquoA comprehensive survey on vehicular Ad Hoc networkrdquoJournal of Network and Computer Applications vol 37 no 1 pp380ndash392 2014

[5] K Tanuja T Sushma M Bharathi and K Arun ldquoA surveyon VANET technologiesrdquo International Journal of ComputerApplications vol 121 no 18 2015

[6] H Y Kim D M Kang J H Lee and T M Chung ldquoA per-formance evaluation of cellular network suitability for VANETrdquoWorld Academy of Science Engineering and Technology Interna-tional Science Index vol 64 no 6 pp 124ndash127 2012

[7] A Fasbender M Gerdes and S Smets ldquoCellular NetworkingTechnologies in ITS Solutions Opportunities and Challengesrdquopp 1ndash13 Springer Berlin Germany 2012

[8] ldquoCellular coverage and tower map for ee network in glasgowrdquohttpswwwcellmappernetmap

[9] GAraniti C CampoloMCondoluci A Iera andAMolinaroldquoLTE for vehicular networking a surveyrdquo IEEECommunicationsMagazine vol 51 no 5 pp 148ndash157 2013

[10] A Vinel ldquo3GPP LTE versus IEEE 80211pWAVE which tech-nology is able to support cooperative vehicular safety applica-tionsrdquo IEEE Wireless Communications Letters vol 1 no 2 pp125ndash128 2012

8 Wireless Communications and Mobile Computing

500m

S-GWP-GWVSA server

400m

500m1 km

500m

Figure 5 MG model covered by 4 sites and 3 cellssite with 1000mintersite distance (ISD)

GCC This change in the network design considers a morerealistic network model employing the site configurationused by UKrsquos mobile operator EE in Glasgow [8] The LTEfunctionality is implemented through the LTE EPC networksimulator (LENA) module [38] that includes highly detailedeNB and UE functionalities as specified by 3GPP standardsTo manage handover and Internet connections eNBs areinterconnected through their X2 interfaces and connected tothe EPC through S1-U interface Interconnection from the P-GW to the VSA server residing in the EPC is modeled usingan error free 10 Gbps point-to-point link and TCPIP version4

The granularity of the LENA module is up to resourceblock (RB) level including the user plane control planeand reference signals used for coherent reception The safetymessages are UDP packets and often might be segmentedat the radio link control (RLC) and MAC layers to betransmitted on the transport shared channel (S-CH) passedto the physical (PHY) layer The size of transport blocks(TB) used at the PHY layer is decided by the MAC sublayerdepending on the channel quality and scheduling decisionsand may be transmitted using several RBs utilizing a mod-ulation and coding scheme (MCS) [38] Transmissions arescheduled every 1ms corresponding to one subframe usingProportional Fair Algorithm which is one of the approachessuitable for applications that include latency constraints [1Chapter 12] For downlink the algorithm utilizes userrsquoschannel quality indicator (CQI isin [0 sdot sdot sdot 15]) measurementreports sent from the vehicles For uplink channel qualitymeasurements are done by the eNB through scheduling

periodic vehicle transmissions of sounding reference signals(SRS)The radio resource control (RRC)module in LENAns-3 assigns the periodicity of SRS as a function of the numberof UEs attached to an eNB according to 3GPP UE-specificprocedure [37]

Each site modeled by an eNB uses 10W (40 dBm) totaltransmission power and a cosine antenna sector model with65∘ half power beamwidth (HPBW) [39] and azimuthaldirection 0∘ 120∘ or 240∘ with respect to north and 1mintersector space We assume utilization of 2 times 10MHzbandwidth whichmeans there are119872 = 50RBs for the uplinkand for the downlink Therefore assuming that the radiopropagation channel is constant in subframe 119905 if vehicle 119894 isscheduled to receive from its serving cell 119895 on resource block119898 isin [0 sdot sdot sdot 49] the received signal power 119875119895119894 can be modeledas follows

119875119895119894 (119898 119905) = 119875119895 (119898 119905) 119866119895 (120579119895119894)119866119894 (120579119894119895)119871119895119894 (119898 119905) (2)

where 119875119895(119898 119905) is the transmitted power for resource block119898119866119895(120579119895119894) is the antenna gain of cell 119895 in the direction of user 119894119866119894(120579119894119895) is the antenna gain of user 119894 in the direction of cell 119895and 119871119895119894 is the path loss from cell 119895 to user 119894

In the single cell scenario there is no intercell interferenceand mutually exclusive RBs are assigned to users Howeverin a more general deployed urban environment the availablespectrum is utilized by multiple cells in the service area andthe intercell interference is one important limiting aspectespecially for cell-edge users [1 Chapter 12]

In order to allocate usersrsquo transmissions while at the sametime maximizing the signal to interference plus noise ratio(SINR) for each RB frequency reuse and handover are usedFrequency allocation has an impact on the achievable systemcapacity and delay since the transport block size MCS andthe transmission schedule are dependent on the SINR [37Chapter 9] In ns-3 if several RBs are used to transmit a TBthe vector composed of the SINRs on each RB is used toevaluate the TB error probability The SINR in RB 119898 whenuser 119894 is scheduled to receive from serving cell 119895 can bewrittenas follows

SINR119894 (119898 119905)= 119875119895 (119898 119905) 119866119895 (120579119895119894)119866119894 (120579119894119895) 119871119895119894 (119898 119905)1198730 + sumforall119897 =119895 (119875119897 (119898 119905) 119866119897 (120579119897119894) 119866119894 (120579119894119897) 119871 119897119894 (119898 119905))

(3)

where1198730 is the noise powerUEs report the SINR mapped to the CQI value corre-

sponding to the MCS (isin [0 sdot sdot sdot 28]) that ensures TB error ratele 01 according to the standard [40 Section 723] In ns-3 thedownlink SINR is used to generate periodic wideband CQIsfeedback (ie an average value of all RBs) and inband CQIs(ie a value for each RB) We set the CQI to be calculated bycombining the received signal power of the reference signalsto the interference power from the physical downlink sharedchannel as an approximation to (3)

For dynamic frequency allocation we use the distributedfractional frequency reuse algorithm that utilizes intercell

Wireless Communications and Mobile Computing 9

S-GWP-GWVSA

TCC

Figure 6 GCC model covered by 6 sites and 3 cellssite where eNB location is determined using UK operator EE mast data [8]

interference coordination (ICIC) [41] The eNB adaptivelyselects a set of RBs referred to as cell-edge subband (set to10RBs) based on information exchanged with its neighborsTherefore according to ICIC a vehicle should be allocatedin the cell-edge subband if its serving cell reference signalreceived quality (RSRQ) gets lower than minus10 dB

Handover decisions are also performed by eNBs based onconfigurable event-triggeredmeasurement reports includingthe RSRQ and the reference signal received power (RSRP)The measurements are performed by UEs usually in orthog-onal frequency division multiplexing (OFDM) symbols car-rying reference signals (RS) for antenna port 0 In ns-3 since the channel is assumed flat over an RB that iscomposed of 12 subcarriers all resource elements (RE) in anRB have the same power Hence assuming orthogonal RSreception the RSRP for cell 119895 measured by vehicle 119894 is givenby

RSRP119895119894 (119905) = sum119872minus1119898=0 sum119870minus1119896=0 (119875119895119894 (119898 119896 119905) 119870)119872= sum119872minus1119898=0 (119875119895119894 (119898 119905) 12)119872

(4)

where 119875119895119894(119898 119896 119905) is the received power of RS 119896 in RB 119898 119872is the number of RBs and 119870 is the number of RS measuredin the RB The average value is passed to higher layers every200ms therefore the index 119905 is omitted for simplicity

The RSRQ for serving cell 119895measured by vehicle 119894 can becomputed by the following [42]

RSRQ119895119894 = 119872 times RSRP119895119894RSSI119895119894

(5)

where RSSI119895119894 is the received signal strength indicator (RSSI)measured by UE 119894when served by cell 119895 and according to the

10 Wireless Communications and Mobile Computing

Table 3 LTE simulation parameters

Parameter ValueSimulation time 100 seconds

Road model Manhattan grid model (MG) (two way roads)Glasgow City Center (GCC) (2 km times 2 km)

LTE network 4 sites with 3 cellssite 1000m ISD for MG6 sites with 3 cellssite UK Operator EE mast data [8]

Transmission power eNB 40 dBm UE 23 dBmCarrier frequency DLUL 2115MHz1715MHzChannel bandwidth 2 times 10MHz (2 times 50RBs)Noise Figure eNB 5 dB UE 9 dBUE antenna model Isotropic (0 dBi)eNB antenna model 15 dB Cosine model 65∘ HPBWScheduling algorithm Proportional Fair

Handover algorithm A2A4RSRQ RSRQ threshold minus5 dBand NeighbourCellOffset = 2 (1 dB)

Frequency reuse Distributed Fractional Freq Reuse

Path loss model LogDistance (120572 = 3) and3GPP Extended Vehicular A model

Safety message format 256 bytes UDPNumber of vehicles 75 100 125 150Average vehiclersquos speed 20 kmh 40 kmhBeacon frequency 1Hz 10HzAwareness range (119877) 100m 250m 500m 750m 1000m

standard comprises the linear average of the total receivedpower observed by the UE from all sources includingcochannel serving and nonserving cells adjacent channelinterference thermal noise and so forth [42] The RSSImeasurement model in ns-3 is computed considering two REper RB (the one that carries the RS) [37] that is

RSSI119895119894 (119905)= 119872minus1sum119898=0

2 times (119873012 +sumforall119897

119875119897 (119898 119905) 119866119897 (120579119897119894) 119866119894 (120579119894119897)12 times 119871 119897119894 (119898 119905) ) (6)

Furthermore in order to perform handover decisions thereported RSRQ for neighboring cell 119897 when vehicle 119894 is con-nected through serving cell 119895 is computed by the following

RSRQ119897119894 = 119872 times RSRP119897119894RSSI119895119894

(7)

The RSRQs for serving and neighboring cells are usedwith theA2A4RsrqHandoverAlgorithm to trigger LTE eventsA2 and A4 [43 Section 554] Event A2 is set to be triggeredwhen the RSRQ of the serving cell becomes worse than minus5 dB(ServingCellThreshold parameter) While event A2 is trueevent A4 is triggered if a neighboring cell becomes betterthan a threshold which is set by the algorithm to a verylow value for the trigger criteria to be true Thus the mea-surement reports received by the eNB are used to consider aneighboring cell as a target cell for handover only if its RSRQ

is 1 dB higher than the serving cell (NeighbourCellOffsetparameter) Table 3 summarizes the simulation parametersfor LTE

31 Vehicular Radio Urban Environment As mentioned inSection 11 previous evaluations lack the use of multipathand multicell environments Use of proper channel modelingis essential in evaluating system models as it poses close toreality system degradations The system model presented inthis paper consists of 6 sites with 3 cells per site Users areassumed to have isotropic antenna (119866119894(120579119894119895) = 0 dBi) and thepath loss (119871119895119894) ismodeled using the Log-distance propagationwith shadow exponent 119899 = 3 [44] plus 3GPP EVA multipathfading modelL119895119894 [34] and is computed by the following

119871119895119894 (119898 119905) = 1198710 + 10119899 log10 (119889119895119894 (119905)1198890 ) +L119895119894 (119898 119905) dB (8)

where 1198710 = 20 log10(120582(41205871198890)) 120582 is the wavelength and1198890 is the reference distance assumed to be 10m in oursimulations Traces for EVA were generated using MATLABwhile modifying the classical Doppler spectrum for each tapat vehicular speed (V) of 40 kmh and 60 kmh as

119878 (119891) prop 1(1 minus (119891119891119863)2)05 (9)

where 119891119863 = V120582 These EVA traces specified by 3GPP in[34] improvise the NLOS conditions relative to the speed

Wireless Communications and Mobile Computing 11

of UE and the transmission bandwidth Simulation of themultipath fading component in ns-3 uses the trace startingat a random point along the time domain for a given user andserving cell The channel object created with the rayleighchanfunction is used for filtering a discrete-time impulse signal inorder to obtain the channel impulse response The filteringis repeated for different Transmission Time Interval (TTI)thus yielding subsequent time-correlated channel responses(one per TTI) This channel response is then processedwith the pwelch function for obtaining its power spectraldensity values which are then saved in a file with the formatcompatible with the simulator model

32 SAI Implementation Model Archer and Vogel in [45]carried out a survey on traffic safety problems in urbanareas It was concluded that overtaking tends to occur lessfrequently within urban areas where the speed limit isless than 50 kmh Lane changing however occurs quitefrequently within urban areas due to low speed limits Thenumber of rear-end accidents is greater within urban areasthan in rural areas and similarly due to higher level ofcongestion and higher number of traffic junctions there isa greater number of opportunities for turning and crossingaccidents within urban area According to another survey of4500 crashes by the Insurance Institute of Highway Safety[46] 13 of accidents are during lane change maneuvers12 are intersection collisions 22 are due to traffic lightviolation 9 are head on collisions 18 include left turningcrashes 12 are due to emergency vehicles parked on blindturns and 14 are due to over speeding In the light ofthese statistics for our adopted urban environment the SAIs(Table 2) we utilize in our simulations are 1 (25) 3 (22) 5(9) 6 (18) 7 (12) and 9 (14)

33 Background Traffic In order to model close to realscenarios we added background traffic into our simulationsModeling of such traffic is done bymaking 50of the existingvehicles use voice application and 25 stream videos Voiceapplication is modeled by implementing constant bit rate(CBR) traffic generator on G711 codec using onoffapplicationon ns-3 [37] Voice payload is 172 bytes and the data ratechosen is 688 Kbps The reason for using a high data rate isto cater worst case scenario where an operator would preferhaving a high quality voice application At the same timevideo streaming is being carried out using evalvid simulatormodule on H263 codec with a payload of 1460 bytes at a datarate of 122Mbps [47] Both these voice and video applicationsare assigned their respective QCIs

34 Performance Measures The primary performance mea-sures used are the end-to-end packet delay the packet deliveryratio and the downlink goodputThe goodput is defined as thenumber of useful information bits received at the applicationlayer per unit of time

In addition we use the notation |F119894| to represent theaverage number of neighbors calculated over the number ofreceivers whenever a beacon is disseminated The end-to-enddelay of a beacon (119863119894119896) is defined as the time measured fromthe start of its transmission at vehicle 119894rsquos application layer until

its successful reception at vehicle 119896rsquos application layer PDR isdefined as the beacon success rate within the awareness range119877 [48] that is for a beacon sent fromvehicle 119894 its PDR is givenby

119875119894 (119877) = sum119896isinF119894

1198681198941198961003816100381610038161003816F1198941003816100381610038161003816 1003816100381610038161003816F1198941003816100381610038161003816 gt 0 (10)

where | sdot | indicates the cardinality of the set and 119868119894119896 (isin 0 1)is set to 1 if the beacon is received by vehicle 1198964 Simulation Results

The success of cellular networks relies on the scalability ofthe system capacity achieved by reusing the spectrum Thecommon approach implemented by operators is to caterthe capacity needs as the traffic demand grows Thereforefirst we analyze the capacity limit which results from thesingle cell case compared to multiple cells using the Friisradio propagation environment and afterwards we analyzethe impact of the urban radio environment as a function ofthe awareness range

Figure 7(a) shows the cumulative distribution function(CDF) of the end-to-end delay when the network is com-posed of 100 vehicles moving at an average speed of 40 kmhThe service area is covered using a single cell with an isotropicantenna located at the upper left corner of Manhattan Gridmodel as utilized by [11] in a Friis radio environment wherevehicles transmit at 1 Hz beacon frequency We observe thatfor awareness range of up to 500m the end-to-end delay isle100ms with around 95 probability Probability of higherend-to-end delay increases more between 250m to 500mas compared to the increase from 500m to 750m That ismainly the result of the increase in traffic but also as theawareness range increases cars located on roads near the edgeof the Manhattan model will have neighbors towards the celldirection with better signal quality which is further exploitedby the scheduling algorithm Figure 7(a) also shows that thereis no significant increase in end-to-end delay from 750mto 1000m This is due to the border effect incurred by thecar located at the edge of the model having less number ofneighbors when increasing the awareness range as comparedto the cars located at the center of the cell Nevertheless theseresults show that up to 100 vehicles were possible for thesingle cell case at a beacon frequency of 1Hz under Friis radioenvironment In order to analyze the intercell interferenceimpact on the delay while increasing the number of vehiclesand beacon frequency we increase the network capacity bydeploying 4 sites with 3 sectorssite (the average number ofvehicles per sector is reduced to 125) Figure 7(b) shows thatthe end-to-enddelay is then further improved to below 50mswith 91 success rate when awareness range of 1000m at thebeacon frequency of 1Hz was utilized We can also noticethat in this case there are some beacons that experiencedlonger delays compared to the single cell scenario causedby low quality conditions of cell-edge users due to intercellinterference and handover However with 1000m 934 ofthe forwarded beacons were received with end-to-end delayle 100ms and 982 for 250m

12 Wireless Communications and Mobile Computing

0 50 100 150 200 250 3000

01

02

03

04

05

06

07

08

09

1CD

F

End-to-end delay (ms)

R = 100

R = 250

R = 500

R = 750

R = 1000

(a)

0

01

02

03

04

05

06

07

08

09

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 1

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 35

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 14

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 25

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 41

(b)

0

01

02

03

04

05

06

07

08

09

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 1

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 35

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 14

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 25

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 41

(c)

Figure 7 LTE end-to-end delay for 100 vehicles at a beacon frequency 1Hz and average speed 40 kmh with Friis radio environment (a)single cell (b) multicell (4 sites 3 sectorssite) (c) multicell with EVA fading radio environment

The results in Figure 7(b) show that under the Friis radiopropagation environment the capacity increases and thesystem performance is not significantly affected in terms ofthe end-to-end delay in presence of intercell interference andhandover since the scheduler effectively managed interfer-ence by time and frequency allocations However in an urbanenvironment it is expected that fading caused by buildingsand Doppler shift due to the relative velocity of vehicles havea very important impact on the received signal quality [49]Figure 7(c) shows the results when the radio propagationenvironment was changed to Log-distance-dependent and

3GPP EVA under the same site configurations Under thisradio environment the probability for the end-to-end delayto be less than or equal to 100ms was around 76ndash78 forawareness range between 100m and 500m 70 for 750mand 66 for 1000m Hence the significant effect on the end-to-end delay is because of path loss and frequency selectivefading imposed by the EVA fading environmentThe end-to-enddelay for beacons to and fromcell-edge users significantlyincreases Similarly for inner cell users the end-to-end delayremained close to the Friis results and for up to 500m theend-to-end delay was le50ms with 74 to 71 probability

Wireless Communications and Mobile Computing 13

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 17

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 53

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 20

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 37

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 62

(a)

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 11

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 69

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 21

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 47

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 73

(b)

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 13

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 63

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 24

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 47

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 74

(c)

Figure 8 End-to-end delay for sample networks at average speed 40 kmh with 150 vehicles at beacon frequency (a) 1Hz with MG model(b) 1Hz with GCC model (c) 10Hz with GCC model

65 for 750m and 59 for 1000m Therefore it can bededuced that with the addition of multipath fading alongwith critical radio propagation environment probability ofdelay being le50ms decreases by 19 to 30

Continuing on with multicell and EVA fading radioenvironment by increasing the network size the number ofvehicles in the awareness range increases and the end-to-enddelay increases as shown in Figure 8(a) Nevertheless theend-to-end delay remained below 100ms for up to 250mwith82 probability for 100 vehicles and similar rate was obtainedwith up to 150 vehicles When the awareness range increasesto 500m the probability was 80 for 100 vehicles while for150 vehicles further reduces to 73 For awareness range of750m and 1000m the probability reduced to 75 and 71 for100 vehicles while for 150 vehicles reduces to 69 and 60respectively

Changing the simulation model along with mobilitytraces to Glasgow city center the effect on delay performancewith 150 vehicles at 1 Hz and 10Hz is shown in Figures 8(b)and 8(c) respectively It is interesting to observe the effect ofchanging the mobility model to a European city Where each

city block is 400m in theMGmodel average city block size inGCC is around 200m increasing vehicle density eventuallygrowing the average number of neighbors for GCC modelThis can be observed for an awareness range of 250m theprobability of delay to be le50ms decreases from 77 to 68while for 1000m this deterioration goes from 54 to 37hence an overall decrease of about 9 to 17 for variousawareness ranges Similarly in Figure 8(c) it is observedthat changing the beacon frequency from 1Hz to 10Hz withan awareness range of 500m the probability of delay to bele50ms drastically decreases from 55 to 22 Whereas foran awareness range of 1000m degradation of about 29 isobserved Performance with a high beacon frequency suchas 10Hz is overall seen to be poor That is because packetsthat arrive late are discarded and are mainly sent by cell-edgeusers for which their neighborhood very likely includes usersat the cell edge Hence the delay from the uplink is indirectlyused as a spatial filtering of low quality cell-edge users Thissuggests that if the information is not useful when receivedafter 100ms the better use of resources is to drop the packetat the minimum latency requirement by the server

14 Wireless Communications and Mobile Computing

0010203040506070809

1

0 20 40 60 80 100 120 140 160 180 200

CDF

End-to-end delay (ms)

SAI = 11003816100381610038161003816Fi

1003816100381610038161003816 = 19

SAI = 31003816100381610038161003816Fi

1003816100381610038161003816 = 16

SAI = 51003816100381610038161003816Fi

1003816100381610038161003816 = 41

SAI = 61003816100381610038161003816Fi

1003816100381610038161003816 = 14

SAI = 71003816100381610038161003816Fi

1003816100381610038161003816 = 56

SAI = 91003816100381610038161003816Fi

1003816100381610038161003816 = 54

Without SAI 1003816100381610038161003816Fi1003816100381610038161003816 = 16

(a)

0010203040506070809

1

0 20 40 60 80 100 120 140 160 180 200

CDF

End-to-end delay (ms)

SAI = 11003816100381610038161003816Fi

1003816100381610038161003816 = 26

SAI = 31003816100381610038161003816Fi

1003816100381610038161003816 = 2

SAI = 51003816100381610038161003816Fi

1003816100381610038161003816 = 35

SAI = 61003816100381610038161003816Fi

1003816100381610038161003816 = 22

SAI = 71003816100381610038161003816Fi

1003816100381610038161003816 = 78

SAI = 91003816100381610038161003816Fi

1003816100381610038161003816 = 94

Without SAI 1003816100381610038161003816Fi1003816100381610038161003816 = 21

(b)

Figure 9 LTE end-to-end delay at average speed 40 kmh (GCC) and (a) 100 vehicles and (b) 150 vehicles

Next we observe and compare the results after theproposed algorithm implementation Figure 9 shows theCDF of the end-to-end delay for 100 and 150 vehicles atan average speed of 40 kmh with and without the imple-mentation of SAI algorithm The scenario being comparedhas the awareness range set to 500m and beacon frequencyto 10Hz We observe a significant decrease in the end-to-end delay after SAI algorithm implementation Probabilityfor the end-to-end delay to be less than 100ms is above80 in both scenarios whereas without the algorithm thisprobability is below 70 for 100 vehicles (Figure 9(a)) andbelow 60 for 150 vehicles (Figure 9(b)) The variation inthis delay between various SAIs is because of the differencein transmission parameters Larger awareness range leadingto higher number of neighbors |F119894| and higher beaconfrequency results in more congestion and large queuingtimes The reason for better delay values with SAI algorithmis mainly the result from restricting resources to the requiredtransmission parameters for safety applicationsmentioned inSection 32

Figure 10 shows the aggregate downlink goodput for thesame scenario investigated for end-to-end delay After theimplementation of SAI algorithm the downlink goodput for100 vehicles dropped from 746Mbps to 622Mbps and for150 vehicles it dropped from 2167Mbps to 1376Mbps Thissignificant decrease in the downlink direction is again due tothe restriction of unnecessary data dissemination from theVSA server to vehicles Eventually this decrease of goodputresults in less load on the LTE system The increase in thedownlink goodput for the scenario without SAI explains thehigh end-to-end delay observed in Figure 9 showing a burstof packet in the downlink leading to waiting time in thequeue

Furthermore the impact of regular cellular traffic on ourLTE vehicular network is studied in order to validate ourfindings and algorithm A broad picture of our observa-tions under different scenarios is shown in Figure 11 whereprobability of end-to-end delay being less than or equal to50ms is studied As expected this probability decreases byabout 3-4 when background traffic is added However with

100 150

Number of vehicles

25

20

15

10

5

0

Dow

nlin

k go

odpu

t (M

bps)

Without SAIWith SAI

746622

2167

1376

Figure 10 Downlink goodput for 100 and 150 vehicles at an averagespeed 40 kmh (GCC)

SAI algorithm probability of end-to-end delay being lessthan 50ms stays above 90 even after adding backgroundtraffic

Finally we include the SAI into MAC layer PDU con-trol elements to prove our concept of D2D like unicasttransmission between sender and receivers Modeling of thesystem is carried out in accordance with the description inSection 22 Figure 12 shows end-to-end delay for 100 and150 vehicles with and without the inclusion of SAI in MAClayer It is evident that 150 vehicles experience almost thesame probability for end-to-end delay being less than 50msas experienced by 100 vehicles This shows that with the useof MAC layer control elements around 50 more vehiclescan be accommodated by LTE network increasing the spec-trum efficiency eventually leading to higher capacity of thesystem

Wireless Communications and Mobile Computing 15

Simple case With backgroundtraffic

With SAI With backgroundand SAI

6065707580859095

100

150 vehicles100 vehicles

Prob

abili

ty o

f end

-to-e

ndde

layle

50

ms

Scenarios

Figure 11 Probability of end-to-end delay being le50ms for 100 and150 vehicles at 40 kmh in various scenarios

With SAI in APP layer With SAI in MAC layer80828486889092949698

100

100 vehicles150 vehicles

p

roba

bilit

y fo

r end

-to-e

ndde

layle50

ms

Figure 12 Probability of end-to-end delay being le50ms for 100 and150 vehicles with and without MAC layer inclusion

5 Conclusion

This paper proposes the safety application identifier conceptin the form of an algorithm that is tested and analyzedwithin the impact of vehicular urbanmulticell radio environ-ment employing multipath fading channels and backgroundtraffic The proposed algorithm uses dynamic adaptation oftransmission parameters in order to fulfill stringent vehicu-lar application requirements while minimizing latency andreducing system load

With the help of extensive system-level simulations thecrucial impact of channel modeling and mobility traces isevident with its influence on the received signal qualitydecreasing the probability of end-to-end delay to be le50msfor various awareness ranges by 19 to 30 and 9 to17 respectively Studying the impact of awareness rangeit is observed that with range up to 1000m for 1Hz 500mfor 2Hz and 250m for 10Hz beacon frequency the systemperformed well With the help of these findings the SAIalgorithm is proposed and implemented Probability of expe-riencing an end-to-end delay of less than 100ms increasedby around 20 and the downlink goodput decreased signif-icantly from 2167Mbps to 1376Mbps for 150 vehicles

Furthermore the impact of having background traffic isalso taken into consideration Results show that the systemis slightly affected by having regular background trafficHowever the probability of end-to-end delay being le50ms

still stays satisfactorily above 85 with the use of SAI Thispaper also proposes the inclusion of SAI inMAC layer controlelements producing a D2D type of communication withinvehicles After modeling this system it was found that thesame amount of system resources and requirements are metwith additional 50 vehicles increasing the capacity of ourLTE vehicular network Finally in future works we plan tointegrate SAI incorporated within MAC layer in a hetero-geneous network with DSRC while exploring the networkslicing in 5G cellular networks for vehicular communicationscomplementing each otherrsquos weaknesses and strengths inorder to meet the vehicular network requirements

Conflicts of Interest

The authors declare that they have no conflicts of interest

Acknowledgments

Marvin Sanchez would like to acknowledge the support ofLAMENITEC Erasmus Mundus Program of the EuropeanUnion to be a guest researcher at GCU Results were obtainedusing the EPSRC funded ARCHIE-WeSt High PerformanceComputer (httpswwwarchie-westacuk) EPSRC Grantno EPK0005861

References

[1] S Sesia I Toufik and M Baker LTE the UMTS long termevolution from theory to practice Wiley Publishing 2009

[2] H Hartenstein and K P Laberteaux ldquoA tutorial survey onvehicular ad hoc networksrdquo IEEE Communications Magazinevol 46 no 6 pp 164ndash171 2008

[3] D Jiang and L Delgrossi ldquoTowards an international standardfor wireless access in vehicular environmentsrdquo in Proceedings ofthe 67th IEEE Vehicular Technology Conference (VTC rsquo08) pp2036ndash2040 May 2008

[4] S Al-Sultan M M Al-Doori A H Al-Bayatti and H ZedanldquoA comprehensive survey on vehicular Ad Hoc networkrdquoJournal of Network and Computer Applications vol 37 no 1 pp380ndash392 2014

[5] K Tanuja T Sushma M Bharathi and K Arun ldquoA surveyon VANET technologiesrdquo International Journal of ComputerApplications vol 121 no 18 2015

[6] H Y Kim D M Kang J H Lee and T M Chung ldquoA per-formance evaluation of cellular network suitability for VANETrdquoWorld Academy of Science Engineering and Technology Interna-tional Science Index vol 64 no 6 pp 124ndash127 2012

[7] A Fasbender M Gerdes and S Smets ldquoCellular NetworkingTechnologies in ITS Solutions Opportunities and Challengesrdquopp 1ndash13 Springer Berlin Germany 2012

[8] ldquoCellular coverage and tower map for ee network in glasgowrdquohttpswwwcellmappernetmap

[9] GAraniti C CampoloMCondoluci A Iera andAMolinaroldquoLTE for vehicular networking a surveyrdquo IEEECommunicationsMagazine vol 51 no 5 pp 148ndash157 2013

[10] A Vinel ldquo3GPP LTE versus IEEE 80211pWAVE which tech-nology is able to support cooperative vehicular safety applica-tionsrdquo IEEE Wireless Communications Letters vol 1 no 2 pp125ndash128 2012

Wireless Communications and Mobile Computing 9

S-GWP-GWVSA

TCC

Figure 6 GCC model covered by 6 sites and 3 cellssite where eNB location is determined using UK operator EE mast data [8]

interference coordination (ICIC) [41] The eNB adaptivelyselects a set of RBs referred to as cell-edge subband (set to10RBs) based on information exchanged with its neighborsTherefore according to ICIC a vehicle should be allocatedin the cell-edge subband if its serving cell reference signalreceived quality (RSRQ) gets lower than minus10 dB

Handover decisions are also performed by eNBs based onconfigurable event-triggeredmeasurement reports includingthe RSRQ and the reference signal received power (RSRP)The measurements are performed by UEs usually in orthog-onal frequency division multiplexing (OFDM) symbols car-rying reference signals (RS) for antenna port 0 In ns-3 since the channel is assumed flat over an RB that iscomposed of 12 subcarriers all resource elements (RE) in anRB have the same power Hence assuming orthogonal RSreception the RSRP for cell 119895 measured by vehicle 119894 is givenby

RSRP119895119894 (119905) = sum119872minus1119898=0 sum119870minus1119896=0 (119875119895119894 (119898 119896 119905) 119870)119872= sum119872minus1119898=0 (119875119895119894 (119898 119905) 12)119872

(4)

where 119875119895119894(119898 119896 119905) is the received power of RS 119896 in RB 119898 119872is the number of RBs and 119870 is the number of RS measuredin the RB The average value is passed to higher layers every200ms therefore the index 119905 is omitted for simplicity

The RSRQ for serving cell 119895measured by vehicle 119894 can becomputed by the following [42]

RSRQ119895119894 = 119872 times RSRP119895119894RSSI119895119894

(5)

where RSSI119895119894 is the received signal strength indicator (RSSI)measured by UE 119894when served by cell 119895 and according to the

10 Wireless Communications and Mobile Computing

Table 3 LTE simulation parameters

Parameter ValueSimulation time 100 seconds

Road model Manhattan grid model (MG) (two way roads)Glasgow City Center (GCC) (2 km times 2 km)

LTE network 4 sites with 3 cellssite 1000m ISD for MG6 sites with 3 cellssite UK Operator EE mast data [8]

Transmission power eNB 40 dBm UE 23 dBmCarrier frequency DLUL 2115MHz1715MHzChannel bandwidth 2 times 10MHz (2 times 50RBs)Noise Figure eNB 5 dB UE 9 dBUE antenna model Isotropic (0 dBi)eNB antenna model 15 dB Cosine model 65∘ HPBWScheduling algorithm Proportional Fair

Handover algorithm A2A4RSRQ RSRQ threshold minus5 dBand NeighbourCellOffset = 2 (1 dB)

Frequency reuse Distributed Fractional Freq Reuse

Path loss model LogDistance (120572 = 3) and3GPP Extended Vehicular A model

Safety message format 256 bytes UDPNumber of vehicles 75 100 125 150Average vehiclersquos speed 20 kmh 40 kmhBeacon frequency 1Hz 10HzAwareness range (119877) 100m 250m 500m 750m 1000m

standard comprises the linear average of the total receivedpower observed by the UE from all sources includingcochannel serving and nonserving cells adjacent channelinterference thermal noise and so forth [42] The RSSImeasurement model in ns-3 is computed considering two REper RB (the one that carries the RS) [37] that is

RSSI119895119894 (119905)= 119872minus1sum119898=0

2 times (119873012 +sumforall119897

119875119897 (119898 119905) 119866119897 (120579119897119894) 119866119894 (120579119894119897)12 times 119871 119897119894 (119898 119905) ) (6)

Furthermore in order to perform handover decisions thereported RSRQ for neighboring cell 119897 when vehicle 119894 is con-nected through serving cell 119895 is computed by the following

RSRQ119897119894 = 119872 times RSRP119897119894RSSI119895119894

(7)

The RSRQs for serving and neighboring cells are usedwith theA2A4RsrqHandoverAlgorithm to trigger LTE eventsA2 and A4 [43 Section 554] Event A2 is set to be triggeredwhen the RSRQ of the serving cell becomes worse than minus5 dB(ServingCellThreshold parameter) While event A2 is trueevent A4 is triggered if a neighboring cell becomes betterthan a threshold which is set by the algorithm to a verylow value for the trigger criteria to be true Thus the mea-surement reports received by the eNB are used to consider aneighboring cell as a target cell for handover only if its RSRQ

is 1 dB higher than the serving cell (NeighbourCellOffsetparameter) Table 3 summarizes the simulation parametersfor LTE

31 Vehicular Radio Urban Environment As mentioned inSection 11 previous evaluations lack the use of multipathand multicell environments Use of proper channel modelingis essential in evaluating system models as it poses close toreality system degradations The system model presented inthis paper consists of 6 sites with 3 cells per site Users areassumed to have isotropic antenna (119866119894(120579119894119895) = 0 dBi) and thepath loss (119871119895119894) ismodeled using the Log-distance propagationwith shadow exponent 119899 = 3 [44] plus 3GPP EVA multipathfading modelL119895119894 [34] and is computed by the following

119871119895119894 (119898 119905) = 1198710 + 10119899 log10 (119889119895119894 (119905)1198890 ) +L119895119894 (119898 119905) dB (8)

where 1198710 = 20 log10(120582(41205871198890)) 120582 is the wavelength and1198890 is the reference distance assumed to be 10m in oursimulations Traces for EVA were generated using MATLABwhile modifying the classical Doppler spectrum for each tapat vehicular speed (V) of 40 kmh and 60 kmh as

119878 (119891) prop 1(1 minus (119891119891119863)2)05 (9)

where 119891119863 = V120582 These EVA traces specified by 3GPP in[34] improvise the NLOS conditions relative to the speed

Wireless Communications and Mobile Computing 11

of UE and the transmission bandwidth Simulation of themultipath fading component in ns-3 uses the trace startingat a random point along the time domain for a given user andserving cell The channel object created with the rayleighchanfunction is used for filtering a discrete-time impulse signal inorder to obtain the channel impulse response The filteringis repeated for different Transmission Time Interval (TTI)thus yielding subsequent time-correlated channel responses(one per TTI) This channel response is then processedwith the pwelch function for obtaining its power spectraldensity values which are then saved in a file with the formatcompatible with the simulator model

32 SAI Implementation Model Archer and Vogel in [45]carried out a survey on traffic safety problems in urbanareas It was concluded that overtaking tends to occur lessfrequently within urban areas where the speed limit isless than 50 kmh Lane changing however occurs quitefrequently within urban areas due to low speed limits Thenumber of rear-end accidents is greater within urban areasthan in rural areas and similarly due to higher level ofcongestion and higher number of traffic junctions there isa greater number of opportunities for turning and crossingaccidents within urban area According to another survey of4500 crashes by the Insurance Institute of Highway Safety[46] 13 of accidents are during lane change maneuvers12 are intersection collisions 22 are due to traffic lightviolation 9 are head on collisions 18 include left turningcrashes 12 are due to emergency vehicles parked on blindturns and 14 are due to over speeding In the light ofthese statistics for our adopted urban environment the SAIs(Table 2) we utilize in our simulations are 1 (25) 3 (22) 5(9) 6 (18) 7 (12) and 9 (14)

33 Background Traffic In order to model close to realscenarios we added background traffic into our simulationsModeling of such traffic is done bymaking 50of the existingvehicles use voice application and 25 stream videos Voiceapplication is modeled by implementing constant bit rate(CBR) traffic generator on G711 codec using onoffapplicationon ns-3 [37] Voice payload is 172 bytes and the data ratechosen is 688 Kbps The reason for using a high data rate isto cater worst case scenario where an operator would preferhaving a high quality voice application At the same timevideo streaming is being carried out using evalvid simulatormodule on H263 codec with a payload of 1460 bytes at a datarate of 122Mbps [47] Both these voice and video applicationsare assigned their respective QCIs

34 Performance Measures The primary performance mea-sures used are the end-to-end packet delay the packet deliveryratio and the downlink goodputThe goodput is defined as thenumber of useful information bits received at the applicationlayer per unit of time

In addition we use the notation |F119894| to represent theaverage number of neighbors calculated over the number ofreceivers whenever a beacon is disseminated The end-to-enddelay of a beacon (119863119894119896) is defined as the time measured fromthe start of its transmission at vehicle 119894rsquos application layer until

its successful reception at vehicle 119896rsquos application layer PDR isdefined as the beacon success rate within the awareness range119877 [48] that is for a beacon sent fromvehicle 119894 its PDR is givenby

119875119894 (119877) = sum119896isinF119894

1198681198941198961003816100381610038161003816F1198941003816100381610038161003816 1003816100381610038161003816F1198941003816100381610038161003816 gt 0 (10)

where | sdot | indicates the cardinality of the set and 119868119894119896 (isin 0 1)is set to 1 if the beacon is received by vehicle 1198964 Simulation Results

The success of cellular networks relies on the scalability ofthe system capacity achieved by reusing the spectrum Thecommon approach implemented by operators is to caterthe capacity needs as the traffic demand grows Thereforefirst we analyze the capacity limit which results from thesingle cell case compared to multiple cells using the Friisradio propagation environment and afterwards we analyzethe impact of the urban radio environment as a function ofthe awareness range

Figure 7(a) shows the cumulative distribution function(CDF) of the end-to-end delay when the network is com-posed of 100 vehicles moving at an average speed of 40 kmhThe service area is covered using a single cell with an isotropicantenna located at the upper left corner of Manhattan Gridmodel as utilized by [11] in a Friis radio environment wherevehicles transmit at 1 Hz beacon frequency We observe thatfor awareness range of up to 500m the end-to-end delay isle100ms with around 95 probability Probability of higherend-to-end delay increases more between 250m to 500mas compared to the increase from 500m to 750m That ismainly the result of the increase in traffic but also as theawareness range increases cars located on roads near the edgeof the Manhattan model will have neighbors towards the celldirection with better signal quality which is further exploitedby the scheduling algorithm Figure 7(a) also shows that thereis no significant increase in end-to-end delay from 750mto 1000m This is due to the border effect incurred by thecar located at the edge of the model having less number ofneighbors when increasing the awareness range as comparedto the cars located at the center of the cell Nevertheless theseresults show that up to 100 vehicles were possible for thesingle cell case at a beacon frequency of 1Hz under Friis radioenvironment In order to analyze the intercell interferenceimpact on the delay while increasing the number of vehiclesand beacon frequency we increase the network capacity bydeploying 4 sites with 3 sectorssite (the average number ofvehicles per sector is reduced to 125) Figure 7(b) shows thatthe end-to-enddelay is then further improved to below 50mswith 91 success rate when awareness range of 1000m at thebeacon frequency of 1Hz was utilized We can also noticethat in this case there are some beacons that experiencedlonger delays compared to the single cell scenario causedby low quality conditions of cell-edge users due to intercellinterference and handover However with 1000m 934 ofthe forwarded beacons were received with end-to-end delayle 100ms and 982 for 250m

12 Wireless Communications and Mobile Computing

0 50 100 150 200 250 3000

01

02

03

04

05

06

07

08

09

1CD

F

End-to-end delay (ms)

R = 100

R = 250

R = 500

R = 750

R = 1000

(a)

0

01

02

03

04

05

06

07

08

09

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 1

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 35

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 14

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 25

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 41

(b)

0

01

02

03

04

05

06

07

08

09

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 1

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 35

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 14

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 25

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 41

(c)

Figure 7 LTE end-to-end delay for 100 vehicles at a beacon frequency 1Hz and average speed 40 kmh with Friis radio environment (a)single cell (b) multicell (4 sites 3 sectorssite) (c) multicell with EVA fading radio environment

The results in Figure 7(b) show that under the Friis radiopropagation environment the capacity increases and thesystem performance is not significantly affected in terms ofthe end-to-end delay in presence of intercell interference andhandover since the scheduler effectively managed interfer-ence by time and frequency allocations However in an urbanenvironment it is expected that fading caused by buildingsand Doppler shift due to the relative velocity of vehicles havea very important impact on the received signal quality [49]Figure 7(c) shows the results when the radio propagationenvironment was changed to Log-distance-dependent and

3GPP EVA under the same site configurations Under thisradio environment the probability for the end-to-end delayto be less than or equal to 100ms was around 76ndash78 forawareness range between 100m and 500m 70 for 750mand 66 for 1000m Hence the significant effect on the end-to-end delay is because of path loss and frequency selectivefading imposed by the EVA fading environmentThe end-to-enddelay for beacons to and fromcell-edge users significantlyincreases Similarly for inner cell users the end-to-end delayremained close to the Friis results and for up to 500m theend-to-end delay was le50ms with 74 to 71 probability

Wireless Communications and Mobile Computing 13

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 17

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 53

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 20

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 37

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 62

(a)

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 11

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 69

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 21

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 47

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 73

(b)

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 13

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 63

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 24

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 47

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 74

(c)

Figure 8 End-to-end delay for sample networks at average speed 40 kmh with 150 vehicles at beacon frequency (a) 1Hz with MG model(b) 1Hz with GCC model (c) 10Hz with GCC model

65 for 750m and 59 for 1000m Therefore it can bededuced that with the addition of multipath fading alongwith critical radio propagation environment probability ofdelay being le50ms decreases by 19 to 30

Continuing on with multicell and EVA fading radioenvironment by increasing the network size the number ofvehicles in the awareness range increases and the end-to-enddelay increases as shown in Figure 8(a) Nevertheless theend-to-end delay remained below 100ms for up to 250mwith82 probability for 100 vehicles and similar rate was obtainedwith up to 150 vehicles When the awareness range increasesto 500m the probability was 80 for 100 vehicles while for150 vehicles further reduces to 73 For awareness range of750m and 1000m the probability reduced to 75 and 71 for100 vehicles while for 150 vehicles reduces to 69 and 60respectively

Changing the simulation model along with mobilitytraces to Glasgow city center the effect on delay performancewith 150 vehicles at 1 Hz and 10Hz is shown in Figures 8(b)and 8(c) respectively It is interesting to observe the effect ofchanging the mobility model to a European city Where each

city block is 400m in theMGmodel average city block size inGCC is around 200m increasing vehicle density eventuallygrowing the average number of neighbors for GCC modelThis can be observed for an awareness range of 250m theprobability of delay to be le50ms decreases from 77 to 68while for 1000m this deterioration goes from 54 to 37hence an overall decrease of about 9 to 17 for variousawareness ranges Similarly in Figure 8(c) it is observedthat changing the beacon frequency from 1Hz to 10Hz withan awareness range of 500m the probability of delay to bele50ms drastically decreases from 55 to 22 Whereas foran awareness range of 1000m degradation of about 29 isobserved Performance with a high beacon frequency suchas 10Hz is overall seen to be poor That is because packetsthat arrive late are discarded and are mainly sent by cell-edgeusers for which their neighborhood very likely includes usersat the cell edge Hence the delay from the uplink is indirectlyused as a spatial filtering of low quality cell-edge users Thissuggests that if the information is not useful when receivedafter 100ms the better use of resources is to drop the packetat the minimum latency requirement by the server

14 Wireless Communications and Mobile Computing

0010203040506070809

1

0 20 40 60 80 100 120 140 160 180 200

CDF

End-to-end delay (ms)

SAI = 11003816100381610038161003816Fi

1003816100381610038161003816 = 19

SAI = 31003816100381610038161003816Fi

1003816100381610038161003816 = 16

SAI = 51003816100381610038161003816Fi

1003816100381610038161003816 = 41

SAI = 61003816100381610038161003816Fi

1003816100381610038161003816 = 14

SAI = 71003816100381610038161003816Fi

1003816100381610038161003816 = 56

SAI = 91003816100381610038161003816Fi

1003816100381610038161003816 = 54

Without SAI 1003816100381610038161003816Fi1003816100381610038161003816 = 16

(a)

0010203040506070809

1

0 20 40 60 80 100 120 140 160 180 200

CDF

End-to-end delay (ms)

SAI = 11003816100381610038161003816Fi

1003816100381610038161003816 = 26

SAI = 31003816100381610038161003816Fi

1003816100381610038161003816 = 2

SAI = 51003816100381610038161003816Fi

1003816100381610038161003816 = 35

SAI = 61003816100381610038161003816Fi

1003816100381610038161003816 = 22

SAI = 71003816100381610038161003816Fi

1003816100381610038161003816 = 78

SAI = 91003816100381610038161003816Fi

1003816100381610038161003816 = 94

Without SAI 1003816100381610038161003816Fi1003816100381610038161003816 = 21

(b)

Figure 9 LTE end-to-end delay at average speed 40 kmh (GCC) and (a) 100 vehicles and (b) 150 vehicles

Next we observe and compare the results after theproposed algorithm implementation Figure 9 shows theCDF of the end-to-end delay for 100 and 150 vehicles atan average speed of 40 kmh with and without the imple-mentation of SAI algorithm The scenario being comparedhas the awareness range set to 500m and beacon frequencyto 10Hz We observe a significant decrease in the end-to-end delay after SAI algorithm implementation Probabilityfor the end-to-end delay to be less than 100ms is above80 in both scenarios whereas without the algorithm thisprobability is below 70 for 100 vehicles (Figure 9(a)) andbelow 60 for 150 vehicles (Figure 9(b)) The variation inthis delay between various SAIs is because of the differencein transmission parameters Larger awareness range leadingto higher number of neighbors |F119894| and higher beaconfrequency results in more congestion and large queuingtimes The reason for better delay values with SAI algorithmis mainly the result from restricting resources to the requiredtransmission parameters for safety applicationsmentioned inSection 32

Figure 10 shows the aggregate downlink goodput for thesame scenario investigated for end-to-end delay After theimplementation of SAI algorithm the downlink goodput for100 vehicles dropped from 746Mbps to 622Mbps and for150 vehicles it dropped from 2167Mbps to 1376Mbps Thissignificant decrease in the downlink direction is again due tothe restriction of unnecessary data dissemination from theVSA server to vehicles Eventually this decrease of goodputresults in less load on the LTE system The increase in thedownlink goodput for the scenario without SAI explains thehigh end-to-end delay observed in Figure 9 showing a burstof packet in the downlink leading to waiting time in thequeue

Furthermore the impact of regular cellular traffic on ourLTE vehicular network is studied in order to validate ourfindings and algorithm A broad picture of our observa-tions under different scenarios is shown in Figure 11 whereprobability of end-to-end delay being less than or equal to50ms is studied As expected this probability decreases byabout 3-4 when background traffic is added However with

100 150

Number of vehicles

25

20

15

10

5

0

Dow

nlin

k go

odpu

t (M

bps)

Without SAIWith SAI

746622

2167

1376

Figure 10 Downlink goodput for 100 and 150 vehicles at an averagespeed 40 kmh (GCC)

SAI algorithm probability of end-to-end delay being lessthan 50ms stays above 90 even after adding backgroundtraffic

Finally we include the SAI into MAC layer PDU con-trol elements to prove our concept of D2D like unicasttransmission between sender and receivers Modeling of thesystem is carried out in accordance with the description inSection 22 Figure 12 shows end-to-end delay for 100 and150 vehicles with and without the inclusion of SAI in MAClayer It is evident that 150 vehicles experience almost thesame probability for end-to-end delay being less than 50msas experienced by 100 vehicles This shows that with the useof MAC layer control elements around 50 more vehiclescan be accommodated by LTE network increasing the spec-trum efficiency eventually leading to higher capacity of thesystem

Wireless Communications and Mobile Computing 15

Simple case With backgroundtraffic

With SAI With backgroundand SAI

6065707580859095

100

150 vehicles100 vehicles

Prob

abili

ty o

f end

-to-e

ndde

layle

50

ms

Scenarios

Figure 11 Probability of end-to-end delay being le50ms for 100 and150 vehicles at 40 kmh in various scenarios

With SAI in APP layer With SAI in MAC layer80828486889092949698

100

100 vehicles150 vehicles

p

roba

bilit

y fo

r end

-to-e

ndde

layle50

ms

Figure 12 Probability of end-to-end delay being le50ms for 100 and150 vehicles with and without MAC layer inclusion

5 Conclusion

This paper proposes the safety application identifier conceptin the form of an algorithm that is tested and analyzedwithin the impact of vehicular urbanmulticell radio environ-ment employing multipath fading channels and backgroundtraffic The proposed algorithm uses dynamic adaptation oftransmission parameters in order to fulfill stringent vehicu-lar application requirements while minimizing latency andreducing system load

With the help of extensive system-level simulations thecrucial impact of channel modeling and mobility traces isevident with its influence on the received signal qualitydecreasing the probability of end-to-end delay to be le50msfor various awareness ranges by 19 to 30 and 9 to17 respectively Studying the impact of awareness rangeit is observed that with range up to 1000m for 1Hz 500mfor 2Hz and 250m for 10Hz beacon frequency the systemperformed well With the help of these findings the SAIalgorithm is proposed and implemented Probability of expe-riencing an end-to-end delay of less than 100ms increasedby around 20 and the downlink goodput decreased signif-icantly from 2167Mbps to 1376Mbps for 150 vehicles

Furthermore the impact of having background traffic isalso taken into consideration Results show that the systemis slightly affected by having regular background trafficHowever the probability of end-to-end delay being le50ms

still stays satisfactorily above 85 with the use of SAI Thispaper also proposes the inclusion of SAI inMAC layer controlelements producing a D2D type of communication withinvehicles After modeling this system it was found that thesame amount of system resources and requirements are metwith additional 50 vehicles increasing the capacity of ourLTE vehicular network Finally in future works we plan tointegrate SAI incorporated within MAC layer in a hetero-geneous network with DSRC while exploring the networkslicing in 5G cellular networks for vehicular communicationscomplementing each otherrsquos weaknesses and strengths inorder to meet the vehicular network requirements

Conflicts of Interest

The authors declare that they have no conflicts of interest

Acknowledgments

Marvin Sanchez would like to acknowledge the support ofLAMENITEC Erasmus Mundus Program of the EuropeanUnion to be a guest researcher at GCU Results were obtainedusing the EPSRC funded ARCHIE-WeSt High PerformanceComputer (httpswwwarchie-westacuk) EPSRC Grantno EPK0005861

References

[1] S Sesia I Toufik and M Baker LTE the UMTS long termevolution from theory to practice Wiley Publishing 2009

[2] H Hartenstein and K P Laberteaux ldquoA tutorial survey onvehicular ad hoc networksrdquo IEEE Communications Magazinevol 46 no 6 pp 164ndash171 2008

[3] D Jiang and L Delgrossi ldquoTowards an international standardfor wireless access in vehicular environmentsrdquo in Proceedings ofthe 67th IEEE Vehicular Technology Conference (VTC rsquo08) pp2036ndash2040 May 2008

[4] S Al-Sultan M M Al-Doori A H Al-Bayatti and H ZedanldquoA comprehensive survey on vehicular Ad Hoc networkrdquoJournal of Network and Computer Applications vol 37 no 1 pp380ndash392 2014

[5] K Tanuja T Sushma M Bharathi and K Arun ldquoA surveyon VANET technologiesrdquo International Journal of ComputerApplications vol 121 no 18 2015

[6] H Y Kim D M Kang J H Lee and T M Chung ldquoA per-formance evaluation of cellular network suitability for VANETrdquoWorld Academy of Science Engineering and Technology Interna-tional Science Index vol 64 no 6 pp 124ndash127 2012

[7] A Fasbender M Gerdes and S Smets ldquoCellular NetworkingTechnologies in ITS Solutions Opportunities and Challengesrdquopp 1ndash13 Springer Berlin Germany 2012

[8] ldquoCellular coverage and tower map for ee network in glasgowrdquohttpswwwcellmappernetmap

[9] GAraniti C CampoloMCondoluci A Iera andAMolinaroldquoLTE for vehicular networking a surveyrdquo IEEECommunicationsMagazine vol 51 no 5 pp 148ndash157 2013

[10] A Vinel ldquo3GPP LTE versus IEEE 80211pWAVE which tech-nology is able to support cooperative vehicular safety applica-tionsrdquo IEEE Wireless Communications Letters vol 1 no 2 pp125ndash128 2012

10 Wireless Communications and Mobile Computing

Table 3 LTE simulation parameters

Parameter ValueSimulation time 100 seconds

Road model Manhattan grid model (MG) (two way roads)Glasgow City Center (GCC) (2 km times 2 km)

LTE network 4 sites with 3 cellssite 1000m ISD for MG6 sites with 3 cellssite UK Operator EE mast data [8]

Transmission power eNB 40 dBm UE 23 dBmCarrier frequency DLUL 2115MHz1715MHzChannel bandwidth 2 times 10MHz (2 times 50RBs)Noise Figure eNB 5 dB UE 9 dBUE antenna model Isotropic (0 dBi)eNB antenna model 15 dB Cosine model 65∘ HPBWScheduling algorithm Proportional Fair

Handover algorithm A2A4RSRQ RSRQ threshold minus5 dBand NeighbourCellOffset = 2 (1 dB)

Frequency reuse Distributed Fractional Freq Reuse

Path loss model LogDistance (120572 = 3) and3GPP Extended Vehicular A model

Safety message format 256 bytes UDPNumber of vehicles 75 100 125 150Average vehiclersquos speed 20 kmh 40 kmhBeacon frequency 1Hz 10HzAwareness range (119877) 100m 250m 500m 750m 1000m

standard comprises the linear average of the total receivedpower observed by the UE from all sources includingcochannel serving and nonserving cells adjacent channelinterference thermal noise and so forth [42] The RSSImeasurement model in ns-3 is computed considering two REper RB (the one that carries the RS) [37] that is

RSSI119895119894 (119905)= 119872minus1sum119898=0

2 times (119873012 +sumforall119897

119875119897 (119898 119905) 119866119897 (120579119897119894) 119866119894 (120579119894119897)12 times 119871 119897119894 (119898 119905) ) (6)

Furthermore in order to perform handover decisions thereported RSRQ for neighboring cell 119897 when vehicle 119894 is con-nected through serving cell 119895 is computed by the following

RSRQ119897119894 = 119872 times RSRP119897119894RSSI119895119894

(7)

The RSRQs for serving and neighboring cells are usedwith theA2A4RsrqHandoverAlgorithm to trigger LTE eventsA2 and A4 [43 Section 554] Event A2 is set to be triggeredwhen the RSRQ of the serving cell becomes worse than minus5 dB(ServingCellThreshold parameter) While event A2 is trueevent A4 is triggered if a neighboring cell becomes betterthan a threshold which is set by the algorithm to a verylow value for the trigger criteria to be true Thus the mea-surement reports received by the eNB are used to consider aneighboring cell as a target cell for handover only if its RSRQ

is 1 dB higher than the serving cell (NeighbourCellOffsetparameter) Table 3 summarizes the simulation parametersfor LTE

31 Vehicular Radio Urban Environment As mentioned inSection 11 previous evaluations lack the use of multipathand multicell environments Use of proper channel modelingis essential in evaluating system models as it poses close toreality system degradations The system model presented inthis paper consists of 6 sites with 3 cells per site Users areassumed to have isotropic antenna (119866119894(120579119894119895) = 0 dBi) and thepath loss (119871119895119894) ismodeled using the Log-distance propagationwith shadow exponent 119899 = 3 [44] plus 3GPP EVA multipathfading modelL119895119894 [34] and is computed by the following

119871119895119894 (119898 119905) = 1198710 + 10119899 log10 (119889119895119894 (119905)1198890 ) +L119895119894 (119898 119905) dB (8)

where 1198710 = 20 log10(120582(41205871198890)) 120582 is the wavelength and1198890 is the reference distance assumed to be 10m in oursimulations Traces for EVA were generated using MATLABwhile modifying the classical Doppler spectrum for each tapat vehicular speed (V) of 40 kmh and 60 kmh as

119878 (119891) prop 1(1 minus (119891119891119863)2)05 (9)

where 119891119863 = V120582 These EVA traces specified by 3GPP in[34] improvise the NLOS conditions relative to the speed

Wireless Communications and Mobile Computing 11

of UE and the transmission bandwidth Simulation of themultipath fading component in ns-3 uses the trace startingat a random point along the time domain for a given user andserving cell The channel object created with the rayleighchanfunction is used for filtering a discrete-time impulse signal inorder to obtain the channel impulse response The filteringis repeated for different Transmission Time Interval (TTI)thus yielding subsequent time-correlated channel responses(one per TTI) This channel response is then processedwith the pwelch function for obtaining its power spectraldensity values which are then saved in a file with the formatcompatible with the simulator model

32 SAI Implementation Model Archer and Vogel in [45]carried out a survey on traffic safety problems in urbanareas It was concluded that overtaking tends to occur lessfrequently within urban areas where the speed limit isless than 50 kmh Lane changing however occurs quitefrequently within urban areas due to low speed limits Thenumber of rear-end accidents is greater within urban areasthan in rural areas and similarly due to higher level ofcongestion and higher number of traffic junctions there isa greater number of opportunities for turning and crossingaccidents within urban area According to another survey of4500 crashes by the Insurance Institute of Highway Safety[46] 13 of accidents are during lane change maneuvers12 are intersection collisions 22 are due to traffic lightviolation 9 are head on collisions 18 include left turningcrashes 12 are due to emergency vehicles parked on blindturns and 14 are due to over speeding In the light ofthese statistics for our adopted urban environment the SAIs(Table 2) we utilize in our simulations are 1 (25) 3 (22) 5(9) 6 (18) 7 (12) and 9 (14)

33 Background Traffic In order to model close to realscenarios we added background traffic into our simulationsModeling of such traffic is done bymaking 50of the existingvehicles use voice application and 25 stream videos Voiceapplication is modeled by implementing constant bit rate(CBR) traffic generator on G711 codec using onoffapplicationon ns-3 [37] Voice payload is 172 bytes and the data ratechosen is 688 Kbps The reason for using a high data rate isto cater worst case scenario where an operator would preferhaving a high quality voice application At the same timevideo streaming is being carried out using evalvid simulatormodule on H263 codec with a payload of 1460 bytes at a datarate of 122Mbps [47] Both these voice and video applicationsare assigned their respective QCIs

34 Performance Measures The primary performance mea-sures used are the end-to-end packet delay the packet deliveryratio and the downlink goodputThe goodput is defined as thenumber of useful information bits received at the applicationlayer per unit of time

In addition we use the notation |F119894| to represent theaverage number of neighbors calculated over the number ofreceivers whenever a beacon is disseminated The end-to-enddelay of a beacon (119863119894119896) is defined as the time measured fromthe start of its transmission at vehicle 119894rsquos application layer until

its successful reception at vehicle 119896rsquos application layer PDR isdefined as the beacon success rate within the awareness range119877 [48] that is for a beacon sent fromvehicle 119894 its PDR is givenby

119875119894 (119877) = sum119896isinF119894

1198681198941198961003816100381610038161003816F1198941003816100381610038161003816 1003816100381610038161003816F1198941003816100381610038161003816 gt 0 (10)

where | sdot | indicates the cardinality of the set and 119868119894119896 (isin 0 1)is set to 1 if the beacon is received by vehicle 1198964 Simulation Results

The success of cellular networks relies on the scalability ofthe system capacity achieved by reusing the spectrum Thecommon approach implemented by operators is to caterthe capacity needs as the traffic demand grows Thereforefirst we analyze the capacity limit which results from thesingle cell case compared to multiple cells using the Friisradio propagation environment and afterwards we analyzethe impact of the urban radio environment as a function ofthe awareness range

Figure 7(a) shows the cumulative distribution function(CDF) of the end-to-end delay when the network is com-posed of 100 vehicles moving at an average speed of 40 kmhThe service area is covered using a single cell with an isotropicantenna located at the upper left corner of Manhattan Gridmodel as utilized by [11] in a Friis radio environment wherevehicles transmit at 1 Hz beacon frequency We observe thatfor awareness range of up to 500m the end-to-end delay isle100ms with around 95 probability Probability of higherend-to-end delay increases more between 250m to 500mas compared to the increase from 500m to 750m That ismainly the result of the increase in traffic but also as theawareness range increases cars located on roads near the edgeof the Manhattan model will have neighbors towards the celldirection with better signal quality which is further exploitedby the scheduling algorithm Figure 7(a) also shows that thereis no significant increase in end-to-end delay from 750mto 1000m This is due to the border effect incurred by thecar located at the edge of the model having less number ofneighbors when increasing the awareness range as comparedto the cars located at the center of the cell Nevertheless theseresults show that up to 100 vehicles were possible for thesingle cell case at a beacon frequency of 1Hz under Friis radioenvironment In order to analyze the intercell interferenceimpact on the delay while increasing the number of vehiclesand beacon frequency we increase the network capacity bydeploying 4 sites with 3 sectorssite (the average number ofvehicles per sector is reduced to 125) Figure 7(b) shows thatthe end-to-enddelay is then further improved to below 50mswith 91 success rate when awareness range of 1000m at thebeacon frequency of 1Hz was utilized We can also noticethat in this case there are some beacons that experiencedlonger delays compared to the single cell scenario causedby low quality conditions of cell-edge users due to intercellinterference and handover However with 1000m 934 ofthe forwarded beacons were received with end-to-end delayle 100ms and 982 for 250m

12 Wireless Communications and Mobile Computing

0 50 100 150 200 250 3000

01

02

03

04

05

06

07

08

09

1CD

F

End-to-end delay (ms)

R = 100

R = 250

R = 500

R = 750

R = 1000

(a)

0

01

02

03

04

05

06

07

08

09

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 1

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 35

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 14

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 25

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 41

(b)

0

01

02

03

04

05

06

07

08

09

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 1

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 35

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 14

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 25

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 41

(c)

Figure 7 LTE end-to-end delay for 100 vehicles at a beacon frequency 1Hz and average speed 40 kmh with Friis radio environment (a)single cell (b) multicell (4 sites 3 sectorssite) (c) multicell with EVA fading radio environment

The results in Figure 7(b) show that under the Friis radiopropagation environment the capacity increases and thesystem performance is not significantly affected in terms ofthe end-to-end delay in presence of intercell interference andhandover since the scheduler effectively managed interfer-ence by time and frequency allocations However in an urbanenvironment it is expected that fading caused by buildingsand Doppler shift due to the relative velocity of vehicles havea very important impact on the received signal quality [49]Figure 7(c) shows the results when the radio propagationenvironment was changed to Log-distance-dependent and

3GPP EVA under the same site configurations Under thisradio environment the probability for the end-to-end delayto be less than or equal to 100ms was around 76ndash78 forawareness range between 100m and 500m 70 for 750mand 66 for 1000m Hence the significant effect on the end-to-end delay is because of path loss and frequency selectivefading imposed by the EVA fading environmentThe end-to-enddelay for beacons to and fromcell-edge users significantlyincreases Similarly for inner cell users the end-to-end delayremained close to the Friis results and for up to 500m theend-to-end delay was le50ms with 74 to 71 probability

Wireless Communications and Mobile Computing 13

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 17

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 53

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 20

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 37

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 62

(a)

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 11

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 69

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 21

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 47

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 73

(b)

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 13

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 63

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 24

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 47

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 74

(c)

Figure 8 End-to-end delay for sample networks at average speed 40 kmh with 150 vehicles at beacon frequency (a) 1Hz with MG model(b) 1Hz with GCC model (c) 10Hz with GCC model

65 for 750m and 59 for 1000m Therefore it can bededuced that with the addition of multipath fading alongwith critical radio propagation environment probability ofdelay being le50ms decreases by 19 to 30

Continuing on with multicell and EVA fading radioenvironment by increasing the network size the number ofvehicles in the awareness range increases and the end-to-enddelay increases as shown in Figure 8(a) Nevertheless theend-to-end delay remained below 100ms for up to 250mwith82 probability for 100 vehicles and similar rate was obtainedwith up to 150 vehicles When the awareness range increasesto 500m the probability was 80 for 100 vehicles while for150 vehicles further reduces to 73 For awareness range of750m and 1000m the probability reduced to 75 and 71 for100 vehicles while for 150 vehicles reduces to 69 and 60respectively

Changing the simulation model along with mobilitytraces to Glasgow city center the effect on delay performancewith 150 vehicles at 1 Hz and 10Hz is shown in Figures 8(b)and 8(c) respectively It is interesting to observe the effect ofchanging the mobility model to a European city Where each

city block is 400m in theMGmodel average city block size inGCC is around 200m increasing vehicle density eventuallygrowing the average number of neighbors for GCC modelThis can be observed for an awareness range of 250m theprobability of delay to be le50ms decreases from 77 to 68while for 1000m this deterioration goes from 54 to 37hence an overall decrease of about 9 to 17 for variousawareness ranges Similarly in Figure 8(c) it is observedthat changing the beacon frequency from 1Hz to 10Hz withan awareness range of 500m the probability of delay to bele50ms drastically decreases from 55 to 22 Whereas foran awareness range of 1000m degradation of about 29 isobserved Performance with a high beacon frequency suchas 10Hz is overall seen to be poor That is because packetsthat arrive late are discarded and are mainly sent by cell-edgeusers for which their neighborhood very likely includes usersat the cell edge Hence the delay from the uplink is indirectlyused as a spatial filtering of low quality cell-edge users Thissuggests that if the information is not useful when receivedafter 100ms the better use of resources is to drop the packetat the minimum latency requirement by the server

14 Wireless Communications and Mobile Computing

0010203040506070809

1

0 20 40 60 80 100 120 140 160 180 200

CDF

End-to-end delay (ms)

SAI = 11003816100381610038161003816Fi

1003816100381610038161003816 = 19

SAI = 31003816100381610038161003816Fi

1003816100381610038161003816 = 16

SAI = 51003816100381610038161003816Fi

1003816100381610038161003816 = 41

SAI = 61003816100381610038161003816Fi

1003816100381610038161003816 = 14

SAI = 71003816100381610038161003816Fi

1003816100381610038161003816 = 56

SAI = 91003816100381610038161003816Fi

1003816100381610038161003816 = 54

Without SAI 1003816100381610038161003816Fi1003816100381610038161003816 = 16

(a)

0010203040506070809

1

0 20 40 60 80 100 120 140 160 180 200

CDF

End-to-end delay (ms)

SAI = 11003816100381610038161003816Fi

1003816100381610038161003816 = 26

SAI = 31003816100381610038161003816Fi

1003816100381610038161003816 = 2

SAI = 51003816100381610038161003816Fi

1003816100381610038161003816 = 35

SAI = 61003816100381610038161003816Fi

1003816100381610038161003816 = 22

SAI = 71003816100381610038161003816Fi

1003816100381610038161003816 = 78

SAI = 91003816100381610038161003816Fi

1003816100381610038161003816 = 94

Without SAI 1003816100381610038161003816Fi1003816100381610038161003816 = 21

(b)

Figure 9 LTE end-to-end delay at average speed 40 kmh (GCC) and (a) 100 vehicles and (b) 150 vehicles

Next we observe and compare the results after theproposed algorithm implementation Figure 9 shows theCDF of the end-to-end delay for 100 and 150 vehicles atan average speed of 40 kmh with and without the imple-mentation of SAI algorithm The scenario being comparedhas the awareness range set to 500m and beacon frequencyto 10Hz We observe a significant decrease in the end-to-end delay after SAI algorithm implementation Probabilityfor the end-to-end delay to be less than 100ms is above80 in both scenarios whereas without the algorithm thisprobability is below 70 for 100 vehicles (Figure 9(a)) andbelow 60 for 150 vehicles (Figure 9(b)) The variation inthis delay between various SAIs is because of the differencein transmission parameters Larger awareness range leadingto higher number of neighbors |F119894| and higher beaconfrequency results in more congestion and large queuingtimes The reason for better delay values with SAI algorithmis mainly the result from restricting resources to the requiredtransmission parameters for safety applicationsmentioned inSection 32

Figure 10 shows the aggregate downlink goodput for thesame scenario investigated for end-to-end delay After theimplementation of SAI algorithm the downlink goodput for100 vehicles dropped from 746Mbps to 622Mbps and for150 vehicles it dropped from 2167Mbps to 1376Mbps Thissignificant decrease in the downlink direction is again due tothe restriction of unnecessary data dissemination from theVSA server to vehicles Eventually this decrease of goodputresults in less load on the LTE system The increase in thedownlink goodput for the scenario without SAI explains thehigh end-to-end delay observed in Figure 9 showing a burstof packet in the downlink leading to waiting time in thequeue

Furthermore the impact of regular cellular traffic on ourLTE vehicular network is studied in order to validate ourfindings and algorithm A broad picture of our observa-tions under different scenarios is shown in Figure 11 whereprobability of end-to-end delay being less than or equal to50ms is studied As expected this probability decreases byabout 3-4 when background traffic is added However with

100 150

Number of vehicles

25

20

15

10

5

0

Dow

nlin

k go

odpu

t (M

bps)

Without SAIWith SAI

746622

2167

1376

Figure 10 Downlink goodput for 100 and 150 vehicles at an averagespeed 40 kmh (GCC)

SAI algorithm probability of end-to-end delay being lessthan 50ms stays above 90 even after adding backgroundtraffic

Finally we include the SAI into MAC layer PDU con-trol elements to prove our concept of D2D like unicasttransmission between sender and receivers Modeling of thesystem is carried out in accordance with the description inSection 22 Figure 12 shows end-to-end delay for 100 and150 vehicles with and without the inclusion of SAI in MAClayer It is evident that 150 vehicles experience almost thesame probability for end-to-end delay being less than 50msas experienced by 100 vehicles This shows that with the useof MAC layer control elements around 50 more vehiclescan be accommodated by LTE network increasing the spec-trum efficiency eventually leading to higher capacity of thesystem

Wireless Communications and Mobile Computing 15

Simple case With backgroundtraffic

With SAI With backgroundand SAI

6065707580859095

100

150 vehicles100 vehicles

Prob

abili

ty o

f end

-to-e

ndde

layle

50

ms

Scenarios

Figure 11 Probability of end-to-end delay being le50ms for 100 and150 vehicles at 40 kmh in various scenarios

With SAI in APP layer With SAI in MAC layer80828486889092949698

100

100 vehicles150 vehicles

p

roba

bilit

y fo

r end

-to-e

ndde

layle50

ms

Figure 12 Probability of end-to-end delay being le50ms for 100 and150 vehicles with and without MAC layer inclusion

5 Conclusion

This paper proposes the safety application identifier conceptin the form of an algorithm that is tested and analyzedwithin the impact of vehicular urbanmulticell radio environ-ment employing multipath fading channels and backgroundtraffic The proposed algorithm uses dynamic adaptation oftransmission parameters in order to fulfill stringent vehicu-lar application requirements while minimizing latency andreducing system load

With the help of extensive system-level simulations thecrucial impact of channel modeling and mobility traces isevident with its influence on the received signal qualitydecreasing the probability of end-to-end delay to be le50msfor various awareness ranges by 19 to 30 and 9 to17 respectively Studying the impact of awareness rangeit is observed that with range up to 1000m for 1Hz 500mfor 2Hz and 250m for 10Hz beacon frequency the systemperformed well With the help of these findings the SAIalgorithm is proposed and implemented Probability of expe-riencing an end-to-end delay of less than 100ms increasedby around 20 and the downlink goodput decreased signif-icantly from 2167Mbps to 1376Mbps for 150 vehicles

Furthermore the impact of having background traffic isalso taken into consideration Results show that the systemis slightly affected by having regular background trafficHowever the probability of end-to-end delay being le50ms

still stays satisfactorily above 85 with the use of SAI Thispaper also proposes the inclusion of SAI inMAC layer controlelements producing a D2D type of communication withinvehicles After modeling this system it was found that thesame amount of system resources and requirements are metwith additional 50 vehicles increasing the capacity of ourLTE vehicular network Finally in future works we plan tointegrate SAI incorporated within MAC layer in a hetero-geneous network with DSRC while exploring the networkslicing in 5G cellular networks for vehicular communicationscomplementing each otherrsquos weaknesses and strengths inorder to meet the vehicular network requirements

Conflicts of Interest

The authors declare that they have no conflicts of interest

Acknowledgments

Marvin Sanchez would like to acknowledge the support ofLAMENITEC Erasmus Mundus Program of the EuropeanUnion to be a guest researcher at GCU Results were obtainedusing the EPSRC funded ARCHIE-WeSt High PerformanceComputer (httpswwwarchie-westacuk) EPSRC Grantno EPK0005861

References

[1] S Sesia I Toufik and M Baker LTE the UMTS long termevolution from theory to practice Wiley Publishing 2009

[2] H Hartenstein and K P Laberteaux ldquoA tutorial survey onvehicular ad hoc networksrdquo IEEE Communications Magazinevol 46 no 6 pp 164ndash171 2008

[3] D Jiang and L Delgrossi ldquoTowards an international standardfor wireless access in vehicular environmentsrdquo in Proceedings ofthe 67th IEEE Vehicular Technology Conference (VTC rsquo08) pp2036ndash2040 May 2008

[4] S Al-Sultan M M Al-Doori A H Al-Bayatti and H ZedanldquoA comprehensive survey on vehicular Ad Hoc networkrdquoJournal of Network and Computer Applications vol 37 no 1 pp380ndash392 2014

[5] K Tanuja T Sushma M Bharathi and K Arun ldquoA surveyon VANET technologiesrdquo International Journal of ComputerApplications vol 121 no 18 2015

[6] H Y Kim D M Kang J H Lee and T M Chung ldquoA per-formance evaluation of cellular network suitability for VANETrdquoWorld Academy of Science Engineering and Technology Interna-tional Science Index vol 64 no 6 pp 124ndash127 2012

[7] A Fasbender M Gerdes and S Smets ldquoCellular NetworkingTechnologies in ITS Solutions Opportunities and Challengesrdquopp 1ndash13 Springer Berlin Germany 2012

[8] ldquoCellular coverage and tower map for ee network in glasgowrdquohttpswwwcellmappernetmap

[9] GAraniti C CampoloMCondoluci A Iera andAMolinaroldquoLTE for vehicular networking a surveyrdquo IEEECommunicationsMagazine vol 51 no 5 pp 148ndash157 2013

[10] A Vinel ldquo3GPP LTE versus IEEE 80211pWAVE which tech-nology is able to support cooperative vehicular safety applica-tionsrdquo IEEE Wireless Communications Letters vol 1 no 2 pp125ndash128 2012

Wireless Communications and Mobile Computing 11

of UE and the transmission bandwidth Simulation of themultipath fading component in ns-3 uses the trace startingat a random point along the time domain for a given user andserving cell The channel object created with the rayleighchanfunction is used for filtering a discrete-time impulse signal inorder to obtain the channel impulse response The filteringis repeated for different Transmission Time Interval (TTI)thus yielding subsequent time-correlated channel responses(one per TTI) This channel response is then processedwith the pwelch function for obtaining its power spectraldensity values which are then saved in a file with the formatcompatible with the simulator model

32 SAI Implementation Model Archer and Vogel in [45]carried out a survey on traffic safety problems in urbanareas It was concluded that overtaking tends to occur lessfrequently within urban areas where the speed limit isless than 50 kmh Lane changing however occurs quitefrequently within urban areas due to low speed limits Thenumber of rear-end accidents is greater within urban areasthan in rural areas and similarly due to higher level ofcongestion and higher number of traffic junctions there isa greater number of opportunities for turning and crossingaccidents within urban area According to another survey of4500 crashes by the Insurance Institute of Highway Safety[46] 13 of accidents are during lane change maneuvers12 are intersection collisions 22 are due to traffic lightviolation 9 are head on collisions 18 include left turningcrashes 12 are due to emergency vehicles parked on blindturns and 14 are due to over speeding In the light ofthese statistics for our adopted urban environment the SAIs(Table 2) we utilize in our simulations are 1 (25) 3 (22) 5(9) 6 (18) 7 (12) and 9 (14)

33 Background Traffic In order to model close to realscenarios we added background traffic into our simulationsModeling of such traffic is done bymaking 50of the existingvehicles use voice application and 25 stream videos Voiceapplication is modeled by implementing constant bit rate(CBR) traffic generator on G711 codec using onoffapplicationon ns-3 [37] Voice payload is 172 bytes and the data ratechosen is 688 Kbps The reason for using a high data rate isto cater worst case scenario where an operator would preferhaving a high quality voice application At the same timevideo streaming is being carried out using evalvid simulatormodule on H263 codec with a payload of 1460 bytes at a datarate of 122Mbps [47] Both these voice and video applicationsare assigned their respective QCIs

34 Performance Measures The primary performance mea-sures used are the end-to-end packet delay the packet deliveryratio and the downlink goodputThe goodput is defined as thenumber of useful information bits received at the applicationlayer per unit of time

In addition we use the notation |F119894| to represent theaverage number of neighbors calculated over the number ofreceivers whenever a beacon is disseminated The end-to-enddelay of a beacon (119863119894119896) is defined as the time measured fromthe start of its transmission at vehicle 119894rsquos application layer until

its successful reception at vehicle 119896rsquos application layer PDR isdefined as the beacon success rate within the awareness range119877 [48] that is for a beacon sent fromvehicle 119894 its PDR is givenby

119875119894 (119877) = sum119896isinF119894

1198681198941198961003816100381610038161003816F1198941003816100381610038161003816 1003816100381610038161003816F1198941003816100381610038161003816 gt 0 (10)

where | sdot | indicates the cardinality of the set and 119868119894119896 (isin 0 1)is set to 1 if the beacon is received by vehicle 1198964 Simulation Results

The success of cellular networks relies on the scalability ofthe system capacity achieved by reusing the spectrum Thecommon approach implemented by operators is to caterthe capacity needs as the traffic demand grows Thereforefirst we analyze the capacity limit which results from thesingle cell case compared to multiple cells using the Friisradio propagation environment and afterwards we analyzethe impact of the urban radio environment as a function ofthe awareness range

Figure 7(a) shows the cumulative distribution function(CDF) of the end-to-end delay when the network is com-posed of 100 vehicles moving at an average speed of 40 kmhThe service area is covered using a single cell with an isotropicantenna located at the upper left corner of Manhattan Gridmodel as utilized by [11] in a Friis radio environment wherevehicles transmit at 1 Hz beacon frequency We observe thatfor awareness range of up to 500m the end-to-end delay isle100ms with around 95 probability Probability of higherend-to-end delay increases more between 250m to 500mas compared to the increase from 500m to 750m That ismainly the result of the increase in traffic but also as theawareness range increases cars located on roads near the edgeof the Manhattan model will have neighbors towards the celldirection with better signal quality which is further exploitedby the scheduling algorithm Figure 7(a) also shows that thereis no significant increase in end-to-end delay from 750mto 1000m This is due to the border effect incurred by thecar located at the edge of the model having less number ofneighbors when increasing the awareness range as comparedto the cars located at the center of the cell Nevertheless theseresults show that up to 100 vehicles were possible for thesingle cell case at a beacon frequency of 1Hz under Friis radioenvironment In order to analyze the intercell interferenceimpact on the delay while increasing the number of vehiclesand beacon frequency we increase the network capacity bydeploying 4 sites with 3 sectorssite (the average number ofvehicles per sector is reduced to 125) Figure 7(b) shows thatthe end-to-enddelay is then further improved to below 50mswith 91 success rate when awareness range of 1000m at thebeacon frequency of 1Hz was utilized We can also noticethat in this case there are some beacons that experiencedlonger delays compared to the single cell scenario causedby low quality conditions of cell-edge users due to intercellinterference and handover However with 1000m 934 ofthe forwarded beacons were received with end-to-end delayle 100ms and 982 for 250m

12 Wireless Communications and Mobile Computing

0 50 100 150 200 250 3000

01

02

03

04

05

06

07

08

09

1CD

F

End-to-end delay (ms)

R = 100

R = 250

R = 500

R = 750

R = 1000

(a)

0

01

02

03

04

05

06

07

08

09

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 1

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 35

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 14

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 25

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 41

(b)

0

01

02

03

04

05

06

07

08

09

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 1

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 35

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 14

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 25

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 41

(c)

Figure 7 LTE end-to-end delay for 100 vehicles at a beacon frequency 1Hz and average speed 40 kmh with Friis radio environment (a)single cell (b) multicell (4 sites 3 sectorssite) (c) multicell with EVA fading radio environment

The results in Figure 7(b) show that under the Friis radiopropagation environment the capacity increases and thesystem performance is not significantly affected in terms ofthe end-to-end delay in presence of intercell interference andhandover since the scheduler effectively managed interfer-ence by time and frequency allocations However in an urbanenvironment it is expected that fading caused by buildingsand Doppler shift due to the relative velocity of vehicles havea very important impact on the received signal quality [49]Figure 7(c) shows the results when the radio propagationenvironment was changed to Log-distance-dependent and

3GPP EVA under the same site configurations Under thisradio environment the probability for the end-to-end delayto be less than or equal to 100ms was around 76ndash78 forawareness range between 100m and 500m 70 for 750mand 66 for 1000m Hence the significant effect on the end-to-end delay is because of path loss and frequency selectivefading imposed by the EVA fading environmentThe end-to-enddelay for beacons to and fromcell-edge users significantlyincreases Similarly for inner cell users the end-to-end delayremained close to the Friis results and for up to 500m theend-to-end delay was le50ms with 74 to 71 probability

Wireless Communications and Mobile Computing 13

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 17

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 53

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 20

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 37

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 62

(a)

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 11

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 69

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 21

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 47

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 73

(b)

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 13

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 63

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 24

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 47

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 74

(c)

Figure 8 End-to-end delay for sample networks at average speed 40 kmh with 150 vehicles at beacon frequency (a) 1Hz with MG model(b) 1Hz with GCC model (c) 10Hz with GCC model

65 for 750m and 59 for 1000m Therefore it can bededuced that with the addition of multipath fading alongwith critical radio propagation environment probability ofdelay being le50ms decreases by 19 to 30

Continuing on with multicell and EVA fading radioenvironment by increasing the network size the number ofvehicles in the awareness range increases and the end-to-enddelay increases as shown in Figure 8(a) Nevertheless theend-to-end delay remained below 100ms for up to 250mwith82 probability for 100 vehicles and similar rate was obtainedwith up to 150 vehicles When the awareness range increasesto 500m the probability was 80 for 100 vehicles while for150 vehicles further reduces to 73 For awareness range of750m and 1000m the probability reduced to 75 and 71 for100 vehicles while for 150 vehicles reduces to 69 and 60respectively

Changing the simulation model along with mobilitytraces to Glasgow city center the effect on delay performancewith 150 vehicles at 1 Hz and 10Hz is shown in Figures 8(b)and 8(c) respectively It is interesting to observe the effect ofchanging the mobility model to a European city Where each

city block is 400m in theMGmodel average city block size inGCC is around 200m increasing vehicle density eventuallygrowing the average number of neighbors for GCC modelThis can be observed for an awareness range of 250m theprobability of delay to be le50ms decreases from 77 to 68while for 1000m this deterioration goes from 54 to 37hence an overall decrease of about 9 to 17 for variousawareness ranges Similarly in Figure 8(c) it is observedthat changing the beacon frequency from 1Hz to 10Hz withan awareness range of 500m the probability of delay to bele50ms drastically decreases from 55 to 22 Whereas foran awareness range of 1000m degradation of about 29 isobserved Performance with a high beacon frequency suchas 10Hz is overall seen to be poor That is because packetsthat arrive late are discarded and are mainly sent by cell-edgeusers for which their neighborhood very likely includes usersat the cell edge Hence the delay from the uplink is indirectlyused as a spatial filtering of low quality cell-edge users Thissuggests that if the information is not useful when receivedafter 100ms the better use of resources is to drop the packetat the minimum latency requirement by the server

14 Wireless Communications and Mobile Computing

0010203040506070809

1

0 20 40 60 80 100 120 140 160 180 200

CDF

End-to-end delay (ms)

SAI = 11003816100381610038161003816Fi

1003816100381610038161003816 = 19

SAI = 31003816100381610038161003816Fi

1003816100381610038161003816 = 16

SAI = 51003816100381610038161003816Fi

1003816100381610038161003816 = 41

SAI = 61003816100381610038161003816Fi

1003816100381610038161003816 = 14

SAI = 71003816100381610038161003816Fi

1003816100381610038161003816 = 56

SAI = 91003816100381610038161003816Fi

1003816100381610038161003816 = 54

Without SAI 1003816100381610038161003816Fi1003816100381610038161003816 = 16

(a)

0010203040506070809

1

0 20 40 60 80 100 120 140 160 180 200

CDF

End-to-end delay (ms)

SAI = 11003816100381610038161003816Fi

1003816100381610038161003816 = 26

SAI = 31003816100381610038161003816Fi

1003816100381610038161003816 = 2

SAI = 51003816100381610038161003816Fi

1003816100381610038161003816 = 35

SAI = 61003816100381610038161003816Fi

1003816100381610038161003816 = 22

SAI = 71003816100381610038161003816Fi

1003816100381610038161003816 = 78

SAI = 91003816100381610038161003816Fi

1003816100381610038161003816 = 94

Without SAI 1003816100381610038161003816Fi1003816100381610038161003816 = 21

(b)

Figure 9 LTE end-to-end delay at average speed 40 kmh (GCC) and (a) 100 vehicles and (b) 150 vehicles

Next we observe and compare the results after theproposed algorithm implementation Figure 9 shows theCDF of the end-to-end delay for 100 and 150 vehicles atan average speed of 40 kmh with and without the imple-mentation of SAI algorithm The scenario being comparedhas the awareness range set to 500m and beacon frequencyto 10Hz We observe a significant decrease in the end-to-end delay after SAI algorithm implementation Probabilityfor the end-to-end delay to be less than 100ms is above80 in both scenarios whereas without the algorithm thisprobability is below 70 for 100 vehicles (Figure 9(a)) andbelow 60 for 150 vehicles (Figure 9(b)) The variation inthis delay between various SAIs is because of the differencein transmission parameters Larger awareness range leadingto higher number of neighbors |F119894| and higher beaconfrequency results in more congestion and large queuingtimes The reason for better delay values with SAI algorithmis mainly the result from restricting resources to the requiredtransmission parameters for safety applicationsmentioned inSection 32

Figure 10 shows the aggregate downlink goodput for thesame scenario investigated for end-to-end delay After theimplementation of SAI algorithm the downlink goodput for100 vehicles dropped from 746Mbps to 622Mbps and for150 vehicles it dropped from 2167Mbps to 1376Mbps Thissignificant decrease in the downlink direction is again due tothe restriction of unnecessary data dissemination from theVSA server to vehicles Eventually this decrease of goodputresults in less load on the LTE system The increase in thedownlink goodput for the scenario without SAI explains thehigh end-to-end delay observed in Figure 9 showing a burstof packet in the downlink leading to waiting time in thequeue

Furthermore the impact of regular cellular traffic on ourLTE vehicular network is studied in order to validate ourfindings and algorithm A broad picture of our observa-tions under different scenarios is shown in Figure 11 whereprobability of end-to-end delay being less than or equal to50ms is studied As expected this probability decreases byabout 3-4 when background traffic is added However with

100 150

Number of vehicles

25

20

15

10

5

0

Dow

nlin

k go

odpu

t (M

bps)

Without SAIWith SAI

746622

2167

1376

Figure 10 Downlink goodput for 100 and 150 vehicles at an averagespeed 40 kmh (GCC)

SAI algorithm probability of end-to-end delay being lessthan 50ms stays above 90 even after adding backgroundtraffic

Finally we include the SAI into MAC layer PDU con-trol elements to prove our concept of D2D like unicasttransmission between sender and receivers Modeling of thesystem is carried out in accordance with the description inSection 22 Figure 12 shows end-to-end delay for 100 and150 vehicles with and without the inclusion of SAI in MAClayer It is evident that 150 vehicles experience almost thesame probability for end-to-end delay being less than 50msas experienced by 100 vehicles This shows that with the useof MAC layer control elements around 50 more vehiclescan be accommodated by LTE network increasing the spec-trum efficiency eventually leading to higher capacity of thesystem

Wireless Communications and Mobile Computing 15

Simple case With backgroundtraffic

With SAI With backgroundand SAI

6065707580859095

100

150 vehicles100 vehicles

Prob

abili

ty o

f end

-to-e

ndde

layle

50

ms

Scenarios

Figure 11 Probability of end-to-end delay being le50ms for 100 and150 vehicles at 40 kmh in various scenarios

With SAI in APP layer With SAI in MAC layer80828486889092949698

100

100 vehicles150 vehicles

p

roba

bilit

y fo

r end

-to-e

ndde

layle50

ms

Figure 12 Probability of end-to-end delay being le50ms for 100 and150 vehicles with and without MAC layer inclusion

5 Conclusion

This paper proposes the safety application identifier conceptin the form of an algorithm that is tested and analyzedwithin the impact of vehicular urbanmulticell radio environ-ment employing multipath fading channels and backgroundtraffic The proposed algorithm uses dynamic adaptation oftransmission parameters in order to fulfill stringent vehicu-lar application requirements while minimizing latency andreducing system load

With the help of extensive system-level simulations thecrucial impact of channel modeling and mobility traces isevident with its influence on the received signal qualitydecreasing the probability of end-to-end delay to be le50msfor various awareness ranges by 19 to 30 and 9 to17 respectively Studying the impact of awareness rangeit is observed that with range up to 1000m for 1Hz 500mfor 2Hz and 250m for 10Hz beacon frequency the systemperformed well With the help of these findings the SAIalgorithm is proposed and implemented Probability of expe-riencing an end-to-end delay of less than 100ms increasedby around 20 and the downlink goodput decreased signif-icantly from 2167Mbps to 1376Mbps for 150 vehicles

Furthermore the impact of having background traffic isalso taken into consideration Results show that the systemis slightly affected by having regular background trafficHowever the probability of end-to-end delay being le50ms

still stays satisfactorily above 85 with the use of SAI Thispaper also proposes the inclusion of SAI inMAC layer controlelements producing a D2D type of communication withinvehicles After modeling this system it was found that thesame amount of system resources and requirements are metwith additional 50 vehicles increasing the capacity of ourLTE vehicular network Finally in future works we plan tointegrate SAI incorporated within MAC layer in a hetero-geneous network with DSRC while exploring the networkslicing in 5G cellular networks for vehicular communicationscomplementing each otherrsquos weaknesses and strengths inorder to meet the vehicular network requirements

Conflicts of Interest

The authors declare that they have no conflicts of interest

Acknowledgments

Marvin Sanchez would like to acknowledge the support ofLAMENITEC Erasmus Mundus Program of the EuropeanUnion to be a guest researcher at GCU Results were obtainedusing the EPSRC funded ARCHIE-WeSt High PerformanceComputer (httpswwwarchie-westacuk) EPSRC Grantno EPK0005861

References

[1] S Sesia I Toufik and M Baker LTE the UMTS long termevolution from theory to practice Wiley Publishing 2009

[2] H Hartenstein and K P Laberteaux ldquoA tutorial survey onvehicular ad hoc networksrdquo IEEE Communications Magazinevol 46 no 6 pp 164ndash171 2008

[3] D Jiang and L Delgrossi ldquoTowards an international standardfor wireless access in vehicular environmentsrdquo in Proceedings ofthe 67th IEEE Vehicular Technology Conference (VTC rsquo08) pp2036ndash2040 May 2008

[4] S Al-Sultan M M Al-Doori A H Al-Bayatti and H ZedanldquoA comprehensive survey on vehicular Ad Hoc networkrdquoJournal of Network and Computer Applications vol 37 no 1 pp380ndash392 2014

[5] K Tanuja T Sushma M Bharathi and K Arun ldquoA surveyon VANET technologiesrdquo International Journal of ComputerApplications vol 121 no 18 2015

[6] H Y Kim D M Kang J H Lee and T M Chung ldquoA per-formance evaluation of cellular network suitability for VANETrdquoWorld Academy of Science Engineering and Technology Interna-tional Science Index vol 64 no 6 pp 124ndash127 2012

[7] A Fasbender M Gerdes and S Smets ldquoCellular NetworkingTechnologies in ITS Solutions Opportunities and Challengesrdquopp 1ndash13 Springer Berlin Germany 2012

[8] ldquoCellular coverage and tower map for ee network in glasgowrdquohttpswwwcellmappernetmap

[9] GAraniti C CampoloMCondoluci A Iera andAMolinaroldquoLTE for vehicular networking a surveyrdquo IEEECommunicationsMagazine vol 51 no 5 pp 148ndash157 2013

[10] A Vinel ldquo3GPP LTE versus IEEE 80211pWAVE which tech-nology is able to support cooperative vehicular safety applica-tionsrdquo IEEE Wireless Communications Letters vol 1 no 2 pp125ndash128 2012

12 Wireless Communications and Mobile Computing

0 50 100 150 200 250 3000

01

02

03

04

05

06

07

08

09

1CD

F

End-to-end delay (ms)

R = 100

R = 250

R = 500

R = 750

R = 1000

(a)

0

01

02

03

04

05

06

07

08

09

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 1

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 35

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 14

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 25

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 41

(b)

0

01

02

03

04

05

06

07

08

09

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 1

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 35

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 14

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 25

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 41

(c)

Figure 7 LTE end-to-end delay for 100 vehicles at a beacon frequency 1Hz and average speed 40 kmh with Friis radio environment (a)single cell (b) multicell (4 sites 3 sectorssite) (c) multicell with EVA fading radio environment

The results in Figure 7(b) show that under the Friis radiopropagation environment the capacity increases and thesystem performance is not significantly affected in terms ofthe end-to-end delay in presence of intercell interference andhandover since the scheduler effectively managed interfer-ence by time and frequency allocations However in an urbanenvironment it is expected that fading caused by buildingsand Doppler shift due to the relative velocity of vehicles havea very important impact on the received signal quality [49]Figure 7(c) shows the results when the radio propagationenvironment was changed to Log-distance-dependent and

3GPP EVA under the same site configurations Under thisradio environment the probability for the end-to-end delayto be less than or equal to 100ms was around 76ndash78 forawareness range between 100m and 500m 70 for 750mand 66 for 1000m Hence the significant effect on the end-to-end delay is because of path loss and frequency selectivefading imposed by the EVA fading environmentThe end-to-enddelay for beacons to and fromcell-edge users significantlyincreases Similarly for inner cell users the end-to-end delayremained close to the Friis results and for up to 500m theend-to-end delay was le50ms with 74 to 71 probability

Wireless Communications and Mobile Computing 13

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 17

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 53

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 20

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 37

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 62

(a)

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 11

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 69

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 21

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 47

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 73

(b)

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 13

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 63

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 24

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 47

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 74

(c)

Figure 8 End-to-end delay for sample networks at average speed 40 kmh with 150 vehicles at beacon frequency (a) 1Hz with MG model(b) 1Hz with GCC model (c) 10Hz with GCC model

65 for 750m and 59 for 1000m Therefore it can bededuced that with the addition of multipath fading alongwith critical radio propagation environment probability ofdelay being le50ms decreases by 19 to 30

Continuing on with multicell and EVA fading radioenvironment by increasing the network size the number ofvehicles in the awareness range increases and the end-to-enddelay increases as shown in Figure 8(a) Nevertheless theend-to-end delay remained below 100ms for up to 250mwith82 probability for 100 vehicles and similar rate was obtainedwith up to 150 vehicles When the awareness range increasesto 500m the probability was 80 for 100 vehicles while for150 vehicles further reduces to 73 For awareness range of750m and 1000m the probability reduced to 75 and 71 for100 vehicles while for 150 vehicles reduces to 69 and 60respectively

Changing the simulation model along with mobilitytraces to Glasgow city center the effect on delay performancewith 150 vehicles at 1 Hz and 10Hz is shown in Figures 8(b)and 8(c) respectively It is interesting to observe the effect ofchanging the mobility model to a European city Where each

city block is 400m in theMGmodel average city block size inGCC is around 200m increasing vehicle density eventuallygrowing the average number of neighbors for GCC modelThis can be observed for an awareness range of 250m theprobability of delay to be le50ms decreases from 77 to 68while for 1000m this deterioration goes from 54 to 37hence an overall decrease of about 9 to 17 for variousawareness ranges Similarly in Figure 8(c) it is observedthat changing the beacon frequency from 1Hz to 10Hz withan awareness range of 500m the probability of delay to bele50ms drastically decreases from 55 to 22 Whereas foran awareness range of 1000m degradation of about 29 isobserved Performance with a high beacon frequency suchas 10Hz is overall seen to be poor That is because packetsthat arrive late are discarded and are mainly sent by cell-edgeusers for which their neighborhood very likely includes usersat the cell edge Hence the delay from the uplink is indirectlyused as a spatial filtering of low quality cell-edge users Thissuggests that if the information is not useful when receivedafter 100ms the better use of resources is to drop the packetat the minimum latency requirement by the server

14 Wireless Communications and Mobile Computing

0010203040506070809

1

0 20 40 60 80 100 120 140 160 180 200

CDF

End-to-end delay (ms)

SAI = 11003816100381610038161003816Fi

1003816100381610038161003816 = 19

SAI = 31003816100381610038161003816Fi

1003816100381610038161003816 = 16

SAI = 51003816100381610038161003816Fi

1003816100381610038161003816 = 41

SAI = 61003816100381610038161003816Fi

1003816100381610038161003816 = 14

SAI = 71003816100381610038161003816Fi

1003816100381610038161003816 = 56

SAI = 91003816100381610038161003816Fi

1003816100381610038161003816 = 54

Without SAI 1003816100381610038161003816Fi1003816100381610038161003816 = 16

(a)

0010203040506070809

1

0 20 40 60 80 100 120 140 160 180 200

CDF

End-to-end delay (ms)

SAI = 11003816100381610038161003816Fi

1003816100381610038161003816 = 26

SAI = 31003816100381610038161003816Fi

1003816100381610038161003816 = 2

SAI = 51003816100381610038161003816Fi

1003816100381610038161003816 = 35

SAI = 61003816100381610038161003816Fi

1003816100381610038161003816 = 22

SAI = 71003816100381610038161003816Fi

1003816100381610038161003816 = 78

SAI = 91003816100381610038161003816Fi

1003816100381610038161003816 = 94

Without SAI 1003816100381610038161003816Fi1003816100381610038161003816 = 21

(b)

Figure 9 LTE end-to-end delay at average speed 40 kmh (GCC) and (a) 100 vehicles and (b) 150 vehicles

Next we observe and compare the results after theproposed algorithm implementation Figure 9 shows theCDF of the end-to-end delay for 100 and 150 vehicles atan average speed of 40 kmh with and without the imple-mentation of SAI algorithm The scenario being comparedhas the awareness range set to 500m and beacon frequencyto 10Hz We observe a significant decrease in the end-to-end delay after SAI algorithm implementation Probabilityfor the end-to-end delay to be less than 100ms is above80 in both scenarios whereas without the algorithm thisprobability is below 70 for 100 vehicles (Figure 9(a)) andbelow 60 for 150 vehicles (Figure 9(b)) The variation inthis delay between various SAIs is because of the differencein transmission parameters Larger awareness range leadingto higher number of neighbors |F119894| and higher beaconfrequency results in more congestion and large queuingtimes The reason for better delay values with SAI algorithmis mainly the result from restricting resources to the requiredtransmission parameters for safety applicationsmentioned inSection 32

Figure 10 shows the aggregate downlink goodput for thesame scenario investigated for end-to-end delay After theimplementation of SAI algorithm the downlink goodput for100 vehicles dropped from 746Mbps to 622Mbps and for150 vehicles it dropped from 2167Mbps to 1376Mbps Thissignificant decrease in the downlink direction is again due tothe restriction of unnecessary data dissemination from theVSA server to vehicles Eventually this decrease of goodputresults in less load on the LTE system The increase in thedownlink goodput for the scenario without SAI explains thehigh end-to-end delay observed in Figure 9 showing a burstof packet in the downlink leading to waiting time in thequeue

Furthermore the impact of regular cellular traffic on ourLTE vehicular network is studied in order to validate ourfindings and algorithm A broad picture of our observa-tions under different scenarios is shown in Figure 11 whereprobability of end-to-end delay being less than or equal to50ms is studied As expected this probability decreases byabout 3-4 when background traffic is added However with

100 150

Number of vehicles

25

20

15

10

5

0

Dow

nlin

k go

odpu

t (M

bps)

Without SAIWith SAI

746622

2167

1376

Figure 10 Downlink goodput for 100 and 150 vehicles at an averagespeed 40 kmh (GCC)

SAI algorithm probability of end-to-end delay being lessthan 50ms stays above 90 even after adding backgroundtraffic

Finally we include the SAI into MAC layer PDU con-trol elements to prove our concept of D2D like unicasttransmission between sender and receivers Modeling of thesystem is carried out in accordance with the description inSection 22 Figure 12 shows end-to-end delay for 100 and150 vehicles with and without the inclusion of SAI in MAClayer It is evident that 150 vehicles experience almost thesame probability for end-to-end delay being less than 50msas experienced by 100 vehicles This shows that with the useof MAC layer control elements around 50 more vehiclescan be accommodated by LTE network increasing the spec-trum efficiency eventually leading to higher capacity of thesystem

Wireless Communications and Mobile Computing 15

Simple case With backgroundtraffic

With SAI With backgroundand SAI

6065707580859095

100

150 vehicles100 vehicles

Prob

abili

ty o

f end

-to-e

ndde

layle

50

ms

Scenarios

Figure 11 Probability of end-to-end delay being le50ms for 100 and150 vehicles at 40 kmh in various scenarios

With SAI in APP layer With SAI in MAC layer80828486889092949698

100

100 vehicles150 vehicles

p

roba

bilit

y fo

r end

-to-e

ndde

layle50

ms

Figure 12 Probability of end-to-end delay being le50ms for 100 and150 vehicles with and without MAC layer inclusion

5 Conclusion

This paper proposes the safety application identifier conceptin the form of an algorithm that is tested and analyzedwithin the impact of vehicular urbanmulticell radio environ-ment employing multipath fading channels and backgroundtraffic The proposed algorithm uses dynamic adaptation oftransmission parameters in order to fulfill stringent vehicu-lar application requirements while minimizing latency andreducing system load

With the help of extensive system-level simulations thecrucial impact of channel modeling and mobility traces isevident with its influence on the received signal qualitydecreasing the probability of end-to-end delay to be le50msfor various awareness ranges by 19 to 30 and 9 to17 respectively Studying the impact of awareness rangeit is observed that with range up to 1000m for 1Hz 500mfor 2Hz and 250m for 10Hz beacon frequency the systemperformed well With the help of these findings the SAIalgorithm is proposed and implemented Probability of expe-riencing an end-to-end delay of less than 100ms increasedby around 20 and the downlink goodput decreased signif-icantly from 2167Mbps to 1376Mbps for 150 vehicles

Furthermore the impact of having background traffic isalso taken into consideration Results show that the systemis slightly affected by having regular background trafficHowever the probability of end-to-end delay being le50ms

still stays satisfactorily above 85 with the use of SAI Thispaper also proposes the inclusion of SAI inMAC layer controlelements producing a D2D type of communication withinvehicles After modeling this system it was found that thesame amount of system resources and requirements are metwith additional 50 vehicles increasing the capacity of ourLTE vehicular network Finally in future works we plan tointegrate SAI incorporated within MAC layer in a hetero-geneous network with DSRC while exploring the networkslicing in 5G cellular networks for vehicular communicationscomplementing each otherrsquos weaknesses and strengths inorder to meet the vehicular network requirements

Conflicts of Interest

The authors declare that they have no conflicts of interest

Acknowledgments

Marvin Sanchez would like to acknowledge the support ofLAMENITEC Erasmus Mundus Program of the EuropeanUnion to be a guest researcher at GCU Results were obtainedusing the EPSRC funded ARCHIE-WeSt High PerformanceComputer (httpswwwarchie-westacuk) EPSRC Grantno EPK0005861

References

[1] S Sesia I Toufik and M Baker LTE the UMTS long termevolution from theory to practice Wiley Publishing 2009

[2] H Hartenstein and K P Laberteaux ldquoA tutorial survey onvehicular ad hoc networksrdquo IEEE Communications Magazinevol 46 no 6 pp 164ndash171 2008

[3] D Jiang and L Delgrossi ldquoTowards an international standardfor wireless access in vehicular environmentsrdquo in Proceedings ofthe 67th IEEE Vehicular Technology Conference (VTC rsquo08) pp2036ndash2040 May 2008

[4] S Al-Sultan M M Al-Doori A H Al-Bayatti and H ZedanldquoA comprehensive survey on vehicular Ad Hoc networkrdquoJournal of Network and Computer Applications vol 37 no 1 pp380ndash392 2014

[5] K Tanuja T Sushma M Bharathi and K Arun ldquoA surveyon VANET technologiesrdquo International Journal of ComputerApplications vol 121 no 18 2015

[6] H Y Kim D M Kang J H Lee and T M Chung ldquoA per-formance evaluation of cellular network suitability for VANETrdquoWorld Academy of Science Engineering and Technology Interna-tional Science Index vol 64 no 6 pp 124ndash127 2012

[7] A Fasbender M Gerdes and S Smets ldquoCellular NetworkingTechnologies in ITS Solutions Opportunities and Challengesrdquopp 1ndash13 Springer Berlin Germany 2012

[8] ldquoCellular coverage and tower map for ee network in glasgowrdquohttpswwwcellmappernetmap

[9] GAraniti C CampoloMCondoluci A Iera andAMolinaroldquoLTE for vehicular networking a surveyrdquo IEEECommunicationsMagazine vol 51 no 5 pp 148ndash157 2013

[10] A Vinel ldquo3GPP LTE versus IEEE 80211pWAVE which tech-nology is able to support cooperative vehicular safety applica-tionsrdquo IEEE Wireless Communications Letters vol 1 no 2 pp125ndash128 2012

Wireless Communications and Mobile Computing 13

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 17

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 53

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 20

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 37

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 62

(a)

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 11

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 69

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 21

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 47

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 73

(b)

0010203040506070809

1

0 50 100 150 200 250 300

CDF

End-to-end delay (ms)

R = 1001003816100381610038161003816Fi

1003816100381610038161003816 = 13

R = 2501003816100381610038161003816Fi

1003816100381610038161003816 = 63

R = 5001003816100381610038161003816Fi

1003816100381610038161003816 = 24

R = 7501003816100381610038161003816Fi

1003816100381610038161003816 = 47

R = 10001003816100381610038161003816Fi

1003816100381610038161003816 = 74

(c)

Figure 8 End-to-end delay for sample networks at average speed 40 kmh with 150 vehicles at beacon frequency (a) 1Hz with MG model(b) 1Hz with GCC model (c) 10Hz with GCC model

65 for 750m and 59 for 1000m Therefore it can bededuced that with the addition of multipath fading alongwith critical radio propagation environment probability ofdelay being le50ms decreases by 19 to 30

Continuing on with multicell and EVA fading radioenvironment by increasing the network size the number ofvehicles in the awareness range increases and the end-to-enddelay increases as shown in Figure 8(a) Nevertheless theend-to-end delay remained below 100ms for up to 250mwith82 probability for 100 vehicles and similar rate was obtainedwith up to 150 vehicles When the awareness range increasesto 500m the probability was 80 for 100 vehicles while for150 vehicles further reduces to 73 For awareness range of750m and 1000m the probability reduced to 75 and 71 for100 vehicles while for 150 vehicles reduces to 69 and 60respectively

Changing the simulation model along with mobilitytraces to Glasgow city center the effect on delay performancewith 150 vehicles at 1 Hz and 10Hz is shown in Figures 8(b)and 8(c) respectively It is interesting to observe the effect ofchanging the mobility model to a European city Where each

city block is 400m in theMGmodel average city block size inGCC is around 200m increasing vehicle density eventuallygrowing the average number of neighbors for GCC modelThis can be observed for an awareness range of 250m theprobability of delay to be le50ms decreases from 77 to 68while for 1000m this deterioration goes from 54 to 37hence an overall decrease of about 9 to 17 for variousawareness ranges Similarly in Figure 8(c) it is observedthat changing the beacon frequency from 1Hz to 10Hz withan awareness range of 500m the probability of delay to bele50ms drastically decreases from 55 to 22 Whereas foran awareness range of 1000m degradation of about 29 isobserved Performance with a high beacon frequency suchas 10Hz is overall seen to be poor That is because packetsthat arrive late are discarded and are mainly sent by cell-edgeusers for which their neighborhood very likely includes usersat the cell edge Hence the delay from the uplink is indirectlyused as a spatial filtering of low quality cell-edge users Thissuggests that if the information is not useful when receivedafter 100ms the better use of resources is to drop the packetat the minimum latency requirement by the server

14 Wireless Communications and Mobile Computing

0010203040506070809

1

0 20 40 60 80 100 120 140 160 180 200

CDF

End-to-end delay (ms)

SAI = 11003816100381610038161003816Fi

1003816100381610038161003816 = 19

SAI = 31003816100381610038161003816Fi

1003816100381610038161003816 = 16

SAI = 51003816100381610038161003816Fi

1003816100381610038161003816 = 41

SAI = 61003816100381610038161003816Fi

1003816100381610038161003816 = 14

SAI = 71003816100381610038161003816Fi

1003816100381610038161003816 = 56

SAI = 91003816100381610038161003816Fi

1003816100381610038161003816 = 54

Without SAI 1003816100381610038161003816Fi1003816100381610038161003816 = 16

(a)

0010203040506070809

1

0 20 40 60 80 100 120 140 160 180 200

CDF

End-to-end delay (ms)

SAI = 11003816100381610038161003816Fi

1003816100381610038161003816 = 26

SAI = 31003816100381610038161003816Fi

1003816100381610038161003816 = 2

SAI = 51003816100381610038161003816Fi

1003816100381610038161003816 = 35

SAI = 61003816100381610038161003816Fi

1003816100381610038161003816 = 22

SAI = 71003816100381610038161003816Fi

1003816100381610038161003816 = 78

SAI = 91003816100381610038161003816Fi

1003816100381610038161003816 = 94

Without SAI 1003816100381610038161003816Fi1003816100381610038161003816 = 21

(b)

Figure 9 LTE end-to-end delay at average speed 40 kmh (GCC) and (a) 100 vehicles and (b) 150 vehicles

Next we observe and compare the results after theproposed algorithm implementation Figure 9 shows theCDF of the end-to-end delay for 100 and 150 vehicles atan average speed of 40 kmh with and without the imple-mentation of SAI algorithm The scenario being comparedhas the awareness range set to 500m and beacon frequencyto 10Hz We observe a significant decrease in the end-to-end delay after SAI algorithm implementation Probabilityfor the end-to-end delay to be less than 100ms is above80 in both scenarios whereas without the algorithm thisprobability is below 70 for 100 vehicles (Figure 9(a)) andbelow 60 for 150 vehicles (Figure 9(b)) The variation inthis delay between various SAIs is because of the differencein transmission parameters Larger awareness range leadingto higher number of neighbors |F119894| and higher beaconfrequency results in more congestion and large queuingtimes The reason for better delay values with SAI algorithmis mainly the result from restricting resources to the requiredtransmission parameters for safety applicationsmentioned inSection 32

Figure 10 shows the aggregate downlink goodput for thesame scenario investigated for end-to-end delay After theimplementation of SAI algorithm the downlink goodput for100 vehicles dropped from 746Mbps to 622Mbps and for150 vehicles it dropped from 2167Mbps to 1376Mbps Thissignificant decrease in the downlink direction is again due tothe restriction of unnecessary data dissemination from theVSA server to vehicles Eventually this decrease of goodputresults in less load on the LTE system The increase in thedownlink goodput for the scenario without SAI explains thehigh end-to-end delay observed in Figure 9 showing a burstof packet in the downlink leading to waiting time in thequeue

Furthermore the impact of regular cellular traffic on ourLTE vehicular network is studied in order to validate ourfindings and algorithm A broad picture of our observa-tions under different scenarios is shown in Figure 11 whereprobability of end-to-end delay being less than or equal to50ms is studied As expected this probability decreases byabout 3-4 when background traffic is added However with

100 150

Number of vehicles

25

20

15

10

5

0

Dow

nlin

k go

odpu

t (M

bps)

Without SAIWith SAI

746622

2167

1376

Figure 10 Downlink goodput for 100 and 150 vehicles at an averagespeed 40 kmh (GCC)

SAI algorithm probability of end-to-end delay being lessthan 50ms stays above 90 even after adding backgroundtraffic

Finally we include the SAI into MAC layer PDU con-trol elements to prove our concept of D2D like unicasttransmission between sender and receivers Modeling of thesystem is carried out in accordance with the description inSection 22 Figure 12 shows end-to-end delay for 100 and150 vehicles with and without the inclusion of SAI in MAClayer It is evident that 150 vehicles experience almost thesame probability for end-to-end delay being less than 50msas experienced by 100 vehicles This shows that with the useof MAC layer control elements around 50 more vehiclescan be accommodated by LTE network increasing the spec-trum efficiency eventually leading to higher capacity of thesystem

Wireless Communications and Mobile Computing 15

Simple case With backgroundtraffic

With SAI With backgroundand SAI

6065707580859095

100

150 vehicles100 vehicles

Prob

abili

ty o

f end

-to-e

ndde

layle

50

ms

Scenarios

Figure 11 Probability of end-to-end delay being le50ms for 100 and150 vehicles at 40 kmh in various scenarios

With SAI in APP layer With SAI in MAC layer80828486889092949698

100

100 vehicles150 vehicles

p

roba

bilit

y fo

r end

-to-e

ndde

layle50

ms

Figure 12 Probability of end-to-end delay being le50ms for 100 and150 vehicles with and without MAC layer inclusion

5 Conclusion

This paper proposes the safety application identifier conceptin the form of an algorithm that is tested and analyzedwithin the impact of vehicular urbanmulticell radio environ-ment employing multipath fading channels and backgroundtraffic The proposed algorithm uses dynamic adaptation oftransmission parameters in order to fulfill stringent vehicu-lar application requirements while minimizing latency andreducing system load

With the help of extensive system-level simulations thecrucial impact of channel modeling and mobility traces isevident with its influence on the received signal qualitydecreasing the probability of end-to-end delay to be le50msfor various awareness ranges by 19 to 30 and 9 to17 respectively Studying the impact of awareness rangeit is observed that with range up to 1000m for 1Hz 500mfor 2Hz and 250m for 10Hz beacon frequency the systemperformed well With the help of these findings the SAIalgorithm is proposed and implemented Probability of expe-riencing an end-to-end delay of less than 100ms increasedby around 20 and the downlink goodput decreased signif-icantly from 2167Mbps to 1376Mbps for 150 vehicles

Furthermore the impact of having background traffic isalso taken into consideration Results show that the systemis slightly affected by having regular background trafficHowever the probability of end-to-end delay being le50ms

still stays satisfactorily above 85 with the use of SAI Thispaper also proposes the inclusion of SAI inMAC layer controlelements producing a D2D type of communication withinvehicles After modeling this system it was found that thesame amount of system resources and requirements are metwith additional 50 vehicles increasing the capacity of ourLTE vehicular network Finally in future works we plan tointegrate SAI incorporated within MAC layer in a hetero-geneous network with DSRC while exploring the networkslicing in 5G cellular networks for vehicular communicationscomplementing each otherrsquos weaknesses and strengths inorder to meet the vehicular network requirements

Conflicts of Interest

The authors declare that they have no conflicts of interest

Acknowledgments

Marvin Sanchez would like to acknowledge the support ofLAMENITEC Erasmus Mundus Program of the EuropeanUnion to be a guest researcher at GCU Results were obtainedusing the EPSRC funded ARCHIE-WeSt High PerformanceComputer (httpswwwarchie-westacuk) EPSRC Grantno EPK0005861

References

[1] S Sesia I Toufik and M Baker LTE the UMTS long termevolution from theory to practice Wiley Publishing 2009

[2] H Hartenstein and K P Laberteaux ldquoA tutorial survey onvehicular ad hoc networksrdquo IEEE Communications Magazinevol 46 no 6 pp 164ndash171 2008

[3] D Jiang and L Delgrossi ldquoTowards an international standardfor wireless access in vehicular environmentsrdquo in Proceedings ofthe 67th IEEE Vehicular Technology Conference (VTC rsquo08) pp2036ndash2040 May 2008

[4] S Al-Sultan M M Al-Doori A H Al-Bayatti and H ZedanldquoA comprehensive survey on vehicular Ad Hoc networkrdquoJournal of Network and Computer Applications vol 37 no 1 pp380ndash392 2014

[5] K Tanuja T Sushma M Bharathi and K Arun ldquoA surveyon VANET technologiesrdquo International Journal of ComputerApplications vol 121 no 18 2015

[6] H Y Kim D M Kang J H Lee and T M Chung ldquoA per-formance evaluation of cellular network suitability for VANETrdquoWorld Academy of Science Engineering and Technology Interna-tional Science Index vol 64 no 6 pp 124ndash127 2012

[7] A Fasbender M Gerdes and S Smets ldquoCellular NetworkingTechnologies in ITS Solutions Opportunities and Challengesrdquopp 1ndash13 Springer Berlin Germany 2012

[8] ldquoCellular coverage and tower map for ee network in glasgowrdquohttpswwwcellmappernetmap

[9] GAraniti C CampoloMCondoluci A Iera andAMolinaroldquoLTE for vehicular networking a surveyrdquo IEEECommunicationsMagazine vol 51 no 5 pp 148ndash157 2013

[10] A Vinel ldquo3GPP LTE versus IEEE 80211pWAVE which tech-nology is able to support cooperative vehicular safety applica-tionsrdquo IEEE Wireless Communications Letters vol 1 no 2 pp125ndash128 2012

14 Wireless Communications and Mobile Computing

0010203040506070809

1

0 20 40 60 80 100 120 140 160 180 200

CDF

End-to-end delay (ms)

SAI = 11003816100381610038161003816Fi

1003816100381610038161003816 = 19

SAI = 31003816100381610038161003816Fi

1003816100381610038161003816 = 16

SAI = 51003816100381610038161003816Fi

1003816100381610038161003816 = 41

SAI = 61003816100381610038161003816Fi

1003816100381610038161003816 = 14

SAI = 71003816100381610038161003816Fi

1003816100381610038161003816 = 56

SAI = 91003816100381610038161003816Fi

1003816100381610038161003816 = 54

Without SAI 1003816100381610038161003816Fi1003816100381610038161003816 = 16

(a)

0010203040506070809

1

0 20 40 60 80 100 120 140 160 180 200

CDF

End-to-end delay (ms)

SAI = 11003816100381610038161003816Fi

1003816100381610038161003816 = 26

SAI = 31003816100381610038161003816Fi

1003816100381610038161003816 = 2

SAI = 51003816100381610038161003816Fi

1003816100381610038161003816 = 35

SAI = 61003816100381610038161003816Fi

1003816100381610038161003816 = 22

SAI = 71003816100381610038161003816Fi

1003816100381610038161003816 = 78

SAI = 91003816100381610038161003816Fi

1003816100381610038161003816 = 94

Without SAI 1003816100381610038161003816Fi1003816100381610038161003816 = 21

(b)

Figure 9 LTE end-to-end delay at average speed 40 kmh (GCC) and (a) 100 vehicles and (b) 150 vehicles

Next we observe and compare the results after theproposed algorithm implementation Figure 9 shows theCDF of the end-to-end delay for 100 and 150 vehicles atan average speed of 40 kmh with and without the imple-mentation of SAI algorithm The scenario being comparedhas the awareness range set to 500m and beacon frequencyto 10Hz We observe a significant decrease in the end-to-end delay after SAI algorithm implementation Probabilityfor the end-to-end delay to be less than 100ms is above80 in both scenarios whereas without the algorithm thisprobability is below 70 for 100 vehicles (Figure 9(a)) andbelow 60 for 150 vehicles (Figure 9(b)) The variation inthis delay between various SAIs is because of the differencein transmission parameters Larger awareness range leadingto higher number of neighbors |F119894| and higher beaconfrequency results in more congestion and large queuingtimes The reason for better delay values with SAI algorithmis mainly the result from restricting resources to the requiredtransmission parameters for safety applicationsmentioned inSection 32

Figure 10 shows the aggregate downlink goodput for thesame scenario investigated for end-to-end delay After theimplementation of SAI algorithm the downlink goodput for100 vehicles dropped from 746Mbps to 622Mbps and for150 vehicles it dropped from 2167Mbps to 1376Mbps Thissignificant decrease in the downlink direction is again due tothe restriction of unnecessary data dissemination from theVSA server to vehicles Eventually this decrease of goodputresults in less load on the LTE system The increase in thedownlink goodput for the scenario without SAI explains thehigh end-to-end delay observed in Figure 9 showing a burstof packet in the downlink leading to waiting time in thequeue

Furthermore the impact of regular cellular traffic on ourLTE vehicular network is studied in order to validate ourfindings and algorithm A broad picture of our observa-tions under different scenarios is shown in Figure 11 whereprobability of end-to-end delay being less than or equal to50ms is studied As expected this probability decreases byabout 3-4 when background traffic is added However with

100 150

Number of vehicles

25

20

15

10

5

0

Dow

nlin

k go

odpu

t (M

bps)

Without SAIWith SAI

746622

2167

1376

Figure 10 Downlink goodput for 100 and 150 vehicles at an averagespeed 40 kmh (GCC)

SAI algorithm probability of end-to-end delay being lessthan 50ms stays above 90 even after adding backgroundtraffic

Finally we include the SAI into MAC layer PDU con-trol elements to prove our concept of D2D like unicasttransmission between sender and receivers Modeling of thesystem is carried out in accordance with the description inSection 22 Figure 12 shows end-to-end delay for 100 and150 vehicles with and without the inclusion of SAI in MAClayer It is evident that 150 vehicles experience almost thesame probability for end-to-end delay being less than 50msas experienced by 100 vehicles This shows that with the useof MAC layer control elements around 50 more vehiclescan be accommodated by LTE network increasing the spec-trum efficiency eventually leading to higher capacity of thesystem

Wireless Communications and Mobile Computing 15

Simple case With backgroundtraffic

With SAI With backgroundand SAI

6065707580859095

100

150 vehicles100 vehicles

Prob

abili

ty o

f end

-to-e

ndde

layle

50

ms

Scenarios

Figure 11 Probability of end-to-end delay being le50ms for 100 and150 vehicles at 40 kmh in various scenarios

With SAI in APP layer With SAI in MAC layer80828486889092949698

100

100 vehicles150 vehicles

p

roba

bilit

y fo

r end

-to-e

ndde

layle50

ms

Figure 12 Probability of end-to-end delay being le50ms for 100 and150 vehicles with and without MAC layer inclusion

5 Conclusion

This paper proposes the safety application identifier conceptin the form of an algorithm that is tested and analyzedwithin the impact of vehicular urbanmulticell radio environ-ment employing multipath fading channels and backgroundtraffic The proposed algorithm uses dynamic adaptation oftransmission parameters in order to fulfill stringent vehicu-lar application requirements while minimizing latency andreducing system load

With the help of extensive system-level simulations thecrucial impact of channel modeling and mobility traces isevident with its influence on the received signal qualitydecreasing the probability of end-to-end delay to be le50msfor various awareness ranges by 19 to 30 and 9 to17 respectively Studying the impact of awareness rangeit is observed that with range up to 1000m for 1Hz 500mfor 2Hz and 250m for 10Hz beacon frequency the systemperformed well With the help of these findings the SAIalgorithm is proposed and implemented Probability of expe-riencing an end-to-end delay of less than 100ms increasedby around 20 and the downlink goodput decreased signif-icantly from 2167Mbps to 1376Mbps for 150 vehicles

Furthermore the impact of having background traffic isalso taken into consideration Results show that the systemis slightly affected by having regular background trafficHowever the probability of end-to-end delay being le50ms

still stays satisfactorily above 85 with the use of SAI Thispaper also proposes the inclusion of SAI inMAC layer controlelements producing a D2D type of communication withinvehicles After modeling this system it was found that thesame amount of system resources and requirements are metwith additional 50 vehicles increasing the capacity of ourLTE vehicular network Finally in future works we plan tointegrate SAI incorporated within MAC layer in a hetero-geneous network with DSRC while exploring the networkslicing in 5G cellular networks for vehicular communicationscomplementing each otherrsquos weaknesses and strengths inorder to meet the vehicular network requirements

Conflicts of Interest

The authors declare that they have no conflicts of interest

Acknowledgments

Marvin Sanchez would like to acknowledge the support ofLAMENITEC Erasmus Mundus Program of the EuropeanUnion to be a guest researcher at GCU Results were obtainedusing the EPSRC funded ARCHIE-WeSt High PerformanceComputer (httpswwwarchie-westacuk) EPSRC Grantno EPK0005861

References

[1] S Sesia I Toufik and M Baker LTE the UMTS long termevolution from theory to practice Wiley Publishing 2009

[2] H Hartenstein and K P Laberteaux ldquoA tutorial survey onvehicular ad hoc networksrdquo IEEE Communications Magazinevol 46 no 6 pp 164ndash171 2008

[3] D Jiang and L Delgrossi ldquoTowards an international standardfor wireless access in vehicular environmentsrdquo in Proceedings ofthe 67th IEEE Vehicular Technology Conference (VTC rsquo08) pp2036ndash2040 May 2008

[4] S Al-Sultan M M Al-Doori A H Al-Bayatti and H ZedanldquoA comprehensive survey on vehicular Ad Hoc networkrdquoJournal of Network and Computer Applications vol 37 no 1 pp380ndash392 2014

[5] K Tanuja T Sushma M Bharathi and K Arun ldquoA surveyon VANET technologiesrdquo International Journal of ComputerApplications vol 121 no 18 2015

[6] H Y Kim D M Kang J H Lee and T M Chung ldquoA per-formance evaluation of cellular network suitability for VANETrdquoWorld Academy of Science Engineering and Technology Interna-tional Science Index vol 64 no 6 pp 124ndash127 2012

[7] A Fasbender M Gerdes and S Smets ldquoCellular NetworkingTechnologies in ITS Solutions Opportunities and Challengesrdquopp 1ndash13 Springer Berlin Germany 2012

[8] ldquoCellular coverage and tower map for ee network in glasgowrdquohttpswwwcellmappernetmap

[9] GAraniti C CampoloMCondoluci A Iera andAMolinaroldquoLTE for vehicular networking a surveyrdquo IEEECommunicationsMagazine vol 51 no 5 pp 148ndash157 2013

[10] A Vinel ldquo3GPP LTE versus IEEE 80211pWAVE which tech-nology is able to support cooperative vehicular safety applica-tionsrdquo IEEE Wireless Communications Letters vol 1 no 2 pp125ndash128 2012

Wireless Communications and Mobile Computing 15

Simple case With backgroundtraffic

With SAI With backgroundand SAI

6065707580859095

100

150 vehicles100 vehicles

Prob

abili

ty o

f end

-to-e

ndde

layle

50

ms

Scenarios

Figure 11 Probability of end-to-end delay being le50ms for 100 and150 vehicles at 40 kmh in various scenarios

With SAI in APP layer With SAI in MAC layer80828486889092949698

100

100 vehicles150 vehicles

p

roba

bilit

y fo

r end

-to-e

ndde

layle50

ms

Figure 12 Probability of end-to-end delay being le50ms for 100 and150 vehicles with and without MAC layer inclusion

5 Conclusion

This paper proposes the safety application identifier conceptin the form of an algorithm that is tested and analyzedwithin the impact of vehicular urbanmulticell radio environ-ment employing multipath fading channels and backgroundtraffic The proposed algorithm uses dynamic adaptation oftransmission parameters in order to fulfill stringent vehicu-lar application requirements while minimizing latency andreducing system load

With the help of extensive system-level simulations thecrucial impact of channel modeling and mobility traces isevident with its influence on the received signal qualitydecreasing the probability of end-to-end delay to be le50msfor various awareness ranges by 19 to 30 and 9 to17 respectively Studying the impact of awareness rangeit is observed that with range up to 1000m for 1Hz 500mfor 2Hz and 250m for 10Hz beacon frequency the systemperformed well With the help of these findings the SAIalgorithm is proposed and implemented Probability of expe-riencing an end-to-end delay of less than 100ms increasedby around 20 and the downlink goodput decreased signif-icantly from 2167Mbps to 1376Mbps for 150 vehicles

Furthermore the impact of having background traffic isalso taken into consideration Results show that the systemis slightly affected by having regular background trafficHowever the probability of end-to-end delay being le50ms

still stays satisfactorily above 85 with the use of SAI Thispaper also proposes the inclusion of SAI inMAC layer controlelements producing a D2D type of communication withinvehicles After modeling this system it was found that thesame amount of system resources and requirements are metwith additional 50 vehicles increasing the capacity of ourLTE vehicular network Finally in future works we plan tointegrate SAI incorporated within MAC layer in a hetero-geneous network with DSRC while exploring the networkslicing in 5G cellular networks for vehicular communicationscomplementing each otherrsquos weaknesses and strengths inorder to meet the vehicular network requirements

Conflicts of Interest

The authors declare that they have no conflicts of interest

Acknowledgments

Marvin Sanchez would like to acknowledge the support ofLAMENITEC Erasmus Mundus Program of the EuropeanUnion to be a guest researcher at GCU Results were obtainedusing the EPSRC funded ARCHIE-WeSt High PerformanceComputer (httpswwwarchie-westacuk) EPSRC Grantno EPK0005861

References

[1] S Sesia I Toufik and M Baker LTE the UMTS long termevolution from theory to practice Wiley Publishing 2009

[2] H Hartenstein and K P Laberteaux ldquoA tutorial survey onvehicular ad hoc networksrdquo IEEE Communications Magazinevol 46 no 6 pp 164ndash171 2008

[3] D Jiang and L Delgrossi ldquoTowards an international standardfor wireless access in vehicular environmentsrdquo in Proceedings ofthe 67th IEEE Vehicular Technology Conference (VTC rsquo08) pp2036ndash2040 May 2008

[4] S Al-Sultan M M Al-Doori A H Al-Bayatti and H ZedanldquoA comprehensive survey on vehicular Ad Hoc networkrdquoJournal of Network and Computer Applications vol 37 no 1 pp380ndash392 2014

[5] K Tanuja T Sushma M Bharathi and K Arun ldquoA surveyon VANET technologiesrdquo International Journal of ComputerApplications vol 121 no 18 2015

[6] H Y Kim D M Kang J H Lee and T M Chung ldquoA per-formance evaluation of cellular network suitability for VANETrdquoWorld Academy of Science Engineering and Technology Interna-tional Science Index vol 64 no 6 pp 124ndash127 2012

[7] A Fasbender M Gerdes and S Smets ldquoCellular NetworkingTechnologies in ITS Solutions Opportunities and Challengesrdquopp 1ndash13 Springer Berlin Germany 2012

[8] ldquoCellular coverage and tower map for ee network in glasgowrdquohttpswwwcellmappernetmap

[9] GAraniti C CampoloMCondoluci A Iera andAMolinaroldquoLTE for vehicular networking a surveyrdquo IEEECommunicationsMagazine vol 51 no 5 pp 148ndash157 2013

[10] A Vinel ldquo3GPP LTE versus IEEE 80211pWAVE which tech-nology is able to support cooperative vehicular safety applica-tionsrdquo IEEE Wireless Communications Letters vol 1 no 2 pp125ndash128 2012

16 Wireless Communications and Mobile Computing

[11] Z M Mir and F Filali ldquoLTE and IEEE 80211p for vehicularnetworking a performance evaluationrdquo EURASIP Journal onWireless Communications and Networking vol 2014 no 1article 89 2014

[12] M Phan R Rembarz and S Sories ldquoA capacity analysis forthe transmission of event- und cooperative awareness messagesin LTE networksrdquo in ITS World Congress 2011 Orlando USAOrlando Florida USA 2011

[13] A Grzybek G Danoy and P Bouvry ldquoGeneration of realistictraces for vehicular mobility simulationsrdquo in Proceedings of the2nd ACM International Symposium on Design and Analysis ofIntelligent Vehicular Networks and Applications DIVANet 2012pp 131ndash138 New York NY USA October 2012

[14] T Cerqueira and M Albano ldquoRoutesmobilitymodel easy real-istic mobility simulation using external information servicesrdquoin Proceedings of the 2015 Workshop on Ns-3 ser WNS3 rsquo15 pp40ndash46 New York NY USA 2015

[15] S Kato M Hiltunen K Joshi and R Schlichting ldquoEnablingvehicular safety applications over LTE networksrdquo in Proceedingsof the 2013 2nd IEEE International Conference on ConnectedVehicles and Expo ICCVE 2013 pp 747ndash752 December 2013

[16] ldquoPublicwarning system (PWS) requirements (release 13)rdquo 3GPPTS 22268 V1300 Dec 2015

[17] ldquoOverall description Stage 2 (Release 13)rdquo 3GPP TS 36300V1320 Dec 2015

[18] G Remy S-M Senouci F Jan and Y Gourhant ldquoLTE4V2XLTE for a centralized VANET organizationrdquo in Proceedings ofthe 54th Annual IEEE Global Telecommunications ConferenceldquoEnergizing Global Communicationsrdquo GLOBECOM 2011 pp 1ndash6 Houston TX USA December 2011

[19] Y Yang P Wang C Wang and F Liu ldquoAn eMBMS basedcongestion control scheme in cellular-VANET heterogeneousnetworksrdquo in Proceedings of the 17th IEEE International Con-ference on Intelligent Transportation Systems (ITSC rsquo14) pp 1ndash5IEEE Qingdao China October 2014

[20] S Taha and X Shen ldquoA physical-layer location privacy-preserving scheme for mobile public hotspots in NEMO-basedVANETsrdquo IEEE Transactions on Intelligent TransportationSystems vol 14 no 4 pp 1665ndash1680 2013

[21] G T V1290 Evolved Universal Terrestrial Radio Access (E-UTRA) Medium Access Control (MAC) protocol specification(Release 12) 3GPP 2016

[22] A Bazzi B M Masini and A Zanella ldquoPerformance analysisof V2v beaconing using LTE in direct mode with full duplexradiosrdquo IEEEWireless Communications Letters vol 4 no 6 pp685ndash688 2015

[23] D Tsolkas E Liotou N Passas and L Merakos LTE-a AccessCore and Protocol Architecture for D2D Communication SMumtaz and J Rodriguez Eds Springer International Publish-ing 2014

[24] Y Bi H Shan X S Shen N Wang and H Zhao ldquoA multi-hop broadcast protocol for emergency message disseminationin urban vehicular ad hoc networksrdquo IEEE Transactions onIntelligent Transportation Systems vol 17 no 3 pp 736ndash7502016

[25] ldquoVehicle safety communications project task 3 final reportrdquoTech Rep US Department of Transportation 2005

[26] Intelligent transport systems (ITS) basic set of applications Part2specification of cooperative awareness basic service ETSI TS 102637-2 V121 2011

[27] Intelligent transport systems (ITS) basic set of applications part3specifications of decentralized environmental notification basicservice ETSI TS 102 637-3 V111 2010

[28] C L Robinson D Caveney L Caminiti G Baliga K La-berteaux and P R Kumar ldquoEfficient message compositionand coding for cooperative vehicular safety applicationsrdquo IEEETransactions on Vehicular Technology vol 56 no 6 I pp 3244ndash3255 2007

[29] D Caveney ldquoCooperative vehicular safety applications col-lision avoidance enabled through geospatial positioning andintervehicular communicationsrdquo IEEE Control Systems Maga-zine vol 30 no 4 pp 38ndash53 2010

[30] W Sun T Fu Y Su F Xia and J Ma ldquoODAM-C an improvedalgorithm for vehicle Ad Hoc networkrdquo in Proceedings of the2011 IEEE International Conference on Internet ofThings iThings2011 and 4th IEEE International Conference on Cyber Physicaland Social Computing CPSCom 2011 pp 152ndash156 October 2011

[31] S Ansari T Boutaleb C Gamio S Sinanovic I Krikidis andM Sanchez ldquoVehicular Safety Application Identifier algorithmfor LTE VANET serverrdquo in Proceedings of the 8th InternationalCongress on Ultra Modern Telecommunications and ControlSystems andWorkshops ICUMT 2016 pp 37ndash42 October 2016

[32] Intelligent transport systems (ITS) communications architectureETSI EN 302 665 V111 2010

[33] S Sinanovic H Burchardt H Haas and G Auer ldquoSum rateincrease via variable interference protectionrdquo IEEETransactionson Mobile Computing vol 11 no 12 pp 2121ndash2132 2012

[34] E-UTRA Base Station (BS) radio transmission and reception(Release 12) 3GPP TS 36104 V12100 2016

[35] ldquoEvolved Universal Terrestrial Radio Access Network (E-UTRAN) Stage 2 functional specification of User Equipment(UE) positioning in E-UTRAN (Release 9)rdquo

[36] G T V1310 Evolved Universal Terrestrial Radio Access (E-UTRA) Medium Access Control (MAC) protocol specification(Release 13) 3GPP 2016

[37] ldquoModel library release ns-32rdquo Ns-3 network simulator 2015httpswwwnsnamorgdocsmodelshtmllte-designhtml

[38] G Piro N Baldo and M Miozzo ldquoAn LTE module for thens-3 network simulatorrdquo in Proceedings of the 4th InternationalICST Conference on Simulation Tools and Techniques 422 serSIMUTools rsquo11 ICST 415 pages March 2011

[39] L Chunjian Efficient antenna patterns for three-sectorWCDMAsystem Master of Science Thesis Chalmers University of Tech-nology Goteborg Sweden 2003

[40] Evolved universal terrestrial radio access (E-UTRA) physicallayer procedures 3GPP TS 36213 V1301 2016

[41] D Kimura and H Seki ldquoInter-cell interference coordination(ICIC) technologyrdquo Fujitsu Scientific amp Technical Journal vol48 no 1 pp 89ndash94 2012

[42] Physical layer measurements (release 13) 3GPP TS 36214V1300 2015

[43] Evolved universal terrestrial radio access (E-UTRA) radioresource control (RRC) protocol specification 3GPP TS 36213V1300 2015

[44] Y S Cho W Y Yang and C Kang IMO-OFDM WirelessCommunications with MATLAB John Wiley amp Sons 2010

[45] J Archer and K VogelThe traffic safety problem in urban areas2000

[46] ldquoUrban crashesrdquo Insurance Institute for Highway Safety 1999

Wireless Communications and Mobile Computing 17

[47] J Klaue B Rathke and A Wolisz EvalVid ndash A Framework forVideo Transmission and Quality Evaluation P Kemper and WH Sanders Eds Springer Berlin Germany 2003

[48] S Ucar S C Ergen and O Ozkasap ldquoMulti-hop clusterbased IEEE 80211p and LTE hybrid architecture for VANETsafety message disseminationrdquo IEEE Transactions on VehicularTechnology vol PP no 99 pp 1-1 2015

[49] L Ward and M Simon ldquoIntelligent transportation systemsusing IEEE80211prdquoRohde and Schwarz -ApplicationNote 2015

International Journal of

AerospaceEngineeringHindawiwwwhindawicom Volume 2018

RoboticsJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Active and Passive Electronic Components

VLSI Design

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Shock and Vibration

Hindawiwwwhindawicom Volume 2018

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawiwwwhindawicom

Volume 2018

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018

Control Scienceand Engineering

Journal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom

Journal ofEngineeringVolume 2018

SensorsJournal of

Hindawiwwwhindawicom Volume 2018

International Journal of

RotatingMachinery

Hindawiwwwhindawicom Volume 2018

Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Navigation and Observation

International Journal of

Hindawi

wwwhindawicom Volume 2018

Advances in

Multimedia

Submit your manuscripts atwwwhindawicom

International Journal of

AerospaceEngineeringHindawiwwwhindawicom Volume 2018

RoboticsJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Active and Passive Electronic Components

VLSI Design

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Shock and Vibration

Hindawiwwwhindawicom Volume 2018

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawiwwwhindawicom

Volume 2018

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018

Control Scienceand Engineering

Journal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom

Journal ofEngineeringVolume 2018

SensorsJournal of

Hindawiwwwhindawicom Volume 2018

International Journal of

RotatingMachinery

Hindawiwwwhindawicom Volume 2018

Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Navigation and Observation

International Journal of

Hindawi

wwwhindawicom Volume 2018

Advances in

Multimedia

Submit your manuscripts atwwwhindawicom