The Coordinated Vehicle Recovery Mechanismin City Environments

Gwo-Jiun Horng1

Published online: 13 April 2016# Springer Science+Business Media New York 2016

Abstract In this paper, an innovative, optimized and coordi-nated vehicle recovery network system is designed and ana-lyzed. This system safeguards private property for all citizens,significantly enhances an existing vehicle recovery and anti-theft system, and improves the success rate for vehicle recov-ery. This study is not only applicable to a single vehicle butalso considers the entire vehicle anti-theft communication net-work and connects all individual vehicles to an innova-tive, optimized and coordinated vehicle recovery network.Algorithms are investigated and integrated to create a highlyaccurate, fast, safe, and efficient innovative system that is ca-pable of locating targets in any environment, regardless of thelength of the communication distance and whether the com-munication comprises indoor or outdoor communication with-in a small area. The current study evaluates the performance ofthe approach by conducting computer simulations. The simu-lation results reveal the strengths of the proposed optimizedand coordinated vehicle recovery system in terms of a reducedsearch time, an increased success rate of tracking and an im-proved success rate for vehicle recovery in VANETs.

Keywords Vehicle recovery . Anti-theft network . VANETs

1 Introduction

In countries with advanced intelligent transportation systems(ITSs), such as the United States, Canada, Japan and the

European Union (EU) countries, development trends in recentyears have gradually shifted from independent developmenttoward integration. The overall development direction is toleverage an integrated vehicle infrastructure system to achievethe goals of energy saving, carbon dioxide (CO2) emissionsreduction, transport system safety improvement, and trafficcongestion control. Typical projects include vehicle infra-structure integration (VII) in the United States, Smartway inJapan, and cooperative vehicle infrastructure systems (CVISs)in the EU. Therefore, an investigation of the progress of VIIintegration system development in developed countries andthe feasibility of the domestic development of VII systems isa major research topic that can provide support for transpor-tation departments in establishing or adjusting policies andstrategies for future ITS development.

ITS architecture provides a framework for the much neededoverhaul of highway transportation infrastructure. The archi-tecture’s immediate effects include alleviating vehicle trafficcongestion and improving operational management to pro-mote public safety by such means as collision-avoidance im-provements. Equipping vehicles with various on-board sen-sors and implementing vehicle-to-vehicle (V2V) communica-tion will allow for large-scale sensing and decision-makingand will help to control actions in support of these goals [1].

Countries that are involved in promoting and developingITS are expected to leverage the momentum of ITS technolo-gy to transform transportation into a positive force that cancreate healthy and sustainable global economies and achievethe goals of expanding motorized power, improving transpor-tation safety, relieving congestion, sustaining economicgrowth and creating sustainable environments. The deploy-ment of ITS can achieve comprehensive and accurate vehiclerecognition, real-time dynamic road condition monitoring,road control, inspection of stolen vehicles [2], and trackingof runaway vehicles in accidents. From a service perspective,

* Gwo-Jiun Hornggrojium@gmail.com

1 Department of Computer Science and Information Engineering,Southern Taiwan University of Science and Technology,Tainan, Taiwan

Mobile Netw Appl (2016) 21:656–667DOI 10.1007/s11036-016-0722-8

communication devices, such as mobile phones, can beemployed to publish road condition information via multipleplatforms in a timely manner, guide and divert motor vehicles,and effectively relieve traffic congestion.

Many developed countries have invested a significantamount of human resources and capital to conduct tests usinglarge-scale Internet of Vehicles technology [3]. The Instituteof Electronics and Electrical Engineers (IEEE) published802.11p as the technology for wireless access in vehicularenvironments (WAVE). Based on the Internet of Things, theInternet of Vehicles will become another symbol of futureintelligent cities. 802.11p is a communication protocol thatis extended from the IEEE 802.11 standard, which is primarilyemployed in dedicated short-range communication in ITS.802.11p supports the data exchange between high-speed ve-hicles and between vehicles and the ITS roadside infrastruc-ture. 802.11p has multiple enhancements for the unique ve-hicular environment; it also employs more advancedswitchover mechanisms and mobile operations, as well asenhanced safety [4–6], identification, and point-to-pointauthentication.

Vehicular information and communication technologyintegrates communication and information. Vehicular in-formation and communication technology has introducedfunctions and services, such as video and audio enter-tainment inside vehicles, global positioning system(GPS) positioning and navigation, sightseeing spotsearching, parking lot and gas station guides, and 3Gwireless communication. Next-generation vehicular infor-mation and communication systems will focus on dedi-cated short-range communication (DSRC) between vehi-cles, between vehicles and roadside units, and [7, 8]between vehicles and public infrastructure to enhancevarious services, such as driving safety alarms, dynamicreal-time navigation, energy savings, carbon reductionand vehicle anti-theft. WAVE/DSRC is expected to be-come a major trend and to provide fast-moving vehicleswith multi-channel operation architecture to enable ve-hicular and roadside devices to establish vehicle-to-vehicle (V2V) and vehicle-to-roadside (V2R) short-range communication in high-speed driving environ-ments. When this system connects to existing communi-cation systems [2, 9, 10], such as Wi-Fi, WiMAX, and3.5 G, features such as dynamic navigation, real-timeroad condition notification and driving safety alarmscan be achieved.

Avehicle will become Bthe 3rd intelligent space^ after thehome and workspace. In the future, people, vehicles, androads will integrate with terminal devices, service centers,service facilities and roadside devices in the environment viaheterogeneous networks to form an interconnected network.Thus, scattered information can be integrated to facilitate theinteraction between vehicle users, vehicles and the

surrounding environment. The use of inter-vehicle communi-cation technology will create new experiences for drivers andpassengers that generate feelings of comfort, convenience,cleanness, energy efficiency, time efficiency and safety.

Vehicles are a natural theft target because they are expen-sive and can be easily disposed of and driven away. The firstrecorded vehicle theft case in history occurred in 1896, tenyears after the invention of the vehicle. Vehicle theft caseshave continued to increase since that time. Statistical datashow that a car is stolen in the United States every 20 secondson average.

Because vehicles have become increasingly popular world-wide, vehicle safety management has gained significant at-tention. Vehicle anti-theft systems have also gained popularity.The anti-theft systems that are currently available in the mar-ket can be divided into two categories: a Global System forMobile Communications (GSM) notification system and asatellite positioning and tracking system. An effective vehiclerecovery and protection system can locate a car in real-time,immediately notify the car owner in the event of an anomalyand obtain the vehicle’s current location and condition.Because the system can locate a vehicle in real-time, it maylead to the arrest of the thief. In recent years, vehicle anti-theftand vehicle recovery systems have integrated GSM/GeneralPacket Radio Service (GPRS) wireless communication withGPS satellite positioning to notify the vehicle owner.However, the scarcity of wireless communication base sta-tions, attenuated signals and insufficient GPS positioninginformation in a small area cause notification failure andhinder locating the vehicle within a short period of time.

This paper contributes to the research literature in four ma-jor areas: 1) using a cyber-physical alarm system to form thevehicle recovery network; 2) increasing the success rate oftracking; 3) reducing the search time for a stolen car; and 4)reducing the packet loss ratio in city environment zones toachieve these goals. In addition to system simulation and anal-ysis, we also implement a modularized prototype of a vehicleroad integration positioning system.

The rest of this paper has the following organization.Section II discusses related research on the VII problem.Section III describes the systemmodel and the VDTNs forthe vehicle recovery system in Section IV. Section V dis-cusses the simulation. We present evaluations of ex-periments and implement the model in Section VI. SectionVII concludes this paper and identifies future researchdirections.

2 Related work

This section reviews important attempts at applying VII com-munication in VANETs. Cloud services and intelligent vehi-cles method are described.

Mobile Netw Appl (2016) 21:656–667 657

2.1 VII system

A VII system focuses on real-time communication betweenBpeople, vehicle and road^ systems. To accommodate differ-ent environments, proper wireless transmission media areemployed for transportation system integration. During theinitial stage of ITS development, terms such as Bvehicle^,Broad^, mean ordinary car and street represent vehicles inland, water and air transport systems and their transport routes.BPeople^ can obtain intelligence and guidance aboutBvehicles^ and Broads^ via information communication orother communication products to improve transportation safe-ty and efficiency and achieve goals of energy saving and car-bon reduction. [11] AVII system is applied research that le-verages a series of advanced technologies to enable directcommunication between vehicles and surrounding auxiliaryfacilities. The original and primary goal is to improve roadsafety. The basic premise of a VII system is to provide a directlink between a moving vehicle and all adjacent vehicles.These vehicles can communicate with each other, exchangeinformation about speed, direction, driving consciousness andintension. Therefore, it can improve safety between vehiclesand simultaneously improve system sensitivities of automaticemergency mechanisms (steering, decelerating, and braking).This system also facilitates communication with road infra-structure, which enables real-time traffic information of entireroad networks to be complete and provide beneficialfeedback.

2.2 Intelligent vehicle and cloud service

Vehicle technology can be extended from basic communica-tion functions between two vehicles. When a user in a vehiclewants to communicate with a user in another vehicle, a highbandwidth is required for real-time video or voice transmis-sion. When the bandwidth is sufficient and all vehicles con-nect to a cloud for communication, communication betweentwo vehicles will become more diversified. In addition to pro-viding information about approaching vehicles, vehicle tech-nology can also be employed by a driver to communicate withother drivers and can directly replace the mobile phone be-cause a driver can communicate with any person online.

2.3 Delay tolerant network (DTN)

In addition to the Internet, numerous communication net-works, including mobile devices with limited electric powerand satellites, are gradually evolving [12, 13]. Their commu-nication technologies need to consider the unique requirementof the communication environment. These communicationnetworks cannot use the transmission control protocol(TCP)/Internet protocol (IP) and cannot communicate witheach other. They can only use a dedicated protocol for

communication within their own system. Each of these com-munication networks constitutes a region, and every connec-tion inside a region exhibits similar communicationcharacteristics.

The boundary of different regions can be determined byconnection delay, connection connectivity, relative data speed,error rate, addressing, reliability, and transmission quality. Incontrast with the Internet, these wireless networks exhibit thefollowing characteristics: long and varying delay, interruptedtransmission, high data error rate, and significantly differentrelative data transmission rates. In addition to the Internet,wireless networks include interstate road networks, combatunit networks in battle fields and space networks. A proxymechanism is required to connect networks in different re-gions, which can accommodate the characteristics of differentnetworks and serve the role of buffering between two net-works with significantly different transmission speeds.

In a normal arbitrary network [14–17], even if node move-ment will result in broken routes, most nodes are interconnect-ed; however, due to some environmental factors and unstabletopology, packet transmission usually experiences some de-lay. This random network is referred to as a delay network. Adelay-tolerant network exhibits characteristics such as highdelay, high error rate, and frequent network blackouts, whichis significantly different from a conventional Internet environ-ment. Several vehicular network architectures are proposedissues, such as DTN [18–20], vehicle delay-tolerant network(VDTN) [21, 22], cooperative DTN [23].

3 System model

In this paper, an innovative, optimized and coordinated vehi-cle recovery network system is designed and analyzed. Thissystem safeguards private property for all citizens, significant-ly enhances an existing vehicle recovery and anti-theft system,and improves the success rate for vehicle recovery. This studyis not only applicable to a single vehicle but also considers theentire vehicle anti-theft communication network and connectsall individual vehicles to an innovative, optimized and coor-dinated vehicle recovery network. Algorithms are investigatedand integrated to create a highly accurate, fast, safe, and effi-cient innovative system that is capable of locating targets inany environment, regardless of the length of the communica-tion distance and whether the communication comprises in-door or outdoor communication within a small area.

In this system, a Wi-Fi base station is equipped with ahotspot network long-distance (HNLD) wireless system.When Wi-Fi base station installation becomes more popular,this system’s service range will be greatly expanded. Avehicleanti-theft communication network that was previously inde-pendent can be securely integrated. We explain how to use aHNLD wireless system to send an alarm message for a stolen

658 Mobile Netw Appl (2016) 21:656–667

vehicle. A black box with a HNLDwireless system is installedin a vehicle.

We apply the concept of a HNLD wireless system and usethis method to design a vehicle black box. The idea is todesign a seamless network that expands from small-scale orindoor communication to large-scale or outdoor communica-tion and is suitable for long-distance communication. Thissystem also exhibits wireless characteristics with superior pen-etration and is suitable for all types of indoor communication.In addition, it is the largest wireless radio signal-based freenetwork community and extends small-scale movable multi-angulation positioning to a large-scale vehicle recovery andanti-theft system.

Because a GPS satellite positioning system cannot receivesufficient data for an alley or indoor area, the signal strengthwill be significantly attenuated in an indoor environment. Ifthe receiver sensitivity enhancement is insufficient, a solutionmust be obtained to address this problem. The mesh networkconcept and HNLD wireless system can be employed for lo-cation measurement and calculation without the need for ad-ditional hardware and network infrastructure. Measurementbased on a HNLD wireless system in conjunction with satel-lite technology can realize fast and complete region position-ing that extends beyond movable, multi-angulation position-ing in a small area.

We plan to employ an algorithm as the movable multi-angulation positioning mechanism, which integrates triangula-tion and a gossip algorithm. It can provide effective locationinformation in a small area for indoor or outdoor environmentsand perform continuous tracking. It can also immediately attainthe accuracy of a small range in various environments, includ-ing urban and indoor areas. Therefore, a user can gain indooraccess to a locating service with reliable operation and fast datacollection. This system not only ensures the availability of fastand accurate location information but also effectively reducesthe cost and power of a terminal device because an expensivenetwork infrastructure update is prevented.

This study employs a movable multi-angulation positionmethod to calculate the accurate location. In a conventionalobject positioning method, relevant data are collected for asingle object to initiate the positioning calculation, that is,the object that can be located needs to continuously provideits ownmessages. Some messages may not be accurate, whichwill significantly reduce the positioning accuracy. In a typicalliving environment, some messages are broadcasted or thecollection method may be fast and encompass a large area.The proverb Bgood news stays indoors while bad news haswings^ can be interpreted as Bthe gossip message’s propaga-tion and influence are very impressive.^ We can employ thismethod as the data collection method for the positioning.Therefore, this positioning method will use the triangulationconcept in conjunction with gossip data collection forpositioning.

In this study, we will explain the theory that serves asthe basis of the triangulation. It uses a fixed point ormobile information exchange process for positioning.Regarding the information exchange, we plan to exchangeinformation via the gossip data collection method. Whenvehicles approach each other, they will exchange recentlycollected data, consolidate the data and make decisions.For example, if a vehicle at a crossroad receives a theftalarm message from a stolen vehicle, this message will betransmitted via a gossip transmission network. Over time,all people will receive this information and the police andthe owner of the stolen vehicle will be informed to enablehandling of the case.

We propose an innovative design based on the conceptof free and sharing. As shown in Fig. 1, a black box isinstalled in a vehicle as a hotspot network long-distancewireless signal transmission device. When a vehicle alarmmechanism is triggered and a vehicle is driven by an in-truder, the vehicle owner will receive audio and visualnotification via a handheld receiver that can successfullyreceive information in any environment. At this moment,the vehicle owner can press an emergency button to senda Car ID signal for vehicle locating. We must search thereceived information to locate the stolen vehicle or obtaininformation regarding its destination. The stolen car’s CarID will be simultaneously transmitted by adjacent vehi-cles. If this system is installed in vehicles nationwide, thevehicles can be connected to form a meshed wirelesscommunication network. When an optimized and coor-dinated vehicle recovery network system becomesmore robust and has a large installation base, vehiclerecovery and protection systems in Taiwan will becomemore comprehensive and will significantly decrease thevehicle theft rate.

We denote each vehicle as CK={C1,C2,…,Ck}, each vehi-cle include the license plate number. The CarID comprises thefollowing categories of information: the driver’s gender di ∈dmale, dfemale; vehicle number vID; and vehicle type cvehicle ∈brand of vehicle. The CarID function is assumed to beUnicode and the corresponding signal is modeled asCarID = vdriver[di, cvehicle] + vID, where CarID(di) denotesdriver’s gender; CarID(cvehicle) denotes brand of vehicle; andCarID (vID[n]) denotes vehicle ID number.

As shown in Fig. 2, cities contain many vehicles; thesevehicles may be equipped with a communication device forV2V communication and hotspot communication between avehicle and a stationary building.

When a hotspot device receives a transmitted theftalarm message, it will transmit this message to a cyber-physical alarm system (CPAS) server center. When thesystem receives this message, it will separate the theftalarm into two parts: the message will be calculatedand analyzed via the proposed gossip mechanism and

Mobile Netw Appl (2016) 21:656–667 659

adaptive goal programming and the vehicle position datawill be processed by a maximum likelihood (ML) meth-od to estimate the most likely location and the locationdata will be constantly updated. In addition, the generat-ed data will be sent to a police vehicle and the owner ofthe stolen vehicle.

When the vehicle alarm mechanism is triggered and thevehicle is driven by an intruder, the searcher of the stolenvehicle (alarm message receiver) will locate the vehicle basedon the theft alarm message. At this moment, the owner of thestolen vehicle search for the vehicle using the alarm messagereceiver and numerous vehicles equipped with these devicesalso receive theft alarm messages. Thus, the vehicle ownerwill receive audio and visual notification via a handheld re-ceiver and successfully receives information in anyenvironment.

As shown in Fig. 3, the stolen vehicle will automati-cally send the stolen vehicle’s Car ID, location and direc-tion to adjacent vehicles and stationary hotspot devices.

The message received by mobile devices will be sent tonearby stationary hotspot devices. When a hotspot devicereceives a message (Carstolen, location, and direction), itwill send the message to the CPAS, which will calculateand update Carstolen (location and direction). The calcu-lated data will be transmitted to a police vehicle via sta-tionary hotspot devices, and a message will be constantlyupdated to intercept the stolen vehicle.

We present a connected network of n cars with a con-nectivity graph G= (V,E). This network graph G changesover time in terms of V and E. Each car, i ∈ V, has itsown information denoted by xi. The information xi canbe thought of as a collection of bits, a packet or even areal number. The interest is in spreading or disseminatingthis information to possibly different subsets of nodes inV. The car wishes to compute a function fi(xi,…,xn) usinga gossip mechanism. Under a gossip mechanism, the op-eration of any car, must satisfy the following properties:1) The mechanism should only utilize information

Fig. 1 System networkenvironment

Fig. 2 coordinated vehicle recovery network system model

660 Mobile Netw Appl (2016) 21:656–667

obtained from its neighbors; 2) The mechanism does notrequire synchronization between car i and its neighbors.

Because information is propagated, some informationwill not be transmitted to locations near the stolen vehi-cle. A vehicular system is a mobile system; thus, anadaptive and goal-optimized decision rule should beemployed to calculate and analyze the closest and mostreliable information and to notify the driver and police.This study uses Adaptive Goal Programming to solvean optimization problem. Goal programming is a type oflinear programming that was proposed in 1961 byCharnes et al. Goal programming can solve a goal opti-mization problem and problems with conflicting goals.Goal programming can consider user strategy and knowninformation, decide priority based on each goal’s rela-tive importance, or assign different weights within thesame priority to help evaluate the likelihood to achievepredefined goals. It can also generate concrete numbersto facilitate subsequent comparison, selection and analy-sis. Conventional linear programming only seeks themaximum or minimum value of the objective functionunder constraint. When the maximization or minimiza-tion condition is satisfied, the optimization goal isachieved. Goal programming focuses on the outcome ofthe system operation or resource utilization, plans theoutcome of an individual target for given conditions,and determines the minimum gap between the goal andthe outcome.

We use priority-based goal programming in systemdesign. Goal importance has different priority levels,and each goal has a different level of importance, thatis, the priority of each goal is in hierarchical order. Thus,goals are sorted according to priority and then solved inthis order. The solution for a goal with higher priority isused as the constraint for a goal with lower priority. Forinstance, when the first priority goal has been solved, theobtained solution is employed as the constraint for the

second priority goal; then the second goal is solved. Inparallel goal programming, no priority difference is ob-served between goals and the goal item with greater im-portance will be assigned a higher weight. The generalformula is as follows:

Minz ¼X K

k¼1w−k d

−k þ

X K

k¼1wþk d

þk

s:t:ð1Þ

Goal constraint ∑j=1n ckjxj+dk

−−dk+=gi, ∀ kIn the formula, ∑j=1

n ckjxj represents the attributedk−−dk+ represents the deviationalvariablegi represents the target

Function constraint ∑j=1n aijxj≤ (=,≥)bi,∀ i

Non-negative constraintni; pi≥0 ;∀i x j≥0 ;∀ j

Complementary relation ni×pi=0 , ∀ i

The design in this study uses priority-based goal program-ming, in which the goal priority is assigned by the decisionmaker. Goals at each stage are optimized. The optimum solu-tion for goals with higher priority will not be affected andcompromised by goals with lower priority. Therefore, thissystem’s objective formula is as follows:

Min Z ¼ ni; pif gs:t:X

j¼1

n ai jx j� �þ ni−pi ¼ bi ∀ i ¼ 1; 2;⋯;m j ¼ 1; 2;⋯; n

ni � pi ¼ 0 ; ∀ i ¼ 1; 2;⋯;mni; pi≥0 ; ∀ i ¼ 1; 2;⋯;mxj≥0; ∀ j ¼ 1; 2;⋯; n

ð2Þ

In the formula, xj: each model’s decision variable; bi: eachmodel’s objective; aij regression coefficient; ni: negativedeviational variable when the (i)th objective is underachieved;pi: positive deviational variable when the (i)th objective isexceeded; ni×pi=0: positive and negative deviations cannotco-exist

Triangulation is the most commonly used algorithm in thepositioning system; its principle is to use three anchor pointswith known location information to determine the location ofa test point. If the distances from the test point to the anchorpoints and the coordinates of the anchor points are known, thetest point’s coordinates can be calculated. The target that needsto be located employs distance measurement technology (e.g.,radio signal strength, RSS) to estimate the distance from itselfto each reference point. If the distances to three or more ref-erence points are known, the location can be estimated, asshown in Fig. 4a. For instance, a GPS employs this methodfor positioning.

Figure 4a shows the case of an ideal situation; an actualsituation will be more complex. In most circumstances, three

CarstolenMatch

?

CPAS

CarK

PoliceCarK

Hotspot

(Building)

catch

Loop:

check

Tx Carstolen (I ama stolen car) Update: Tx

Carstolen(location,

direction)

ResponseUpdate:

Carstolen (location, direction)

computing

Tx Carstolen (location,direction)

Tx Carstolen(location,

direction)

Fig. 3 Diagram of the state transition system

Mobile Netw Appl (2016) 21:656–667 661

circles will not intersect at a single point; thus, the condition ofthree concurrent circles cannot be used to determine the finalpositioning point, as shown in Fig. 4b. Therefore, the maxi-mum likelihood method is typically used to estimate a targetobject’s location. rA, rB, and rC .represent the distances esti-mated from the reference points A, B, C, respectively, and theminimum σxy is used to determine the target object’s location.

σx;y ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffix−xAð Þ2 þ y−yAð Þ2−rA

q��������þ

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffix−xBð Þ2 þ y−yBð Þ2−rB

q��������

þffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffix−xCð Þ2 þ y−yCð Þ2−rC

q��������

ð3Þ

If an object that needs to be located is continuously mov-ing, then three or more fixed reference points must be con-stantly identified to estimate the object’s location. This pro-cess will significantly affect positioning accuracy and speed.In this paper, each fixed point is assumed to be equipped witha hotspot network long-distance wireless referral work stationand every mobile point (e.g., vehicle and people) will beequipped with a hotspot network long-distance wireless signaldevice. Vehicles will exchange data with hotspot networklong-distance wireless signal base stations in buildings andadjacent vehicles.

In addition to the use of three fixed points (hotspot networklong-distance wireless signal base station in buildings) as ref-erence, information from mobile points (vehicle or peopleequipped with hotspot network long distance wireless signaldevice) can also be used as a reference. Using this concept, weuse the principle of triangulation and significantly increase thenumber of reference points to obtain more abundant informa-tion for reference, which improves the positioning accuracyand speed.

Suppose there is a sample x1, x2,…,xn of n independentand identically distributed observations coming from adistribution with an unknown probability density functionf0(•). It is, however, surmised that the function f0 belongsto a certain family of distributions {f0(•|θ), θ∈Θ}, where θis a vector of parameters, called the parametric model, sothat f0 = f (•|θ). The value θ0 is unknown and is referred toas the true value of the parameter vector. It is desirable tofind an estimator θ that is as close to the true value θ0 as

possible. Either or both of the observed variables xi andthe parameter θ can be vectors.

To use the method of maximum likelihood, one first spec-ifies the joint density function for all observations. For anindependent and identically distributed sample, this joint den-sity function is

f x1; x2;…; xnjθð Þ ¼ f x1jθð Þ � f x2jθð Þ �… f xnjθð Þ ð4Þ

The maximum likelihood estimator coincides with themost probable Bayesian estimator given a uniform prior dis-tribution of the parameters. The maximum a posteriori esti-mate is the parameter θ that maximizes the probability of θgiven the data, given by Bayes' theorem as:

P θjx1; x2;…; xnð Þ ¼ f x1; x2;…; xnjθð ÞP θð ÞP x1; x2;…; xnð Þ ð5Þ

where P(θ) is the prior distribution for the parameter θ, andP(x1, x2,…, xn) is the probability of the data averaged over allparameters.

Because the denominator is independent of θ, the Bayesianestimator is obtained by maximizing P(x1, x2,…, xn | θ )P(θ)with respect to θ. If we further assume that the prior P(θ) is auniform distribution, the Bayesian estimator is obtained bymaximizing the likelihood function f(x1, x2,…, xn | θ )P(θ).Thus, the Bayesian estimator coincides with the maximum-likelihood estimator for a uniform prior distribution P(θ).

For the normal distribution N(μ, σ2), which has probabilitydensity function

f x���u;σ2

� �¼ 1ffiffiffiffiffiffi

2πp

σe − x−uð Þ2

2σ2

� �ð6Þ

u ¼ x ¼X n

i¼1xi=n

where x is the sample mean.Based on the definitions, the likelihood function and the

maximum likelihood estimator of μ, the mean weight of alllocations, can be identified. Using the given sample, the max-imum likelihood estimate of μ can also be determined. Next,we investigate how to improve the vehicle recovery efficiencyusing the relevant network transmission method.

4 VDTNs for vehicle recovery system

We know that hotspot network long-distance wireless signalcoverage will affect a vehicle recovery system’s success rate.An optimized and coordinated vehicle recovery network sys-tem’s hotspot network long-distance wireless signal systemsare fixed manual installations. However, signal coverage ishighly dependent on the installation location. Due to con-straints such as installation permission, cost and landform, a

(a) (b)

Fig. 4 (a) Schematic diagram of trangulation; (b) Schematic diagram oftrangulation deviation

662 Mobile Netw Appl (2016) 21:656–667

server may not be installed in the originally planned area,which will produce incomplete signal coverage.

When a stolen vehicle travels out of an area that is coveredby a hotspot network long-distance wireless server signal, thebroadcasted alarm information cannot be received and trans-mitted. To address the problem of low vehicle recovery suc-cess rate, we use a VDTNs communication protocol, which isa very popular technology in the area of ultra-long distancenetwork transport, as the solution.

Therefore, the characteristics of a VDTN are employed,and ordinary passing by vehicles will function as cruisers thatcollect the stolen vehicle’s wireless signal and Car ID.When acruiser has a GPS self-positioning signal, two vehicles canexchange information with each other. At this moment, thecruiser will search the access point that can report to the serverand send this discovered signal to the vehicle owner or policestation.

A VDTN is a network environment that is insensitive totime delay, environments of mobile devices with limited elec-tric power, satellite communication, and incomplete networkcoverage. These networks have the following characteristics:long and varying delay, transmission interrupt, high data errorrate, and significantly different relative data transmissionrates, which are situations addressed by a DTN.

In a conventional arbitrary wireless network, therouting protocol assumes that the entire network is con-nected. Even if node mobility will create or terminatesome connections, most nodes are connected. Networkconnection is intermittent; the connection between nodeswill change over time; and new connections will be cre-ated or existing connections will be terminated. In thesesituations, a node typically has a queue for temporarypacket storage; when a new connection emerges, it cancontinue packet transmission. When a network connec-tion is unstable, each packet transmission will experiencea long delay; this vehicular arbitrary wireless networkcan be referred to as a VDTN.

To improve the performance of a VDTN and improve therouting success rate, a buffer is added to a node to temporarilystore unsent packets; when a connection condition improvesin the future, transmission can be continued. The main goal ofa VDTN is to provide a comprehensive bi-directional commu-nication method in heterogeneous networks and to provide anacceptable network performance in the environment of packetloss, excessive delay, error, and temporary connectiondisruption.

Next, we describe a vehicle recovery system’s VDTN and ahotspot network device’s long-distance wireless signal. Signalconnection disruption is caused by the relatively long distancebetween a hotspot network server and a stolen vehicle. Thus,we implement the delayed packet transmission in a vehiclewith a hotspot network transmitter and convert it to a trans-mission relay station between a stolen vehicle and a hotspot

network server. Any vehicle can receive a message from astolen vehicle’s hotspot network device and transfer it to ahotspot network server. Its operational mode and functionare as follows:

When the theft condition is satisfied, a vehicle periodicallysends a stolen vehicle message (SVM) via a hotspot networkdevice, as shown in Fig. 5-(1). The stolen vehicle will besuspended, which will send a Beacon packet to announceinformation, such as subsequent encapsulation action and op-eration schedule, hibernate duration. Then, it will send anSVM, receive server ACK information (SA) from the hotspotnetwork device, trigger sensing and scaring (SS) action in thestolen vehicle, and remain in a hibernation state to save elec-tric power (Inactive). Because a stolen vehicle is in a station-ary state, a thief has limited options or may already haveescaped from the vehicle. Therefore, actions do not need tobe executed in every cycle. If the stolen vehicle is moving, thethief should be inside the vehicle; therefore, SS actions shouldbe continuously executed and monitored in the stolen vehicle,as shown in Fig. 5-(2).

When the stolen vehicle does not receive an ACK re-sponse, the vehicle is not in the hotspot network long-distance wireless signal server’s communication range andthe stolen vehicle’s transmission strategy will switch to DTNmode. As shown in Fig. 5-(3), (4), a stolen vehicle will specifyDTN mode in the Beacon packet, vehicles that pass and re-ceive this Beacon packet will pass the ACK message to thestolen vehicle and transmit the stolen message to the hotspotnetwork long-distance wireless signal server. The differencebetween diagrams 5-(3) and 5-(4) is that the stolen vehicle is ina different state: stationary versus moving.

However, if an ordinary vehicle needs to receive an SVM,the SVM transmission period (SV period) should not be ex-cessive. Whether an ordinary vehicle can receive a Beaconpacket when it travels through a hotspot network long-distance wireless transmission area should be considered.Therefore, the message transmission period and durationshould be determined by factors such as its driving speed,ordinary vehicle’s estimated speed, and hotspot networklong-distance wireless transmission range, which are includedin the following formulas:

SV Period ¼ Tr

Vnv þ VSVð7Þ

where Tr is the hotspot network long-distance wireless trans-mission range (diameter); Vnv is the upper limit for the vehiclespeed on a common road; and Vsv is the speed of the stolenvehicle. We assume that an ordinary vehicle’s driving direc-tion is in the opposite lane, and deduce that the maximumrelative speed of an ordinary vehicle versus stolen vehicle is|Vnv+Vsv|. The calculated SV Period is used as the messagetransmission period.

Mobile Netw Appl (2016) 21:656–667 663

When an ordinary vehicle receives the DTN mode stolenmessage outside of the signal range of the hotspot networklong-distance wireless server, it saves this information packetin a buffer and continues to detect other SVMs or hotspotnetwork long-distance wireless servers. When this vehicle ap-proaches the first hotspot network long-distance wireless serv-er, it sends an SVM to the server to complete its mission as aDTN transmission relay station.

This model assumes that each driver in the following vehi-cle maintains a safe distance from the leading vehicle and thedeceleration factor is taken into account for the braking per-formance and drivers’ behavior. The complete mathematicalmodel is given by

S’ ¼ Lþ β’V þ γV 2 ð8Þwhere S’ is the headway spacing from the rear bumper to rearbumper, L is the effective vehicle length in meters, and V is thevehicle speed in meters/second. β’ is the driver reaction timein seconds, and the γ coefficient is the reciprocal of twice themaximum average deceleration of the following vehicle.

The car-following model is extended to the road level byreplacing the lane-level reaction time β’ with a road-levelinter-arrival time β. The lane-level car-following model canbe generalized as

S’ ¼ Lmin þ βV ð9Þwhere Lmin is the minimum spacing between any two adjacentvehicles, which is assumed to be zero in this study.

This system uses a Fon commercial network system as areference and installs a hotspot network long-distance wirelesssignal transceiver in the Wi-Fi base station. The user can ac-tivate this system’s free notification function by sharingWi-Fi.When the Wi-Fi base station installation becomes more pop-ular, the system’s service range will be significantly extended.The previously disconnected vehicle anti-theft communica-tion network can be securely integrated. This system uses lightradio Wi-Fi. The solution incorporates wireless and IP tech-nology. High-capacity and safe Wi-Fi seamlessly connectsvehicle terminal devices. Light radio multi-standardFemtocell and Metrocell systems incorporate a Wi-Fi option.

Among them, Metrocell provides high network capacity andwide network coverage in densely populated areas such asdowntown areas and airports. Femtocell is employed as apersonal base station to expand residential areas and to enableusers to automatically switchover between mobile network,home Wi-Fi or public Wi-Fi hotspots. Femtocell requires nei-ther a separate login nor extra charge, and the switchoverspeed is extremely fast. This technology will implement thepreviously mentioned function via three methods: it uses soft-ware to help users automatically identify, switchover, and con-nect in areas covered by trusted Wi-Fi networks and to im-prove network capacity.

5 Simulation and discussion

In this section, we present the simulation and discussion of avehicle-infrastructure integration using the CPAS mechanismunder an urban environment. Table 1 shows the simulationparameters and the range of values. The chosen parametersshould resemble those of high network density urban areas.

In the current study, we assume that in a real-world vehiclenetwork environment, a vehicle’s speed is affected by theDoppler effect and by any data communication’s success rateon the spectrum sensing problem. After we simulated andanalyzed the different numbers of vehicles and the differentvehicle speeds through MATLAB, we considered many fac-tors, including the simulation area, number of vehicles, num-ber of lanes, number of intersections, distance between inter-sections, RSUs, and adjustable vehicle speed. This paper eval-uates the performance of the proposed scheme. This workrandomly distributes 250 to 1800 vehicles in a 5.0 x 5.0 km2

field.Figure 6 shows an analysis of the time required to track a

stolen vehicle and the number of vehicles in the city. Threemechanisms are investigated and compared, including the ran-dom work mechanism, the mobile mechanism and the pro-posed CPAS mechanism. The random work mechanism de-scribes the time required for alarm processing when vehicletheft occurs in normal conditions. The police station will

Fig. 5 Vehicle periodically sendsSVM via a hotspot network long-distance wireless signal

664 Mobile Netw Appl (2016) 21:656–667

notify the police in the region to patrol and track within aregion. Although other nearby vehicles may occasionally dis-cover and send notifications, the number of nearby vehicleswill affect the time required to discover the stolen vehicle.Because the stolen vehicle is moving at high speed, it requiresa large amount of time to search and track the stolen vehicleand arrest the thief.

Then, the mobile mode is analyzed and compared. In mo-bile mode, an ordinary vehicle is equipped with a high-techpositioning module. When a vehicle is stolen, it will periodi-cally send vehicle positioning information and vehicle posi-tioning information via a message to the vehicle owner andrelevant vehicle security guard. Due to a vehicle’s highly mo-bile nature, which belongs to the large-area positioning modeand different landforms, positioning information signal maybe attenuated or become completely lost despite the availabil-ity of vehicle positioning information. Although the coordi-nated search time is shorter than the duration in conventionalmode, it requires a significant amount of time.

As shown in Fig. 6, when the number of neighboring ve-hicles increases in the case of the random work and mobilemechanism, an increase in adjacent vehicles around the citymay increase the likelihood of a stolen vehicle case report,which facilitates the analysis in areas with concentrated

message propagation. Thus, police in this area may requireless time to track a stolen vehicle and arrest the thief.However, the proposed CPAS mechanism can take advantageof the increased number of vehicles. The gossip protocol inconjunction with the adaptive goal programming’s selectionmechanism can yield the desired results and quickly searchand retrieve a stolen vehicle.

Next, the number of vehicles that have received stolen ve-hicle message notifications is compared and wait periods priorto the arrest of a vehicle thief are analyzed. Figure 7a showsthe average wait period prior to arrest for three mechanisms,including the random work mechanism, mobile mechanismand proposed CPAS mechanism. The proposed CPAS mech-anism employs gossip protocol to collect messages from ad-jacent vehicles and incorporates the adaptive goal program-ming’s selection mechanism. Thus, the proposed method re-quires less time to complete the arrest. Although the remainingtwo mechanisms have a multi-point case reporting method aswell as information and case location analysis, they cannotnarrow information and do not have an accurate positioningmethod and highly efficient DTN communication protocol.

Due to diversified road conditions in an urban environ-ment, such as buildings, corners, and crossroads, searchand tracking can be time-consuming if police vehicles arethe only employed vehicle. Next, we investigate coordi-nated information transmission using police vehicles, ad-jacent vehicles and stationary buildings to achieve ahigher tracking success rate. As shown in Fig. 7b, theanalysis and comparison of different mechanisms indi-cates a tracking success rate that surpasses other caseswhen vehicle and hotspot percentages are each 50 %.

The proposed CPAS mechanism for this system is investi-gated. A DTN-only configuration is compared with a gossip-only configuration and the packet loss ratio is analyzed, asshown in Fig. 7c. Various communication transmission dis-tances are addressed; when the communication area expands,the DTN protocol can compensate the flaws of wireless trans-mission in a large area. Therefore, the proposed CPAS mech-anism can achieve a low packet loss ratio in various environ-ments, including large or small areas: in small areas, gossipcan be employed; in large areas, a DTN can be employed.

The information propagation success rate is analyzed.Figure 7d shows and compares four different conditions, in-cluding integrated gossip and DTN for vehicles at speeds ofapproximately 50 or 80 mph, and two other cases with onlythe DTN or gossip mechanism. Various communication dis-tances are used to analyze the information propagation successrate. For short communication distances, the success rates offour different conditions exhibit a relatively high level, with anincrease in communication distance, and the success rates fordifferent conditions begin to significantly vary. We also dis-cover that the Doppler effect is insignificant in urban areaswhen vehicle speeds are less than 90 mph.

Table 1 Simulation parameters

Factor Range of values

Simulation area 5.0 × 5.0 km2

Number of vehicles 250-1800

Number of lanes 2 per direction

Distance between intersections 350 m

Distance between RSUs 500 m

Vehicle speed 10 – 70 km/h

0123456789

10

250 500 750 1000 1400 1800Sear

ch �

me

of st

olen

car n

orm

alize

d

Number of vehicles (Ck)

random work

mobile mechanism

our proposed mechanism (CPAS)

Fig. 6 Search time of stolen car for number of cars

Mobile Netw Appl (2016) 21:656–667 665

6 System evaluation

In addition to system simulation and analysis, we alsoimplement a modularized prototype of a vehicle road in-tegration positioning system. Figure 8a shows the trans-mission of information from hotspots installed in station-ary urban buildings to a CPAS server for additional cal-culations. Figure 8b shows that a hotspot is installed in

vehicle to facilitate V2V and vehicle-to-infrastructurewireless information transmission. During software devel-opment, open system and architecture and the OpenWRTdesign model are combined. Different mechanisms andapplications are compared, and the mobile mechanism isalso simulated and analyzed. The modularized prototypeis implemented and a comparative analysis is performed,as shown in Fig. 8c.

(a) (b)

(c) (d)

Fig. 7 (a) Average waiting timefor ratio of number of messages;(b) Success rate of tracking fornumber of cars and hotspots; (c)Packet loss ratio forcommunication range; (d)Dissemination successful forcommunication range

666 Mobile Netw Appl (2016) 21:656–667

Fig. 8 (a) Hotspots installed in stationary urban buildings; (b) shows that a hotspot is installed in vehicle to facilitate V2V and V2I; (c) Mobilemechanism (GPS and SMS)

7 Conclusion

In the paper, we propose an innovative design that opti-mizes a collaborative vehicle-finding network system, thedesign of vehicular cyber-physical alarm system and aninformation fusion platform with the purpose of protectingcitizens’ private property, enhancing anti-theft systems andincreasing the chances of finding cars in a smart city tomake vehicles safer, more convenient and suitable for sus-tainable development in urban areas. We design a cutting-edge algorithm that provides an accurate, fast and safesystem for any environment, communication distance, in-doors and even outdoors. We conclude that our proposedprotocol achieves substantial improvement of the successrate for vehicle recovery, increases the success rate oftracking, and reduces the search time in VANET environ-ments. In the simulation, our mechanism was better thanothers. We conclude that our proposed protocol achievessubstantial increases in the success rate of tracking, re-duces the packet loss ratio, and reduces the search timefor stolen cars in cyber-physical alarm city environments.

Acknowledgments We thank the Ministry of Science andTechnology of Taiwan for funding this research (Project no.:MOST 103-2221-E-218-037).

References

1. Scogna AC, Wang J (2008) BStudy of a conformal hidden wireantenna used for the detection of stolen cars^. IEEE Int SympElectromagnet Compat 1–6

2. Zhang X, Hang S, Chen H-H (2006) Cluster-based multi-channelcommunications protocols in vehicle ad hoc networks^. IEEEWirel Commun 13(5):44–51

3. Jones WD (2001) Keeping cars from crashing. IEEE Spectr 38:40–45

4. Rong D, Chen C, Yang B, Lu N, Guan X, Shen X (2015) BEffectiveurban traffic monitoring by vehicular sensor networks^. IEEETrans Veh Technol 64(1):273–286

5. Chisalita I, Shahmebri N (2004) BVehicular communication- a can-didate technology for traffic safety . 2004 I.E. Int Conf Syst, ManCybernet 4:3903–3908

6. Chen W, Cai S (2005) BAd hoc peer-to-peer network architecturefor vehicle safety communications^. IEEE Commun Mag

7. Benslimane A (2005) BLocalization in vehicular ad-hoc networks^.Proc 2005 Syst Commun 19–25

8. Mori JUJ, Nakamura Y, Horii Y, Okada H (2004) Development ofvehicular-collision avoidance support system by inter-vehicle com-munications. IEEE Vehicul Technol Conf 5:2940–2945

9. Milanes V, Perez J, Onieva E, Gonzalez C (2010) Controller forurban intersections based on wireless communications and fuzzylogic. IEEE Trans Intell Transport Syst 11(1):243–248

10. ASTM E2213-03 (2003) BStandard specification for telecommuni-cations and information exchange between roadside and vehiclesystems: 5 GHz band Dedicated Short Range Communications(DSRC) Medium Access Control (MAC) and Physical Layer(PHY) Specifications.^ ASTM Int

11. Weikl S, Bogenberger K (2013) Relocation strategies and algo-rithms for free-floating car sharing systems. IEEE Intell TranspSyst Mag 5(4):100–111

12. Isento JNG, Rodrigues JJPC, Dias JAFF, Paula MCG, Vinel A(2013) Vehicular delay-tolerant networks? a novel solution for ve-hicular communications. IEEE Intell Transp Syst Mag 5(4):10–19

13. Zhou J, Chen CLP, Chen L, Zhao W (2013) A user-customizableurban traffic information collection method based on wireless sen-sor networks. IEEE Trans Intell Transp Syst 14(3):1119–1128

14. Niyato D, Wang P (2009) BOptimization of the mobile router andtraffic sources in vehicular delay-tolerant network^. IEEE TransVeh Technol 58(9):5095–5104

15. Missoum S, G?rdal Z, Setoodeh S (2005) "Study of a new localupdate scheme for cellular automata in structural design". StructMultidisciplin Optim 29(2). doi: 10.1007/s00158-004-0464-2

16. Kraus S, Parshani R, Shavitt Y (2008) A study on gossiping intransportation networks. IEEE Trans Veh Technol 57(4):2602–2607

17. Liu K, Lee VCS, Ng JK-Y, Chen J, Son SH (2014) Temporal datadissemination in vehicular cyber-physical systems. IEEE TransIntell Transp Syst 15(6):2419–2431

18. Barua M, Liang X, Lu R, Shen X (2014) BRCare: extending securehealth care to rural area using VANETs^. Mobile Networks Appl19(3):318–330

19. Mao Y, Zhu P (2015) BA game theoretical model for energy-awareDTN routing in Manets with nodes’ selfishness^. Mobile NetworksAppl 20(5):593–603

20. Li Z, Liu Y, Zhu H, Sun L (2015) BCoff: contact-duration-awarecellular traffic offloading over delay tolerant networks^. IEEETrans Veh Technol 64(11):5257–5268

21. Liu J, Jiang X, Nishiyama H, Kato N (2013) BPerformance model-ing for relay cooperation in delay tolerant networks^. MobileNetworks Appl 18(2):186–194

22. Tornell SM, Calafate CT, Cano J-C, Manzoni P (2015) BDTN pro-tocols for vehicular networks: an application oriented overview .IEEE Commun Surv Tutor 17(2):868–887

23. Dias JAFF, Rodrigues JJPC, Isento JNG, Niu J (2013) BThe impactof cooperative nodes on the performance of vehicular delay-tolerantnetworks^. Mobile Networks Appl 18(6):867–878

Mobile Netw Appl (2016) 21:656–667 667