application of vehicular communications for improving … avdvisory_418... · cles display lower...

WIRELESS COMMUNICATIONS AND MOBILE COMPUTINGWirel. Commun. Mob. Comput. 2011; 11:1657–1667

Published online 23 November 2011 in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/wcm.1233

SPECIAL ISSUE PAPER

Application of vehicular communications forimproving the efficiency of traffic in urban areasKonstantinos Katsaros1*, Ralf Kernchen1, Mehrdad Dianati1, David Rieck2 andCharalambos Zinoviou1

1 Centre for Communications Systems Research, Department of Electronic Engineering, University of Surrey Guildford, SurreyGU2 7XH, U.K.

2 Fraunhofer Institute for Open Communication Systems, Kaiserin-Augusta-Allee 31, 10589 Berlin, Germany

ABSTRACT

This paper studies the impacts of vehicular communications on efficiency of traffic in urban areas. We consider a GreenLight Optimized Speed Advisory application implementation in a typical reference area and present the results of its per-formance analysis using an integrated cooperative intelligent transportation systems simulation platform. In addition, westudy route alternation using vehicle-to-infrastructure and vehicle-to-vehicle communications. Our interest was to monitorthe impacts of these applications on fuel and traffic efficiency by introducing metrics for average fuel consumption, averagestop time behind a traffic light and average trip time, respectively. For gathering the results, we implemented two trafficscenarios defining routes through an urban area including traffic lights. The simulations are varied for different penetrationrates of application-equipped vehicles, driver’s compliance to the advised speed and traffic density. Our results indicate thatGreen Light Optimized Speed Advisory systems could improve fuel consumption, reduce traffic congestion in junctionsand the total trip time. Copyright © 2011 John Wiley & Sons, Ltd.

KEYWORDS

vehicular communications; traffic light advisory; route advisory; fuel consumption; traffic congestion

*Correspondence

Konstantinos Katsaros, Centre for Communications Systems Research, Department of Electronic Engineering, University of SurreyGuildford, Surrey GU2 7XH, U.K.E-mail: [email protected]

1. INTRODUCTION

Advances in wireless communications and in particularvehicular communications have led to the advent of coop-erative intelligent transportation systems [1,2]. These sys-tems employ vehicular communication technologies suchas IEEE 802.11p [3] to enable deployment of applica-tions that could potentially improve road safety and trafficefficiency and introduce new entertainment and businessapplications [4]. Exploitation of intelligent transportationsystems for traffic congestion control in urban areas aswell as fuel consumption reduction are among the mostpromising applications according to transportation author-ities [5,6]. This can be achieved by vehicle-to-vehicle(V2V) or infrastructure-to-vehicle communications, intel-ligently advising individual drivers about traffic events,such as the traffic light phases and congestion.

In this paper, we study the impacts of application ofvehicular communications on improving the efficiency of

traffic in urban areas. In particular, we focus on two spe-cific applications: Green Light Optimal Speed Advisory(GLOSA) and Adaptive Route Change (ARC), respectively.

First, we design and implement a GLOSA system toreduce traffic congestion by decreasing the average stoptime behind traffic lights and total trip time while reduc-ing fuel consumption and CO2 emissions. The potentialof V2V communication for improving fuel efficiency hasbeen demonstrated in [7], showing that vehicular commu-nications can assist to reduce average fuel consumptionespecially under high traffic density and long traffic lightcycles. Other projects have investigated the impacts on fuelefficiency when using wireless communications betweenvehicles (V2V) or vehicle to infrastructure employing dif-ferent algorithms to smoothly slow down at a red trafficlight or to reach it at the next green phase. Dependingon the used consumption models, different results can beobserved. One important aspect, as introduced in [8], is thedependency of the results on different penetration rates of

Copyright © 2011 John Wiley & Sons, Ltd. 1657

Efficiency of traffic in urban areas K. Katsaros et al.

the communication enabled vehicles. In the same work aswell as in [9], the effect of traffic density is studied. In[8], it is also suggested to either cut the fuel delivery orstop the engine when the vehicle stops at the traffic lightfor long time to achieve less fuel consumption. In [10],the intelligent driver car-following model [11] is used tocontrol a platoon of 10 vehicles, where only the leadingone is equipped with communication capabilities, achiev-ing 30% fuel savings. In [12], when the algorithm is usedonly with one vehicle and one traffic light, fuel savingsreached 20%. Although these results seem trustworthy, thesimulations are conducted with a small number of vehiclesthat do not consider the dynamics of the vehicular environ-ment. When the model in [12] is scaled up including 15traffic light junctions, the results are reduced to 6% reduc-tion in fuel consumption. The optimal activation distancefor the algorithm is also investigated in this work and isfound that a distance of 500 m achieves the best results intheir simulations. Furthermore, in [10] and [9], traffic effi-ciency is examined using different performance metrics. In[10], the increase of average speed is taken as indicatorof traffic efficiency where in [9], the flow and the ratio ofmotionless vehicles are considered. In our implementation,the GLOSA application provides the advantage of timelyand accurate information about traffic lights cycles andtraffic lights position information through infrastructure-to-vehicle communication. It provides drivers with speedadvice guiding them with a more constant speed and withless stopping time through traffic lights. Our results showup to 7% reduction in average fuel consumption and upto 89% in average stop time and 9.85% in trip time whenGLOSA is used in a reference scenario. While conduct-ing the simulations, we found that the optimal GLOSAactivation distance is around 300 m.

Second, we implement an ARC application, whichuses both vehicle-to-infrastructure and vehicle-to-vehicle(V2X) communications to divert traffic away of congestedroads. The avoidance of congested areas by re-routing traf-fic away of these areas was proposed as a use case in [13].Vehicles disseminate messages with information regardingthe traffic conditions around them, for other vehicles toanalyze and make a decision whether a change of routeis needed or not. An advantage that this algorithm pos-sesses is that vehicles also know about the congestion onroads nearby. This aims to prevent vehicles reverting toa route that is also congested. The main metric that thealgorithm uses to implement this feature is the speed ofthe vehicles in the area. The average speed of each vehi-cle passing a road segment is sent to the other vehicles. Onthe receiving edge, vehicles calculate the “edge weights”for these road segments; on the basis of these weights,they calculate the optimal route. Two simulation scenarioswere studied: an urban scenario and a highway scenario.For the purposes of this paper, we discuss solely aboutthe former scenario. Vehicles travel on a predefined route,which becomes congested near traffic lights. Their simu-lation results suggest that depending on application pene-tration rate, the algorithm benefits all vehicles, even those

with no communication support. This happens becauseV2V equipped vehicles take alternative routes, unloadingthe roads that were in their past route. Therefore, all vehi-cles display lower travel times. As for the CO2 emissionsand fuel consumption, they are also proportionally affectedby the number of equipped vehicles. Additional factors likelonger alternative routes reduce the positive effects. Butemissions and fuel consumption are lower compared with100% conventional vehicle scenarios. Finally, the infor-mation exchange that takes place eventually balances thetraffic flow within greater part of the road network. Anotherstudy of such an application is presented in [14]. Resultsshowed that high V2X penetration rate leads to more vehi-cles using alternative routes. About 50% of the total traveltime was reduced for all vehicles at a V2X penetrationrate of 80% or higher, whereas V2X-equipped vehiclesbenefit from much lower penetration rates. Another note-worthy observation is that on penetration rate equal to80%, all vehicles benefitted equally. In our implemen-tation, vehicles inform traffic lights about their positionand status. Traffic lights assess the traffic congestion interms of the number of stopped vehicles using informa-tion gathered from vehicles. When a certain thresholdof congestion is passed, they geo-broadcast this conges-tion information using V2V communication towards pre-vious junctions so the vehicles can adaptively change theirroute. Regarding ARC effects on traffic and fuel efficiency,we observed reduction of 26.7% in stop time and 18%in total trip time. The consumption is also reduced upto 15.86%.

In the aforementioned works, different communicationtechnologies are adopted. The authors in [7,9,12,15] donot discuss in depth the communication mechanisms oftheir simulations. They assume successful disseminationof the messages. Others, such as [10] and [16], use gen-eral purpose wireless communication technologies, suchas wireless sensor networks and IEEE 802.11[17]. Lately,the IEEE 802.11 standard is preferred because of opera-tional cost. The highly dynamic network topology of vehic-ular communications has led to the introduction of IEEE802.11p [3] that is more suitable for such applications andis the one we used in our simulations.

The main challenges in the implementation of the previ-ous applications include the modeling of the vehicle traffic,the communications between traffic lights and vehicles andfinally the driver’s behavior. Individual research has beenperformed for each one of these areas, but complete simu-lations by taking into account the dynamics of all param-eters are scarce. Thus, for our implementation of GLOSAand ARC, we used an integrated simulation tool based onthe Fraunhofer VSimRTI [18], which enables online two-way coupling of different simulators for monitoring theinfluence of GLOSA and ARC application on the traffic andfuel consumption.

The contributions of this paper are as follows:

� Design and implementation of two V2X applica-tions (GLOSA and ARC), which aim at increasing

1658 Wirel. Commun. Mob. Comput. 2011; 11:1657–1667 © 2011 John Wiley & Sons, Ltd.DOI: 10.1002/wcm

K. Katsaros et al. Efficiency of traffic in urban areas

fuel and traffic efficiency, using the FraunhoferVSimRTI platform.

� Extensive simulation experiments using different sce-narios to find the optimal point of GLOSA activation,the influence of application penetration rate on fueland traffic efficiency and the effect of traffic densityon GLOSA performance.

� Design and implementation of a stochastic drivermodel to investigate the influence of the driver’scompliance to the advisory speed.

� Extensive simulation experiments using different sce-narios to find the influence of ARC application pene-tration rate on fuel and traffic efficiency and the effectof traffic density on ARC performance.

The rest of this paper is organized as follows. InSection 2, we present our integrated simulation approach,and in Section 3, we give an overview of the integratedsimulation platform. The design of the GLOSA algorithmis thoroughly discussed in Section 4. In Section 5, the ARCalgorithm is described. In Section 6, the simulation set-upis presented followed by our evaluation results. Finally, inSection 7, we conclude and provide ideas for future work.

2. SYSTEM MODEL

We designed and tested GLOSA and ARC in an integratedsimulation platform. The first step was to define a refer-ence scenario that is depicted in Figure 1. In this referencearea, we placed a traffic light. Vehicles will enter followinga common arrival process such as Poisson distribution. Inthe scenario, the equipped vehicles rate is varied because

we want to investigate the influence of the penetration rateand which is the minimum percentage of equipped vehi-cles for GLOSA and ARC to have a positive impact onthe traffic efficiency and fuel consumption. Road traffic ismodeled using the microscopic Stefan Krauss (SK) model[19], a car-following model with two basic rules. First,vehicles in free motion have a target speed and try to cruiseat it. Second, when a vehicle senses the distance to thevehicle ahead to be less than a certain threshold, it slowsdown keeping a safe distance. The speed of the vehicles iswithin a certain range [Vmin, Vmax] where Vmin is the mini-mum speed that vehicles can cruise without causing furthertraffic congestion and Vmax is the maximum speed that isforced by the speed limit of the area. Acceleration is alsobounded and asymmetric – higher deceleration than accel-eration – for more realistic simulations. The SK model isintegrated in Simulation of Urban Mobility (SUMO) [20]traffic simulator that we used in our work.

We have set up two simulation scenarios for each appli-cation to compare the results. In the first scenario (S0),there is no driver’s assistance information (advisory speedor alternative route), and the traffic is governed by the SKmodel. In the second scenario (S1), we provide informa-tion through the GLOSA or ARC messages, and the driverreceives speed advisory messages or an alternative route.We assume that all drivers who get an advise will follow it.We control the percentage of GLOSA and ARC equippedvehicles to monitor the impact of penetration rate. There-fore, we have defined three performance metrics that wecheck in all scenarios to derive our conclusions. The firstmetric (P1) is the average stop time of vehicles waitingat the intersection behind red lights measuring the traffic

Figure 1. Reference Scenario for Green Light Optimal Speed Advisory.

Wirel. Commun. Mob. Comput. 2011; 11:1657–1667 © 2011 John Wiley & Sons, Ltd.DOI: 10.1002/wcm

1659


efficiency of GLOSA and ARC application. We assume thatthere is no other reason for the vehicles to halt apart fromstopping at a traffic light or a queue caused by the trafficlight (we do not simulate brake downs or accidents in ourscenarios). The second metric (P2) is the average trip timeof vehicles from the moment they enter the reference areauntil the moment they exit. The third metric (P3) is thefuel consumption derived from the fuel consumption andemission model in [21] measuring the fuel efficiency of theapplications. The emissions of CO2 are estimated from thesame model to be proportional to fuel consumption.

3. INTEGRATED SIMULATIONPLATFORM

Simulating the GLOSA and ARC use cases poses a chal-lenge in terms of combining and synchronizing differentsimulation aspects, for example, vehicular traffic, networkcommunication and application handling. For these chal-lenges to be addressed, the integrated simulation platformVSimRTI [18] was used to simulate the use cases.

VSimRTI borrows some concepts of the High LevelArchitecture [22] to enable the coupling of the mostappropriate simulators for a scenario. It is not only alightweight framework interaction between them and life-cycle management of simulators that is responsible fortasks such as synchronization of simulators but also itsapplications. Using VSimRTI enables access to all relevanttraffic objects such as road-side units (RSUs), application-equipped vehicles or traffic lights through common inter-faces regardless of the specific simulator instance, forexample Vissim or SUMO.

For the applications to be simulated, the integrated appli-cation simulator VSimRTI_app was used. It provides acouple of simple JAVA interfaces to create V2X appli-cations, while offering access to all relevant simulationdata such as vehicle status or communication modules.V2X messages sent to specific vehicles are forwarded tothe associated application, and application output can be

directed to a specific vehicle (e.g., giving speed advisoryto a simulated vehicle). Specialized for V2X simulations,VSimRTI and its simulators have been built with currentV2X technologies in mind; that is, all simulations havebeen performed using standardized V2X protocols such asIEEE 802.11p.

Figure 2 shows the general concept of VSimRTI. Eachsimulator is coupled to the runtime infrastructure (RTI)by implementing generic interfaces to communicate to theRTI (VSimRTI Ambassador) or to receive messages fromthe RTI (Federate Ambassador).

4. GREEN LIGHT OPTIMAL SPEEDADVISORY ALGORITHM

The GLOSA algorithm has been implemented to supportthe aforementioned simulation approach and is presentedin Algorithm 1. First, vehicles enter the communication

Federate 1

V2X Simulation Runtime Infrastructure

VSimRTIAmbassador

Time ManagementFederation Management

R

FederateAmbassador

R

R

Simulator 1

Federate X

VSimRTIAmbassador

FederateAmbassador

R

Simulator X

RR

Interaction Management

Simulation Runner

R

Figure 2. VSimRTI system.


zizi

Highlight


range of a traffic light according to the Poisson distribu-tion as mentioned before. The RSU attached to a trafficlight broadcasts periodically cooperative awareness mes-sages including the position, timing information and addi-tional data for the traffic light. When the on-board unitreceives a cooperative awareness message, the algorithmchecks if its source is a traffic light or not. From the posi-tion information within the message and the vehicle’s ownposition and heading, it calculates whether this traffic lightis relevant (on its route) or not (line 1). The application canthen calculate the distance from the traffic light and withthe current speed and acceleration, the time that it wouldtake to reach it (time-to-traffic-light TTL) (line 2). Next, itchecks the traffic light phase at that time (TTL) (line 3). Ifthe traffic light is green when the vehicle reaches it, thenthe vehicle continues its trip trying to reach the maximumspeed limit of the road (lines 4–6). If it is red, it calcu-lates the speed that it should have to reach it in the nextgreen phase (lines 7–9). If it is yellow, depending on theremaining yellow time and the acceleration capabilities ofthe vehicle, it could advice to accelerate or decelerate againwithin the permitted range (lines 10–13). Finally, the drivergets an advise with the speed limited within the permit-ted range [Vmin, Vmax] (line 15). This algorithm runs everysecond, which makes it more robust against external inter-ference, such as other vehicles, that do not follow the sameadvisory speed or are non-equipped.

The algorithm has as input the current speed U0, accel-eration a of the vehicle and the distance to the traffic lightDtl . With the utilization of basic rules of motion, givenby (1),

d D ut C1

2at2 (1)

where d is the distance, u is the initial speed, t is the timeand a is the acceleration; the time to reach the traffic light(TTL) can be calculated as shown in (2).

TTL D

8̂̂̂<ˆ̂̂:

d

uwhen aD 0

�u

aC

su2

aC2d

awhen a¤ 0

(2)

The target speed (Ut ) for the red and yellow light phasesis calculated using (3) after rearranging (1) and settingaD .Ut �U0/=t

Ut D2d

t�U0 (3)

where d is the distance to traffic light (Dtl ), t is the timeto reach the traffic TTL light plus the remaining time forthe next green phase (Tred or Tyellow C Tred, respectively)and U0 is the current speed. The algorithm can be visual-ized with the help of Figure 3. A vehicle without GLOSAaccelerates until it reaches the maximum allowed speedand then suddenly has to break because of the upcoming

0 20 40 60 80 100 1200

2

4

6

8

10

12

14

16

Trip Time (sec)

Veh

icle

Spe

ed (

m/s

)

without GLOSAwith GLOSA

Trip Time Reduction

Figure 3. Vehicle’s speed over time with and without GreenLight Optimal Speed Advisory.

red light. The worst-case scenario is that it has to stop forthe complete duration of traffic light’s red phase (25 s inour example). On the other hand, using GLOSA, a vehicleobtains information about the phases and adjusts accord-ingly its speed so that it reaches the traffic light at themoment it turns green again. Thus, it does not have tocome into a complete halt. Even though the advised speedis lower than the road limit, the vehicle’s average speedand thus trip time is not increased. On the contrary, it isdecreased because of the fact that when the vehicle passesthe traffic light, it has an initial speed greater than zero.

5. ADAPTIVE ROUTE CHANGEALGORITHM

In this section, we describe the ARC algorithm. Vehiclesenter the reference area following the Poisson distributionas before. When they stop at a traffic light, they broadcasta message with information about their position, the roadthey are and their ID. Upon receipt of such a message,a traffic light updates its local database with the stoppedvehicle on that road. When a vehicle departs from a traf-fic light, it sends a second message so that the traffic lightcan keep track of the number of stopped vehicles; thus, thecongestion level on each road. When the congestion levelpasses a certain threshold, the traffic light geo-broadcasts amessage towards the previous junction where the decisionabout an alternative route may be made. The protocol thatis used for geo-broadcasting is Cached Greedy Geocast(CGGC) [23]. This message includes information relatedto the road that the congestion has occurred. However,when the traffic is again lowered, it sends a second messageso that vehicles can switch back their initial route.

6. SIMULATION SET-UP ANDEVALUATION RESULTS

For the evaluation of the GLOSA and ARC application, aseries of simulations were conducted. The configuration of


1661


the environment for the project consists of the SUMO [20]traffic simulator used to produce and cope with the vehicletraffic, the JiST/SWANS [24] used for the communicationsand finally the application simulator described in Section 3that runs the GLOSA and ARC application written in Java.We simulated two road networks. The first underlying sim-ulation scenario is a road network section of Guildfordtown center in the UK as depicted in Figure 4. It consists ofone route where vehicles start from point A on York Roadaccording to a Poisson distribution and travel until point Bon Waterden Road on one lane and without overtaking. Thenumber of vehicles is defined to 100 to gather sufficientdata in terms of time and number of independent vehiclesto obtain statistically accurate results. The travel distanceis approximately 0.6 miles (0.965 km). Within this route,there are two traffic lights (TL1 and TL2). For these twotraffic lights, the timing regarding the previous route is 20–4–6 (green–yellow–red) and 20–4–36 s, respectively. Thedifference in red time for TL2 is because of the LondonRoad’s green phase duration. The second road network isa 2� 2 grid with four traffic lights as depicted in Figure 5.There are multiple routes for this scenario. The basic routefollows intersections A, B and D. The alternative one trav-els through intersections A, C and D, respectively to avoidcongestion at B and D. A third route follows intersectionsC and D only and works as a control for the congestion.The speed limit on the road is set to 15 m/s (54 km/h),

Figure 4. Simulation scenario map number 1.

Figure 5. Simulation scenario map number 2.

which is near the usual limit in an urban area. The mini-mum advisory speed is set to 6 m/s (21.6 km/h) in ordernot to travel too slow and cause more congestion. The sim-ulation runs until every vehicle has left the simulation area.The communication range of the vehicles and traffic lightsis set near 500 m and use IEEE 802.11p communication asaccess mechanism to broadcast their messages.

6.1. Green Light Optimal SpeedAdvisory evaluation

The simulations can be divided into four categories. First,we tested the influence of the activation distance forGLOSA on the overall performance. In these simulations,all vehicles are equipped with the GLOSA application.Also, to check the integrity of the algorithm, we excludedthe first traffic light and run only with one (TL2). The tim-ing for this was also altered to have equal red and greenphases. The results of the two performance metrics can beseen in Figure 6. An optimal point of activation is found ata distance near 350 m. At shorter activation distances, thereaction time (time required for the driver to slow down tothe advised velocity) is not enough to have benefits. Thefuel consumption is also slightly increased because of thefact that the average trip time is increased (vehicles areadvised to travel at lower speeds). At longer activation dis-tances, the benefits regarding fuel consumption are slightlydecreased but remain near the optimal levels. The resultsfor two traffic lights with a distance near 400 m betweenthem (Figure 4) shift this optimal point to a shorter dis-tance of 250 m, which will be the value used to producethe next set of results. This is due to the fact that vehi-cles do not have enough time to accelerate and reach ahigher velocity after the effect the first traffic light has ontheir velocity before they run the GLOSA algorithm onceagain for the second traffic light. Hence, further simula-tions have to be made for larger scale scenario and moretraffic lights to conclude which activation distance to use.Having a shorter activation distance means that we canreduce the transmission power of the RSU and thus having

Figure 6. Influence of activation distance on Green LightOptimal Speed Advisory performance.



better resource allocation and less collisions in the commu-nications. Compared with the work in [12], where the min-imum activation distance is found near 500 m, our workshows better characteristics in this aspect.

0 20 40 60 80 1002

4

6

8

10

12

14

GLOSA Penetration Rate (%)

Ave

rage

Sto

p T

ime

(sec

)

ALL Vehicleswith GLOSAwithout GLOSA

Figure 7. Influence of Green Light Optimal Speed Advisorypenetration rate on average stop time.

0 20 40 60 80 1009.2

9.4

9.6

9.8

10

10.2

10.4

10.6


Ave

rage

Fue

l Con

sum

ptio

n (L

/100

km)

ALL Vehicles

with GLOSA

without GLOSA

Figure 8. Influence of Green Light Optimal Speed Advisorypenetration rate on average fuel consumption.

0 20 40 60 80 10090

92

94

96

98

100

102


Ave

rage

Tri

p T

ime

(sec

)

Figure 9. Influence of Green Light Optimal Speed Advisorypenetration rate on average trip time.

Second, we measured the influence that GLOSA pen-etration rate has on the three performance metrics andhow the non-equipped vehicles are affected. The simula-tions were conducted in a high traffic density environment(Poisson expected value �D 0:2). From [8], we learn thatan increase in penetration rates of equipped vehicles allowsfor a better reduction of fuel consumption in the overalltraffic scenarios. As it can be seen from Figures 7–9, this isverified not only for fuel efficiency but also for traffic effi-ciency. The most interesting outcome from these figures isthat even the non-equipped vehicles are getting affected ina beneficial way from the GLOSA equipped vehicles andthis is due to the SK model. They follow the leading vehi-cle that, if equipped with GLOSA, forces them to adjusttheir speeds accordingly because we assume that there areno overtaking in our simulations. The second notice is thatthe average stop time is reduced even when the penetrationrate is small; but to see positive results in fuel efficiency, weneed at least 50% equipped vehicles. The observed averagemaximum reduction in fuel consumption is 7%, which isslightly higher than the average maximum fuel savings in[12] for their scaled up scenario. Finally, the trip time isreduced by 9.85%. The sharp decline after 70% is due tothe reduction in stop time; vehicles stop for maximum onetraffic light cycle.

Third, we evaluated the impact of the driver’s com-pliance to the advisory speed on the performance of theGLOSA algorithm. Driver’s behavior is simulated using arandom value uniformly distributed, which indicates thecompliance of the driver to the advised speed. This isdescribed by (4). It has to be noted that the driver’s speed isagain constrained by the road’s limit and minimum advisedspeed as described previously. We simulated a scenariowhere all vehicles are equipped with the GLOSA applica-tion, and we vary the maximum deviation from the advi-sory speed. The results presented in Figure 10 indicate thatif the driver does not follow the exact advisory speed, thenthe fuel consumption is potentially increased. Accordingto our simulations, similar results are observed for average

0 5 10 15 209.6

9.62

9.64

9.66

9.68

9.7

9.72

9.74

9.76

9.78

Ave

rage

Fue

l Con

sum

ptio

n (L

/100

km)

Average Increase in advised Speed (%)

Figure 10. Influence of driver’s compliance to the advisedspeed on fuel consumption.


1663


0.08 0.1 0.12 0.14 0.16 0.18 0.2 0.220

10

20

30

40

50

60

70

80

90

100

Traffic Density − Expected value of the Poisson Distribution

Gai

n (%

)Stop TimeFuel Consumption

Figure 11. Influence of vehicle traffic density on the GLOSAperformance

stop time and trip time.

Udriver D .1C a/Ut (4)

where Udriver is the speed that the driver will follow, Utis the advised speed calculated from the GLOSA algorithmand a is a uniformly random number that takes values fromthe range of [0, 0.4], namely an average increase of 20%from the optimal advisory.

Finally, we simulated different traffic densities (high,medium and low) to capture the influence they have onthe overall performance of the GLOSA application. Theresults shown in Figure 11 suggest that the higher the traf-fic density (moving from left to right in the plot), the morebenefits we have regarding fuel efficiency reaching a max-imum of 7% fuel reduction. On the other hand, the benefitswe get regarding traffic efficiency are decreased, which wasalso reported in [9]. This is because the vehicles are morescarcely distributed; therefore, they do not influence eachother, they all follow precisely the advisory speed and thereare smaller queues at the traffic lights making the GLOSAalgorithm work better.

6.2. Adaptive Route Change evaluation

For the evaluation of ARC, we varied the penetration rateof ARC-equipped vehicles. However, those vehicles thatare not ARC equipped have wireless module and act asrelays for the geo-broadcast messages. As presented inFigures 12 and 13, the average trip time is reduced by asmuch as 26.5% and the average stop time up to 32.5% for150 vehicles in the simulation, respectively. In addition,fuel consumption is also reduced by 17.3% as depicted inFigure 14. The minimum values are observed not at 100%penetration but at 70–80%. The reason behind this is thefact that for 100% penetration, all vehicles receive thechange route message; thus, all of them change their routeand cause congestion on the other route. However, whenless vehicles are equipped, the non-equipped ones willcontinue their initial route, which has less vehicles now.

0 20 40 60 80 100350

400

450

500

550

600

ARC Penetration Rate (%)

Ave

rage

Tri

p T

ime

(sec

)

50 vehicles100 vehicles150 vehicles

Figure 12. Influence of Adaptive Route Change penetration rateon average trip time.

0 20 40 60 80 10050

60

70

80

90

100

110


Ave

rage

Sto

p T

ime

(sec

)50 vehicles100 vehicles150 vehicles

Figure 13. Influence of Adaptive Route Change penetration rateon average stop time.

0 20 40 60 80 1008.5

9

9.5

10

10.5

11

11.5

12

12.5


Ave

rage

Fue

l Con

sum

ptio

n (L

/100

km)

50 vehicles

100 vehicles

150 vehicles

Figure 14. Influence of Adaptive Route Change penetration rateon average fuel consumption.



One possible amendment to the algorithm would be a pro-portional selection of the new route, so as not all vehiclesto change their route.

Moreover, the vehicles make an instant estimation ofthe local queue size, calculating the number of stoppedcars in front of them, without using further communica-tion with the traffic light using (5). They measure the dis-tance from the closest traffic light (Dtl ); by subtractingthe distance of the first stopped car from the traffic light(D0 � 2 m, approximated using simulations), they divideit by the length of vehicles (L D 5 m, constant for oursimulations) to find the local queue size (Q). We simulatedtwo scenarios with different number of vehicles running,and the results presented in Figures 15 and 16 suggest thatthe maximum and average queue sizes are decreased asARC penetration is increased. Again the maximum is notobserved at 100% as explained before.

QDDtl �D0

L(5)

0 20 40 60 80 10020

30

40

50

60

70

80


Max

imum

Que

ue S

ize

(# v

ehic

les) 50 vehicles

100 vehicles150 vehicles

Figure 15. Influence of Adaptive Route Change penetration rateon maximum queue size.

0 20 40 60 80 1006

8

10

12

14

16

18

20

22

24


Ave

rage

Que

ue S

ize

(# v

ehic

les)

50 vehicles100 vehicles150 vehicles

Figure 16. Influence of Adaptive Route Change penetration rateon average queue size.

7. CONCLUSIONS AND FUTUREWORK

The results suggest that both GLOSA and ARC applica-tions have a positive effect on all performance metrics.The higher the GLOSA penetration rate is, the more ben-efits we have with a maximum of 80% reduction in stoptime, 9.85% in trip time and up to 7% reduction in fuelconsumption in a high traffic density scenario. There is acritical point of 50% of equipped vehicles where the effectof GLOSA starts to be more visible on fuel consumptionand of 70% where the trip time is sharply decreased. Asthe density decreases, the benefits for fuel efficiency arereduced, but there is still improvement compared with non-equipped vehicles. The traffic efficiency on the other handis increased with the decrease in traffic density reaching89%. There is also an optimal activation distance wherethe GLOSA application should advise the driver; this isnear 300 m from the traffic lights, but it depends slightlyon the road network. Closer to this distance, the time toreact is limited; further away, there are no more benefits.If the complexity of the algorithm is to be increased, thedistance could be also increased. Our simulation results forthe ARC application suggest increase in both traffic andfuel efficiencies. We observe an 18% reduction in averagetrip time and 26.7% reduction in average stop time. Fuelconsumption is reduced by 15.86%. The work presented bythis paper is an example of what can be achieved in termsof fuel and traffic efficiencies when vehicles are enabled tocommunicate with traffic lights and how we can exploit anintegrated simulation platform to achieve this.

There are various ways in which these applications couldbe extended to achieve more accurate results. First of all,for GLOSA we assumed that there are no vehicles waitingat the traffic light, which is not always the case. There-fore, the distance to traffic light could be replaced by thedistance to the end of the queue instead to achieve morereasonable results. The simulation network should alsobe extended to a larger scale scenario using real data forvehicle input. Finally, having results from field tests wouldprovide data to compare field tests and simulations toevaluate the estimations made by the simulations.

ACKNOWLEDGEMENT

The work was carried out within the joint research projectDRIVE C2X, which is funded by DG Infso of the EuropeanCommission within the 7th Framework Program.

REFERENCES

1. Figueiredo L, Jesus I, Machado J, Ferreira J,Martins de Carvalho J. Towards the developmentof intelligent transportation systems. IEEE Intelli-gent Transportation Systems 2001: 1206–1211. DOI:10.1109/ITSC.2001.948835.


1665

zizi

Highlight


2. Hartenstein H, Laberteaux KP. A tutorial surveyon vehicular ad hoc networks. IEEE Communica-tions Magazine 2008; 46(6): 164–171. DOI: 10.1109/MCOM.2008.4539481.

3. Institute of Electrical and Electronics Engineers. IEEEStandard for Information Technology – WirelessLAN Medium Access Control (MAC) and Phys-ical Layer (PHY) Specifications Amendment 6:Wireless Access in Vehicular Environments 2010.DOI: 10.1109/IEEESTD.2010.5514475.

4. CAR 2 CAR Communication Consortium Manifesto.Online document at: www.car-2-car.org.

5. National traffic signal report card, 2007. Online docu-ment at: http://www.ite.org/REPORTCARD.

6. Traffic congestion and urban mobility, 2010. Onlinedocument at: http://tti.tamu.edu/infofor/media/topics/congestion_mobility.htm.

7. Widodo A, Hasegawa T, Tsugawa S. Vehicle fuelconsumption and emission estimation in environment-adaptive driving with or without inter-vehicle com-munications, In IEEE Intelligent Vehicles Symposium,2000; 382–386, DOI: 10.1109/IVS.2000.898373.

8. Wegener A, Hellbruck H, Wewetzer C, Lubke A.VANET simulation environment with feedback loopand its application to traffic light assistance, In IEEEGLOBECOM Workshops, 2008; 1–7, DOI: 10.1109/GLOCOMW.2008.ECP.67.

9. Chao-Qun M, Hai-Jun H, Tie-Qia T. Improving urbantraffic by velocity guidance, In International Con-ference on Intelligent Computation Technology andAutomation, 2008; 383–387, DOI: 10.1109/ICICTA.2008.288.

10. Sanchez M, Cano JC, Kim D. Predicting traffic lightsto improve urban traffic fuel consumption, In Interna-tional Conference on ITS Telecommunications, 2006;331–336, DOI: 10.1109/ITST.2006.288906.

11. Treiber M, Hennecke A, Helbing D. Congested traf-fic states in empirical observations and microscopicsimulations. Physical Review E 2000; 62(2, Part A):1805–1824. DOI: 10.1103/PhysRevE.62.1805.

12. Tielert T, Killat M, Hartenstein H, Luz R,Hausberger S, Benz T. The impct of traffic-light-to-vehicle communication on fuel consumption andemmisions, In Internet of Things 2010 – SecondInternational Conference for Academia and Industry,2010, DOI: 10.1109/IOT.2010.5678454.

13. PRE-DRIVE C2X Deliverable D2.3 – descriptionof communication, traffic and environmental modelsand their integration and validation. Technical Report,June 2010.

14. Wedel JW, Schünemann B, Radusch I. V2X-based traffic congestion recognition and avoidance,In International Symposium on Pervasive Systems,Algorithms, and Networks, 2009; 637–641, DOI: 10.1109/I-SPAN.2009.71.

15. Asadi B, Vahidi A. Predictive cruise control: utilizingupcoming traffic signal information for improving fueleconomy and reducing trip time. IEEE Transactions onControl Systems Technology 2010; 19: 707–714. DOI:10.1109/TCST.2010.2047860.

16. Iglesias I, Isasi L, Larburu M, Martinez V,Molinete B. I2V communication driving assis-tance system: on-board traffic light assistant, InVehicular Technology Conference, 2008; 1–5, DOI:10.1109/VETECF.2008.437.

17. Institute of Electrical and Electronics Engineers. IEEEStandard for Information Technology – Wireless LANMedium Access Control (MAC) and Physical Layer(PHY) Specifications: High-Speed Physical Layer inthe 5 GHz Band 1999. DOI: 10.1109.IEEESTD.1999.90606.

18. Queck T, Schünemann B, Radusch I, Meinel C.Realistic simulation of V2X communication sce-narios, In IEEE Asia-Pacific Services ComputingConference, 2008; 1623–1627, DOI: 10.1109/APSCC.2008.23.

19. Krauss S. Microscopic modeling of traffic flow: inves-tigation of collision free vehicle dynamics, PhDThesis, Mathematisches Institut, Universität zu Köln,1998.

20. Krajzewicz D, Hertkorn G, Rssel C, Wagner P. SUMO(Simulation of Urban MObility) – an open-source traf-fic simulation, In Middle East Symposium on Simula-tion and Modelling, 2002; 183–187.

21. Biggs DC, Akcelik R. An energy-related model ofinstantaneous fuel consumption. Traffic engineering &control 1986; 27: 320–325.

22. Institute of Electrical and Electronics Engineers. IEEEStandard for Modeling and Simulation (M&S) HighLevel Architecture (HLA) – Federate Interface Speci-fication. IEEE Standard 1516.1 2000. DOI: 10.1109/IEEESTD.2001.92421.

23. Maihoefer C, Eberhardt R, Schoch E. CGGC:cached greedy geocast, In International Confer-ence Wired/Wireless Internet Communications, 2004;13–25, DOI: 10.1.1.94.9237.

24. Barr R, Haas ZJ, van Renesse R. JiST: an efficientapproach to simulation using virtual machines. Soft-ware: Practice and Experience 2005; 35: 539–576.DOI: 10.1002/spe.647.



AUTHORS’ BIOGRAPHIES

Konstantinos Katsaros Konstanti-nos Katsaros received his diplomain Electrical and Computer Engineer-ing from the University of Patras(Greece) in 2009 and his MSc inCommunications, Networks and Soft-ware from the University of Surrey(UK) in 2010. Currently, he is pursu-ing his PhD in the Centre for Commu-

nication Systems Research at the Department of ElectronicEngineering of the University of Surrey in the UK. Hehas been involved in PRE-DRIVE C2X and DRIVE C2XFP7 projects. His research interests include vehicular com-munications, intelligent transportation systems and mobilead hoc networks.

Ralf Kernchen Ralf received hisDipl. Inf. (equal to MSc in Com-puter Science) from the TechnicalUniversity of Berlin in 2004. Heworked as a research fellow in theCentre for Communication SystemsResearch (CCSR) at the Universityof Surrey (Guildford, UK) from 2004to 2011. He completed his PhD in

Mobile Communications in 2010 at CCSR. Ralf is aco-founder at Rulemotion Ltd, working on executable busi-ness specifications for information systems.

Mehrdad Dianati Mehrdad Dianatireceived his BSc and MSc degreesin Electrical Engineering from SharifUniversity of Technology (Tehran,Iran) in 1992 and from K. N. ToosiUniversity of Technology (Tehran,Iran) in 1995, respectively. He workedas a hardware/software developer and

technical manager from 1992 to 2002. From 2002 to 2006,he completed his PhD in Electrical and Computer Engi-neering at the University of Waterloo (Ontario, Canada).Mehrdad is currently a Lecturer (Assistant Professor) inthe Centre for Communication Systems Research at theDepartment of Electronic Engineering of the University ofSurrey in the UK.

David Rieck David Rieck is aresearch assistant at the FraunhoferInstitute for Open CommunicationSystems. The main focus of theresearch work of David Rieck isthe development of overall simu-lation architecture for the simula-tion of Vehicle-2-X Applications,the Vehicle-2X Simulation Runtime

Infrastructure.

Charalambos Zinoviou Charalam-bos Zinoviou is a graduate of SoftwareEngineer at Comptel Corporation. Hereceived his masters in Communica-tions Networks and Software fromthe University of Surrey in 2011.He obtained his 5-year PolytechnicDiploma in Information and Commu-nication Systems Engineering from

the University of the Aegean in Greece in 2010. Duringhis industrial placement in 2008, he was involved in Cus-tomer Technical Support at Cyprus TelecommunicationsAuthority. In 2004, he completed his 25-month militaryservice with Cyprus National Guard, where he was enlistedto the Special Forces Unit (Commando) and graduated asa 2nd Lieutenant.


1667

application of vehicular communications for improving … avdvisory_418... · cles display lower...

Documents