Research Article Dubhe: A Reliable and Low-Latency Data ...downloads.hindawi.com/journals/ijdsn/2013/581821.pdf · Research Article Dubhe: A Reliable and Low-Latency Data Dissemination

Hindawi Publishing CorporationInternational Journal of Distributed Sensor NetworksVolume 2013, Article ID 581821, 16 pageshttp://dx.doi.org/10.1155/2013/581821

Research ArticleDubhe: A Reliable and Low-Latency Data DisseminationMechanism for VANETs

Lifeng Zhang1,2 and Beihong Jin1

1 Institute of Software, Chinese Academy of Sciences, Beijing 100190, China2University of Chinese Academy of Sciences, Beijing 100190, China

Correspondence should be addressed to Beihong Jin; [email protected]

Received 14 June 2013; Revised 31 October 2013; Accepted 13 November 2013

Academic Editor: Floriano De Rango

Copyright © 2013 L. Zhang and B. Jin. This is an open access article distributed under the Creative Commons Attribution License,which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Many attractive applications over vehicular ad hoc networks (VANETs) need data to be transmitted to the remote destinationsthroughmultihop data forwarding, but some unique characteristics of VANETs (i.e., high nodemobility, dynamic topology changeswith frequent link breakage, and unstable quality of wireless transmission) incur unstable data delivery performance. In order toreliably and quickly disseminate the data, we present Dubhe which includes a delay model and an improved greedy broadcastalgorithm embedded with a coverage elimination rule. The former is used for making decisions for path selection with the aimof minimizing the transmission latency, while the latter focuses on boosting the reliability of one-hop data transmission. We alsoanalyze the necessity and effectiveness of Dubhe and the retransmission overheads theoretically. It is shown from the experimentsthat Dubhe can achieve high-reliability and low-latency data delivery in comparison with the epidemic-based protocol and thestatic-node-assisted adaptive data delivery protocol.

1. Introduction

As a mobile and self-organizing network, a VANET is builtover two types of nodes on an as-needed basis, that is, themobile nodes equipped with on-board units (OBUs) andthe stationary wireless stations deployed along the roadsknown as roadside units (RSUs). Both OBUs and RSUshave wireless interfaces to connect together or to the otherequipment such as traffic lights. The communication over aVANET can be done by means of vehicle-to-vehicle (V2V)and Vehicle-to-Infrastructure (V2I) communications. Basedon these communications, many promising applicationsover VANETs can be envisioned in the future with thehelp of enabling technologies (e.g., vehicular positioning,digital maps, and sensors), such as active safety appli-cations, intelligent transportation applications, and vari-ous convenience applications for drivers and passengers.More specially, VANETs are also underlying networks toprovide access to the Internet, as well as data transmis-sion in case of emergency (earthquakes, traffic jams, etc.)[1].

In this paper, we focus on the reliable and low-latencylong distance data transmission in urban scenarios, which isindispensable to many applications. For example, a movingvehicle may initiate an inquiry for available parking lotsof a business center several miles away to decide the nexttravel destination; the hazardous traffic situations collectedby vehicles cooperatively need to be diffused timely to theinfostations deployed at adjacent blocks so as to remindthe driver of rerouting. Although RSUs connected to thebackbone networks can provide a stable link for the datatransmission, the high expenses impose restriction on thedeployment of RSUs on a large scale, especially in the initialphase of deploying aVANET.A common idea is to place RSUsat intersections firstly in urban areas, where the RSUs cantake on multiple roles: storing the received data temporally,making the decisions on the data routing, and forwarding thedata to the passing vehicles if necessary. Meanwhile, vehicleshave to cooperate with each other to deliver the data. That is,a vehicle will forward the data in case that there is a chance toestablish a wireless link with other vehicles; otherwise, it willcarry the data by itself.

2 International Journal of Distributed Sensor Networks

The wireless connections among vehicles are extremelyuncertain owing to the rapid movements of vehicles. Whenvehicles are sparse, the VANET falls into a disruptive network[2]. However, a dense vehicle flowmay lead to severe channelcontentions due to the broadcast nature of the VANETs.Therefore, the key issue of the data delivery over a VANET ishow a vehicle node on a road segment determines its action:carrying on or forwarding or dropping the received packets,and how a node (a vehicle or an RSU) at an intersectionchooses the optimal paths for the packets to be delivered.

As far as the selection of relay nodes is concerned,some algorithms including the broadcast suppression tech-niques [3] have been developed to assist in the selection ofrebroadcasting nodes, but they lack the theoretical analysison the reliability of data delivery. On the other hand, asfor the routing path selection, a variety of data deliveryprotocols have been proposed. Typically, while a data packetis transmitted to a node at an intersection, the optimal pathaiming at minimizing the transmission delay is expected.Some protocols require nodes to maintain the routing infor-mation [4, 5], as thus, considerable packets have to beexchanged to keep routing tables up to date. However, someother protocols (e.g., VADD [6]) exploit the historical trafficdata to decide the path for packet forwarding. Obviously,the historical data are almost certainly inconsistent withthe constantly changing traffic situations. So far, lacking anaccurate model of reflecting the time-varying connectivity ofa VANET has become the biggest obstacle to improving theperformance of data delivery, where the connectivity refersto the probability that a node (i.e., a vehicle or an RSU)can communicate with any other nodes through wirelesscommunication.

Motivated by the limitations of the existing protocolsand noting that the vehicles in a typical city scenario arerestricted in grid roads, we present Dubhe, a reliable andlow-latency data delivery mechanism for transmitting datato stationary roadside infrastructures several miles awayfrom source vehicles by unicast communication. Dubhecomprises a delay model, a path choosing algorithm, and animproved greedy broadcast algorithm. The delay model isresponsible for estimating the transmitting delay between theadjacent intersections with RSUs deployed, the path choosingalgorithm is used for determining the optimal path for thepacket arriving at RSUs, and the broadcast algorithm con-centrates on improving the reliability of one-hop data trans-mission.The contributions of this paper include the followingaspects.

(1) We propose a delay model which takes accountof the road topology and the interference of thetraffic signals in an urban scenario. We also give ourpath choosing algorithm on the basis of the delaymodel.

(2) We present an improved greedy broadcast algorithmmerged with a coverage elimination rule. The multi-candidate strategy in algorithm can enhance thereliability of one-hop data transmission.

(3) We analyze the necessity and feasibility of the multi-candidate strategy for 1-hop data transmission and thecorresponding retransmission ratio theoretically.

(4) We evaluate the performance ofDubhe by conductingextensive experiments to examine the performanceand cost metrics (i.e., delivery ratio, transmissiondelay, and message overhead) over three independentfactors (i.e., vehicle density, traffic signal cycle, andconnectivity estimation period).

The rest of this paper is organized as follows. In Section 2,we introduce the existing researches related to data deliv-ery mechanisms over VANETs. In Section 3, we describeour delay model and path choosing algorithm. And then,we present our greedy broadcast algorithm and analyzethe reliability and the retransmission ratio in Section 4. InSection 5,we conduct the performance evaluation. Finally, wesummarize our work in Section 6.

2. Related Work

Rapid changes of network topology, lack of central coordina-tion, and limited bandwidth are stumbling blocks to efficientdata delivery over VANETs [7]. Besides, radio channelsare error-prone. The wireless signals interfere with vehiclesin unpredictable ways, and the fading phenomenon (themultipart of the same wireless signal traveling along differentpaths that interfere with each other) deteriorates the channelquality even in the case of no neighbor trying to trans-mit simultaneously. All lead to apparently nondeterministicpacket reception.

As is known, the best-effort manner is commonly usedfor data delivery over a MANET. For example, following theepidemic based data delivery (EBDD) protocol [8], a nodeexchanges digests of all data packets with other nodes aslong as they can communicate with each other. Then theytransmit the data packets that do not appear in the packet listsof counterparts in turn. However, the performance degradesquickly while facing a high vehicle density owing to theexcessive repeated packet exchanging.

Due to the broadcast nature of wireless signals, a wirelesstransmission with the target to a particular destinationvehicle can be overheard by all neighboring vehicles; how tochoose an appropriate relay vehicle frommultiple candidatesis a tough task. In [9], a decentralized broadcast algorithm ispresented to achieve a good broadcast coverage in a VANET.Beside the broadcast suppression techniques introduced in[3], a connected dominating set (CDS) on the basis of localbeacon messages is proposed in [10] with the expectation ofrepressing the overhead of broadcast. As thus, only the nodeswithin a CDS have a chance to transmit the received packets.Although broadcast messages are partially diminished, theCDS suffers from the overheads for constant CDS construc-tion owing to the fast movements of nodes. Obviously, theirgoal is different from ours.

Further, in order to find an efficient path for forwardingthe packet, some studies propose to propagate the vehicledensity of roads among vehicles to acquire the route decision-making information, for example, periodical probing packets

International Journal of Distributed Sensor Networks 3

are used [4] to collect the vehicle density proactively andthen are exchanged among neighbors. Obviously, packets forprobing and propagating route information induce massiveoverheads. By contrast, [5, 11] employ a reactive route dis-covery approach with the help of the relay nodes. However,the routing tables are out of date quickly because of themobility of the relay nodes. Since the historical data oftraffic flow are an important basis for deciding the optimalpath, vehicle-assisted data delivery (VADD) [6] exploitsthe historical records of the vehicle density marked ona digital map so that a node can decide the minimumdelay paths for the arriving packets. But the time-varyingvehicle density may deviate far from the one marked onthe map. To remedy the weakness, [12] presents a static-node-Assisted adaptive Data delivery protocol for Vehicularnetworks (SADV), which suggests to place static nodes atintersections to assist packet forwarding. Each static nodehas the capability of storing the received packets untilthey can be forwarded to a vehicle travelling toward theoptimal forwarding road. Meanwhile, transmission delay ispiggybacked in data packets to help forwarding decisions.Nevertheless, SADV is designed entirely without considera-tion for the impact of the spatial distribution of vehicles onthe connectivity of VANETs. Similarly, [13] suggests bufferingdata in the parked vehicles along the roads if no appropriateroute path for data forwarding is available in urban areas.But the efficiency of data delivery is unstable, depending onthe number of vehicles which are willing to relay messages asvolunteers.

To enhance the adaptation and scalability of routingprotocols, [14] proposes to select the intermediate nodesaccording to the group mobility patterns, importance of themessages, destination, current location, and speed of thevehicles. Reference [15] periodically evaluates the SIR (signal-to-interference ratio) values on the available transmissionchannels and chooses the channel with the best perceivedSIR to deliver the data. Reference [16] presents a hierarchicalrouting scheme for data forwarding decisions, which makesuse of peer servers to discover the trajectory from a sourceto a destination. However, the overhead of electing peerservers and maintaining the information on the peer serversis not ignorable. On the behalf of enhancing the reliabilityof data transmission over VANETs, [17] proposes to utilizeblack-burst signals to prevent the receivers from concurrentrebroadcasting. At the same time, [17] employs acknowledge-ment packets to improve the reliability and a mechanismsimilar to the RTS/CTS handshake to decrease the effectsof the hidden nodes. But the overheads incurred by theblack-burst and the acknowledgement messages restrict itsscalability.

The network connectivity is crucial to efficiently deliverdata over VANETs. Usually, one-hop connectivity betweennode 𝑖 and node 𝑗 can be specified as 𝑃

𝑟(𝑑(𝑖, 𝑗) < 𝑅), where

𝑑(𝑖, 𝑗) is the Euclidean distance between two nodes, and 𝑅

denotes the radius of wireless transmission (Table 2). Refer-ence [18] examines the connectivity of VANETs in an urbanscenario by experiments, and the results show that the proba-bility of an uninterrupted multi-hop link decreases exponen-tially with the increasing length of the road. More concretely,

when the vehicle density is 4 vehicles/250m, the probabilityof the formation of a 1000m wireless communication is just68%, and 42% as the road length is increased to 2000m.While the vehicle density reduces to 2 vehicles/250m, the cor-responding probabilities decline to 21% and 7%, respectively.Hence, some studies believe that it is unfeasible to achievereliable and efficient data delivery without the support ofRSUs [18]. Further, the experimental results in [19] alsoindicate that once the vehicle density is below a specifiedthreshold (e.g., 60 vehicles/km2), the quality of service of datatransmission cannot be satisfied if it relies solely on the V2Vcommunications.

Although the connectivitymodels are of great importanceto measure the data transmission delay in VANETs, the mostexisting studies only focus on the highway scenarios [20]. Ref-erence [21] employs discrete-time and discrete-spaceMarkovchains to estimate the changing of vehicle velocities so asto infer whether the network partition would occur. Unfor-tunately, the computational complexities of the estimationmethods refrain them from putting into practice for routingdecisions. Considering the significant impact from the trafficsignals on the speeds and spacing distribution of the vehicles,[22] derives the connectivity model of a single road segmentin urban environments. However, [22] needs traffic flowdata of all entrances of the road segments to calculate theconnectivity, so it is improper to be directly used for routingdecision.

In summary, the frequent disruptions of traffic flow inurban roads seriously affect the connectivity of VANETs. Butthe existing data delivery protocols still fail to adapt to thetime-varying vehicle density resulting from the stochasticvehiclemobility andperiodic interference of the traffic signals[23].

3. Connectivity-Aware Data Delivery

We propose deploying RSUs at major intersections to col-lect data for calculating the vehicle arrival ratios and thenexchanging them with the neighboring RSUs periodicallythrough connectivity messages. Next, the RSUs can employthe arrival ratios to estimate the real-time connectivity andexchange the transmitting delay with the adjacent RSUs byconnectivitymessages.Then, the RSUs can choose a forward-ing path for the arriving packets in terms of the minimumdelay.

On the other hand, each vehicle and each RSU areequipped with a GPS device and a digital map on which thepositions of RSUs are marked.

3.1. Estimating Arrival Ratios of Vehicles. The vehicles arerequired to broadcast beacon messages periodically, so anRSU can be aware of the approaching vehicles, and thenestimate the arrival ratios of different directions of an inter-section for every time interval 𝜏. The arrival ratios calculatedby an RSU (denoted as 𝑅

𝑖; see Table 1) are sent to the

neighboring 𝑘-hop RSUs (here, 1-hop RSU is defined as anRSU immediate adjacent to the RSU 𝑅

𝑖) with the assistance

of the passing vehicles. If a neighboring RSU (denoted as 𝑅𝑗)


Table 1: The notations used in the delay model.

𝜏 Time interval for estimating the transmission delay𝜆 Arrival ratio during 𝜏𝛽 The average vehicle density𝐴 The number of arrival vehicles during [𝑡 − 𝜏, 𝑡]

𝐷 The number of vehicles departing in one direction𝜔 Time slices of 𝜏𝐿𝑖𝑗

The road length from RSU 𝑖 to 𝑗V𝑖𝑗

The average vehicle velocity of the road 𝑖𝑗

𝑅 The transmission range of wireless signals𝑑0

Average one-hop wireless transmission delay𝜌𝑖𝑗(𝑡) The vehicle density of road 𝑖𝑗 at time 𝑡

𝑑𝑖𝑗(𝑡) The transmission delay of road 𝑖𝑗 at time 𝑡

𝐸(𝑁𝑖𝑗(𝑡)) The expected number of vehicles of road 𝑖𝑗 at time 𝑡

𝑉𝑠(𝑖𝑗)

The vehicles that go straight through the intersection𝑖 into the road 𝑖𝑗

𝑉𝑡(𝑖𝑗)

The vehicles that enter road 𝑖𝑗 from theperpendicular direction during the green phase

Table 2: Notations list of reliability analysis.

𝑆 The sender node𝑅 The transmission radius𝑡𝑏

The period of beacon messagesV𝑥

The velocity of the vehicles

𝐶0

The furthest vehicle from 𝑆 in the direction of packettransmission

𝐿 The distance from 𝑆 to 𝐶0

𝐵0

The region without vehicle marked with red lines

𝑚The distance from 𝐶

0to the horizontal transmission

boundary of 𝑆𝜃 The horizontal angle

does not receive the reports from 𝑅𝑖within a given interval,

then 𝑅𝑗forecasts it based on the historical arrival ratios of

𝑅𝑖recorded in the local storage by applying the regression

technique (see the following paragraph). If the delayed arrivalratios of 𝑅

𝑖reach later, then 𝑅

𝑗replaces the estimated ones

with the incoming arrival ratios.In order to predict the missing arrival ratios of vehicles at

neighboring intersections that should have been reported, the𝑘-nearest neighbor (𝑘NN) algorithm [24] is used to capturethe traffic patterns from a time series of data, given that thetraffic flow has periodical patterns [25]. As thus, a historicalarrival ratio record turns into a state vector in 𝑘NN whichcontains several attributes depicting the arrival ratios, that is,the time and the date attributes (e.g., hour of the day, the dayof the week), and then it is stored locally in RSUs.

The estimation process is activated when the missing ofcurrent arrival ratio is detected. To simplify the calculation,the historical records are stored in a KD-tree and the recordsretrieved from the local storage are restricted to latest fouritems represented by 𝜆

ℎ1, 𝜆ℎ2, 𝜆ℎ3, and 𝜆

ℎ0.Then, the missing

arrival ratio of vehicles at a neighboring intersection can

be estimated as 𝜆 = ∑𝐻

𝑖=0𝜇𝑖𝜆ℎ𝑖, where 𝐻 = 4 and 𝜇

𝑖=

𝑑𝑖

−1

/∑𝐾

𝑖=0𝑑𝑖

−1. 𝑑𝑖denotes the dissimilarity distance between

the subvector from the historical record and the currentstate and is measured by the Euclidean distance; that is,𝑑𝑖= √∑

𝐾

𝑗=1(𝑝𝑗− 𝑞

𝑗)2, where 𝐾 is the dimension of the state

vector, 𝑝𝑗is the 𝑗th element of the historical record under

consideration, and 𝑞𝑗is the 𝑗th element of the current state.

3.2. Delay Model. All RSUs are assumed to be standalone,they determine the forwarding paths for the arriving packetswith the goal of minimizing the transmission delay. Notethat the transmission delay depends heavily on the wirelessconnectivity of the VANETs on the roads, so the delay modelis derived through the real-time estimation of the expectednumber of vehicles and vehicle distribution between adjacentRSUs.

An RSU works as follows to decide the optimal paths forthe arriving packets.

(1) The RSU builds a Markov model for vehicle queuelength at intersections. At time 𝑡

0, the RSU 𝑅

𝑖esti-

mates the queue length of the neighboring intersec-tions of time 𝑡

0+ 𝜏 in the light of the current arrival

ratios of neighbor RSUs.

(2) TheRSU estimates the transmission delays in terms ofthe real-time varying vehicle number and distributionbetween the adjacent RSUs.

(3) The RSU calculates a minimum-delay forwardingpath to a virtual destination V

𝑑for a packet with an

improved Dijkstra algorithm, where V𝑑refers to the

RSU nearest to the original destination of the packetand can exchange arrival ratios with the intendedforwarding RSU 𝑅

𝑖. All the destinations refer to V

𝑑in

the rest of the paper.

In the following, we elaborate the vehicle queue lengthprediction, delay estimation, and path choosing, respectively.

3.2.1. Vehicle Queue Length Prediction. Thearrival of vehiclesat an intersection can be viewed as a stochastic processfollowing the Poisson distribution [26, 27]. As a result, thequeue length of vehicles at an intersection at time 𝑡 can bemodelled with a random variable 𝐿(𝑡):

𝐿 (𝑡) = max {𝐿 (𝑡 − 𝜏) + 𝐴 − 𝐷, 0} , (1)

where 𝜏 is a small time interval by which a green phaseof traffic signal is divided into 𝑛 equal episodes and 𝐴 is arandom variable standing for the number of arrival vehiclesduring [𝑡 − 𝜏, 𝑡].Meantime, 𝐷 represents the number of


vehicles departing from the intersection in one direction; weassume that the departure ratio 𝐺 is a constant during thegreen phase. Equation (1) points out that the queue lengthat time 𝑡 depends only on the queue length at time 𝑡 − 𝜏,the arrival and departed vehicles during 𝜏; therefore, it is arenewal process. So the RSU can predict the queue length at(𝑘+ 1)𝜏with the arrival ratios at time 𝑘𝜏 (𝑘 and 𝑛 are positiveintegers, 0 < 𝑘 < 𝑛) in an iterative manner. Note that 𝜏 is alsothe period of connectivity messages.

The prediction is based on the transition probability 𝑞𝑖𝑗,

which is the probability that the queue length changes from 𝑖

to 𝑗 in a small interval 𝜔 (𝜏 = 𝑢 × 𝜔, 𝑢 is a positive integer):

𝑞𝑖𝑗(𝜔) = {

𝑃𝑟(𝐴 (𝜔) = 𝑗 − 𝑖 + 𝜔𝐷) 𝑗 > 𝑖 − 𝜔𝐷

0 otherwise,(2)

where 𝑖, 𝑗 > 0. Consequently, the probability that the queuelength changes from 𝑖 to 0 can be calculated by the followingequation:

𝑞𝑖0(𝜔) =

{{

{{

{

𝜔𝐷−𝑖

∑

𝑘=0

𝑃𝑟(𝐴 (𝜔) = 𝑘) 𝑖 ≤ 𝜔𝐷

0 𝑖 > 𝜔𝐷.

(3)

At any interval 𝜔, the probability of finding 𝑖 vehiclesarriving is written as

𝑃 (𝐴 = 𝑖) = 𝑒−𝜆𝜏 (𝜆𝜔)

𝑖

𝑖!, (4)

where 𝜆 is the arrival ratio during 𝜏.Note that the probability distribution of a queue length

at time 𝑘𝜏 during a green phase can be denoted as 𝑃(0) =

[𝑃(0)

0, 𝑃

(0)

1, 𝑃

(0)

2, . . . , 𝑃

(0)

𝑚] (where 𝑃(0)

𝑙denotes the probability

that the queue length equals 𝑙 and𝑃(0) is randomly initialized)[26]; then probability distribution at 𝑘𝜏 +𝜔 can be estimatedby 𝑃

(1)

= 𝑃(0)

𝑞𝑖𝑗(𝜔). Then after 𝑢 iterations, the probability

distribution about the queue length is

𝑃(𝑢)

= 𝑃(𝑢−1)

𝑞𝑖𝑗(𝜔) . (5)

With the queue length at time 𝑘𝜏, the expected queuelength at time (𝑘 + 1)𝜏 is calculated with

𝐸 (𝐿 ((𝑘 + 1) 𝜏)) =

𝑚

∑

𝑗=0

𝑗𝑃(𝑢)

, (6)

where𝑚 is the dimension of the transition probabilitymatrix.During the red phase of traffic signal, (6) is still valid as longas the parameter𝐷 in (2) is set to 0.

3.2.2. Delay Estimation. With the prediction of the changesof the queue length, we then estimate the transmission delayof a specific road segment. To simplify the derivation, wesuppose that all roads are of two lanes and the traffic signalsare synchronized.

The beginning time of the green phase of the traffic signalis denoted as time 0. So the shortest time to form the wireless

links from RSU 𝑖 to 𝑗 is 𝑀𝑇𝑖𝑗

= (𝐿𝑖𝑗− 𝑅)/(2 × V

𝑖𝑗) after

the green phase has started; it is reasonable that we supposethat the spacing of vehicles on the road 𝑖𝑗 follows the Poissondistribution [26] after 𝑀𝑇

𝑖𝑗. Then, the one-hop connectivity

probability between vehicles at time 𝑡 can be specified as

𝑝𝑟(𝑥 < 𝑅) = 1 − 𝑒

−𝜌𝑖𝑗(𝑡)𝑅, (7)

where 𝑥 represents the spacing between vehicles and 𝜌𝑖𝑗(𝑡) is

the vehicle density. Notice that 𝑡 ≥ 𝑀𝑇𝑖𝑗and 𝑡 = 𝑚𝜏; that

is, the estimation of transmission delay is only executed at anintegral multiple of time interval 𝜏.

Since 𝑉𝑠(𝑖𝑗) during the interval 𝜏 can be estimated by

𝑉𝜏

𝑠(𝑖𝑗) = 𝛼max {𝐸 (𝑁

𝑖(𝜏)) + 𝐸 (𝐿

𝑖(𝑡 − 𝜏)) , 𝜏𝐷} , (8)

where 𝐸(𝑁𝑖(𝜏)) denotes the number of the vehicles arriving

intersection 𝑖 during 𝜏 and 𝛼 is the ratio of the vehicles goingstraight through intersection 𝑖 into the road 𝑖𝑗, apparently, thenumber of vehicles on the road 𝑖𝑗 at time 𝑡 can be given by

𝐸 (𝑁𝑖𝑗(𝑡)) = 𝐸 (𝑁

𝑖𝑗(𝑡 − 𝜏)) + 𝑉

𝜏

𝑠(𝑖𝑗) − 𝐸 (𝐴

𝑗(𝜏)) . (9)

Equation (9) implies that the number of vehicles of road𝑖𝑗 at time 𝑡 depends only on the number of vehicles attime 𝑡 − 𝜏, the number of vehicles going straight throughthe intersection, 𝑖 toward intersection 𝑗 and the number ofarriving vehicles at intersection 𝑗. Note that𝐸(𝐴

𝑗(𝜏))needs to

be calculated only after 𝑡 > 𝐿𝑖𝑗/V𝑖𝑗in that no vehicle in 𝑉

𝑠(𝑖𝑗)

could arrive at intersection 𝑗 before 𝐿𝑖𝑗/V𝑖𝑗.

Thus, the vehicle density during the green phase is givenby

𝜌𝑖𝑗(𝑡) =

(𝐸 (𝑁𝑖𝑗(𝑡)) + 𝐸 (𝑁

𝑗𝑖(𝑡)))

𝐿𝑖𝑗

. (10)

During the red phase of the traffic signal, since thewireless links may still be connected before time elapsed𝑀𝑇

𝑖𝑗. Considering that some vehicles (denoted by 𝐸(𝑇

𝑖(𝑡)))

on the road 𝑖𝑗 are entered from the perpendicular directionduring green phase, 𝐸(𝑇

𝑖(𝑡)) is proportional to the vehicles

passed in the perpendicular direction of the intersection (theratio is 1−𝛼). So the number of vehicles on the road 𝑖𝑗 can beestimated as

𝐸 (𝑁𝑖𝑗(𝑡)) = 𝐸 (𝑁

𝑖𝑗(0)) − 𝐸 (𝐴

𝑗(𝑡)) + 𝐸 (𝑇

𝑖(𝑡)) . (11)

From (11) we can observe that the expected number ofvehicles on the road 𝑖𝑗 at time 𝑡 equals the number of vehiclesat the beginning of the red phase subtracting the number ofvehicles arriving at intersection 𝑗 after the beginning of thered phase and adding the vehicles entering the road from theperpendicular direction. Specially, the wireless link of 𝑉

𝑠(𝑖𝑗)

could be possible only after 𝑀𝑇𝑖𝑗passes in the green phase;

meanwhile, 𝑉𝑡(𝑖𝑗) contributes to the connectivity of the road

at a red phase. Therefore, the vehicles on the road 𝑖𝑗 can beregarded as the 𝑉

𝑠(𝑖𝑗) superimposed with 𝑉

𝑡(𝑖𝑗). Thus, the

transmission delay of road 𝑖𝑗 at time 𝑡 of a green phase canbe written as follows:

𝑑𝑖𝑗(𝑡) = {

𝜎𝑖𝑗(𝑡) + 𝐷

(1)

𝑖𝑗(𝑡) 𝑡 ≤ 𝑀𝑇

𝑖𝑗

𝐷𝑖𝑗(𝑡) 𝑡 > 𝑀𝑇

𝑖𝑗.

(12)


Equation (12) indicates that before a wireless link isformed, the packets are carried or transmitted by 𝑉

𝑠(𝑖𝑗) and

entered during the green phase together with the vehiclesturned into the road 𝑖𝑗 during last red phase. 𝜎

𝑖𝑗(𝑡) is the

resulting delay of 𝑉𝑡(𝑖𝑗), the vehicle density of 𝑉

𝑡(𝑖𝑗) is

estimated by the numbers of vehicles passing in the perpen-dicular direction during the green phase of intersection 𝑖 and𝑗, that is, 𝜌(2)

𝑖𝑗(𝑡) = (𝐸(𝑇

𝑖(𝑡)) + 𝐸(𝑇

𝑗(𝑡)))/(𝐿

𝑖𝑗− 2V

𝑖𝑗𝑡), so

𝜎𝑖𝑗(𝑡) = (1 − 𝑒

−𝑅𝜌(2)

𝑖𝑗(𝑡)

)

𝐿𝑖𝑗− 2V

𝑖𝑗𝑡

𝑅𝑑0+ 𝑒

−𝑅𝜌(2)

𝑖𝑗(𝑡)𝐿𝑖𝑗− 2V

𝑖𝑗𝑡

V𝑖𝑗

,

(13)

where 𝑒−𝑅𝜌(2)

𝑖𝑗(𝑡) is the probability that the spacing of vehicles

is larger than 𝑅. If so, the packets have to be carried by thevehicles. As suggested in (13), the transmission delay relieson the vehicle density heavily; a high vehicle density makesfor a small portion of data packets completed by the vehiclemovements;𝐷(1)

𝑖𝑗(𝑡) is the delay caused by the 𝑉

𝑠(𝑖𝑗):

𝐷(1)

𝑖𝑗(𝑡) = (1 − 𝑒

−𝑅𝜌(1)

𝑖𝑗(𝑡)

)

2V𝑖𝑗𝑡

𝑅𝑑0+ 2𝑒

−𝑅𝜌(1)

𝑖𝑗(𝑡)

𝑡. (14)

Similar to (10), the vehicle density of 𝑉𝑠(𝑖𝑗) is 𝜌(1)

𝑖𝑗(𝑡) =

(𝐸(𝑁𝑖𝑗(𝑡)) + 𝐸(𝑁

𝑗𝑖(𝑡)))/(2V

𝑖𝑗𝑡), when the spacing of vehicles

satisfies the exponential distribution after time elapsed𝑀𝑇𝑖𝑗;

the data delivery is performed mainly by 𝑉𝑠(𝑖𝑗) and the

transmission delay is given by

𝐷𝑖𝑗(𝑡) = (1 − 𝑒

−𝑅𝜌𝑖𝑗(𝑡))

𝐿𝑖𝑗

𝑅𝑑0+ 𝑒

−𝑅𝜌𝑖𝑗(𝑡)𝐿𝑖𝑗

V𝑖𝑗

. (15)

𝐷𝑖𝑗(𝑡) indicates that the total delay is composed of the

delay caused by vehicles carrying on with proportion of𝑒−𝑅𝜌𝑖𝑗(𝑡) and the delay incurred by wireless transmission whenthe intervehicles distances are smaller than 𝑅. It must beemphasized that on the calculation of the expected transmit-ting delay between adjacent RSUs, the impacts of the trafficsignal cycle and the reasonable conditions of the exponentialspacing distribution to be fulfilled are considered.

3.3. Path Choosing. When the RSU at an intersection 𝑖

receives a packet with V𝑑as its destination, it calculates

the minimum delay from intersection 𝑖 to V𝑑through an

improved Dijkstra algorithm. Considering that the trafficflow is uneven, we need to scale up the weight of delay𝑑𝑘,𝑘+1

on the roads nearer to the intersection 𝑖 so as toavoid being deceived by the short latencies of the roads faraway from the intersection 𝑖. Therefore, on the basis of theoriginal Dijkstra algorithm, we introduce (16) to calculate theminimum forwarding delay 𝐷

𝑖,V𝑑, of which 𝛽𝑘,𝑘+1

∈ (0, 1) isthe weight of delay 𝑑

𝑘,𝑘+1and 𝑙 is the searching depth from 𝑖

to V𝑑in the algorithm

𝐷𝑖,V𝑑 =

𝑘=𝑙

∑

𝑘=0

(𝛽𝑘,𝑘+1

)𝑘

𝑑𝑘,𝑘+1

. (16)

The parameter 𝛽𝑘,𝑘+1

depends on the average vehicledensity on the road (𝑘, 𝑘 + 1). The higher the vehicle densityis, the greater the value of 𝛽

𝑘,𝑘+1is assigned. On that account,

𝛽𝑘,𝑘+1

is decided by (17), in which the offset 𝜃 representsthe least number of vehicles in the transmission range 𝑅 toform a successive wireless connection [18]; 𝜃 is set to 6 inexperiments:

𝛽𝑘,𝑘+1

=1

1 + 𝑒−(𝜌𝑘,𝑘+1𝑅−𝜃)

. (17)

4. Reliable Data Delivery

The propagation of wireless signals in urban environments isinterfered by the buildings, facilities along the roads, and therapid vehicle movements in different directions, which maygive rise to the loss of data transmitted Therefore, improvingreliability is an imperative requirement for the data delivery.Due to the broadcast nature of the wireless medium inVANETs, a single packet often evokes multiple receptions.Thus, it is crucial in determining the appropriate neighboringvehicles to rebroadcast the data in a distributed way with thepromise of guaranteeing the reliability.

4.1. Improved Greedy Broadcast Algorithm. Dubhe providesan improved greedy broadcast algorithm to choose the properrelay vehicles. In detail, the sending vehicle broadcasts apacket which contains the payload to be sent, the packetID, the list of 1-hop neighboring IDs acquired by beaconmessages, the location of the last RSU passed by, and thedistance from the sending vehicle to the passed RSU. Ifthe vehicle finds that there is no neighbor in the directionof next RSU within the transmission range, it carries thepacket. However, on receiving a packet, the vehicle appliesthe coverage elimination rule; that is, it checks whether allits neighbors are covered by the neighbor list of the packetor not, if covered, then the recipient discards the packet,otherwise it becomes a forwarding candidate. Such a coverageelimination rule can effectively reduce the candidate number.

Being a candidate implies that the vehicle needs to starta timer according to the distance from the sending nodein preparation for the packet rebroadcasting. The longer thedistance is, the shorter the timer is set. In this way, the vehiclefurthest away from the sender has the highest priority torebroadcast the packet. On condition that the furthest onefails, the secondary vehicles can rebroadcast the packet aftera little while. Therefore, a multicandidate situation is formed.

In Dubhe, the packet may be rebroadcast by the vehiclesmoving in the opposite direction, which can improve theconnectivity and increase the efficiency of packet forwarding.But in order to prevent the packet from being propagated inthe inverse direction, the vehicle updates the distance fromthe last passed RSU in the packet before it rebroadcasts thepacket. The improved greedy broadcast algorithm deployedon vehicles is shown in Algorithm 1.

It is clear that the maximum waiting time 𝑇 is a criticalparameter toAlgorithm 1, which closely relies on the one-hop


(1) //Definition: 𝑢 denotes a vehicle in a VANET,𝑁(𝑢)

denotes a neighbor list of 𝑢;(2) //𝑂 denotes the last intersection passed by; dist

(𝑂, 𝑢) denotes the length from 𝑂 to 𝑢;(3) //𝑅 denotes wireless transmission radius;(4) //𝑀 denotes the data packet which needs to be

delivered, and consists of ID, data, sender node 𝑠, aneighbor list of 𝑠 𝑁(𝑠), the coordinates of 𝑂, and thedistance between 𝑂 and 𝑠

(5) On receiving a data packet𝑀 at node 𝑢(6) Extract dist(𝑂, 𝑠), coordinates of 𝑂 and𝑁(𝑠)

from𝑀;(7) If 𝑀 has been received before or 𝑢 ∉ 𝑁(𝑠) then(8) Drop data packet𝑀;(9) If exists any defer timer on𝑀 then(10) Cancel all the timers;(11) return;(12) If dist(𝑂, 𝑢) ≤ dist(𝑂, 𝑠) then(13) Drop data packet𝑀;(14) If𝑁(𝑢) ⊆ 𝑁(𝑠) then(15) Drop data packet𝑀;(16) Set defer timer 𝑑

1time to 𝑇;

(17) On expiration of deferred timer 𝑑1

(18) Broadcast data packet𝑀;(19) Set 𝑡 equals to (𝑅 − (dist(𝑂, 𝑢) mod 𝑅)) ∗ 𝑇/𝑅;(20) Set defer timer 𝑑

2with 𝑡;

(21) On expiration of deferred timer 𝑑2

(22) Update𝑀: replace𝑁(𝑠) with𝑁(𝑢),dist(𝑂, 𝑠)with dist(𝑂, 𝑢)

(23) Broadcast data packet𝑀

Algorithm 1: An improved greedy broadcast algorithm withcoverage elimination.

wireless transmitting time, vehicle density, and wireless com-munication range and queue length of the packet in theMAClayer. The similar deferring policy is utilized in [3], whichadvises to set no less than 5ms when the vehicle density is100 vehicles/km/lane. Taking all factors into consideration, 𝑇is set to 20ms in our experiments. Additionally, in order toavoid the expanding of the neighbor list in the packet whenthe vehicle density is high, the probabilistic data structure(e.g., bloom filter) can be used. Meanwhile, the interval ofbeacon messages can be dynamically adjusted according tothe vehicle density to reduce the communication overheads[11].

4.2. Algorithm Analysis. Algorithm 1 attempts to choose thefurthest vehicle to forward the packets and meanwhile letsmultiple neighboring vehicles be the alternatives in case ofthe failure of the furthest node. Such a design aims to providethe robust packet transmission.

The following subsections will check the reliability ofour algorithm by theoretically analyzing the effects of rapidmoving of vehicles on the reliability of one-hop packettransmission and discussing the treatment of the failures ofRSUs.

4.2.1. Reliability of One-HopData Transmission. Aswe know,the vehicles at the edge of the transmission range of themessage sender may get away in a very short time due tothe fast movement of the vehicles. When the above situationoccurs in Dubhe, the furthest one-hop neighboring vehicleof the sender may not receive the message because theneighboring vehicle is no longer within the transmissionrange of the sender when the sender broadcasts the message.This subsection examines the possibility of such situationcaused by the rapid vehicle mobility, given the ideal wirelesssignal model (i.e., the unit disk model).

Figure 1 illustrates the distribution of vehicles on a 4-laneroad; the packets are propagated from right to left by vehicle𝑆. Note that whether the rebroadcasting vehicles could runaway from the sender or not is decided by the location of thecandidates when they send beacon to 𝑆, the relative velocitybetween the sender and receivers, and packet sending timeelapsed from the time of receiving beacons.

Supposing that the vehicle 𝐶0reports its position by a

beacon message at time 0 and the time that 𝑆 sends a packetis 𝑇

0, therefore, the condition that vehicle 𝐶

0can run away

from 𝑆 at 𝑇0is defined as

∫

𝑇0

0

(V𝑐0+ V

𝑠) 𝑡 𝑑𝑡 > 𝑚. (18)

Since the relative velocity between the vehicles in the oppositedirections is more likely to be greater than that of the samedirection and such situation has more chances to flee fromeach other, so we focus on the case of the opposite directions.

To begin with, we specify the probability that a candidatevehicle 𝐶

𝑖can escape from 𝑆 at time instant 𝑡:

𝑝𝑖(𝑡, V, 𝑙) = ∫

𝑅

0

𝑓 (𝑙) ∫

𝑡𝑏

0

𝑓 (𝑡) ∫

2Vmax

0

𝑓𝑟(V) V𝑡𝑑𝑙 𝑑𝑡 𝑑V, (19)

where 𝑓𝑟(V) is the probability density function (pdf) of the

relative velocity, 𝑓(𝑙) is the pdf of the distance between thecandidates and 𝑆 and 𝑓(𝑡) is the pdf of packet transmissiontime of 𝑆.

(1) Pdf of the relative velocity is as follows. We combinethe fluid model with the stochastic model to characterizethe average vehicle density and the random behavior ofindividual vehicle [22]. In light of the Greenshield model intraffic flow theory, as for a node at location 𝑙 at time 𝑡, therelationship between the velocity V(𝑙, 𝑡) of the node and thevehicle density is depicted as

V (𝑙, 𝑡) = V𝑓(1 −

𝑛 (𝑙, 𝑡)

𝑘𝑗

) , (20)

where V𝑓denotes the vehicle velocity of free flow, 𝑘

𝑗rep-

resents the jamming vehicle density and 𝑘𝑗= 1/𝑙

𝑐, in which

𝑙𝑐is the vehicle length, and 𝑛(𝑙, 𝑡) is the vehicle density of

location 𝑙 at time 𝑡. Taking the random behavior of vehiclesinto account, the vehicle velocity can be redefined as 𝑓(V) =V(𝑙, 𝑡) + 𝜇, in which 𝜇 is a random variable following normal


distribution𝑁(0, 𝜎2

). Since the maximum relative velocity ofthemoving vehicles in the opposite directions is 2Vmax and theadditive of two hormal distributions follows𝑁(0, 2𝜎

2

), hence,𝑓𝑟(V) is

𝑓𝑟(V) = 𝑓 (V

𝑐) + 𝑓 (V

𝑠)

= V𝑐(𝑙, 𝑡) + V

𝑠(𝑙, 𝑡) +

1

√4𝜋𝜎2exp(− V2

4𝜎2) .

(21)

(2) Pdf of vehicle distances is as follows. The exponentialdistribution is turned out to be appropriate for describingthe distances between the vehicles travelling along a 1-laneroad; therefore, the number of vehicles in the 1-lane roadcan be estimated by a Poisson distribution. However, vehiclesin Figure 1 are located in an approximate rectangle regionformed by the 4-lane road intersecting with the wirelesscoverage area of 𝑆. In view of this, we consider that thevehicles may be located randomly in the rectangle regionwhich is modelled by a 2-dimension Poisson distribution.To simplify the analysis, we turn to the distance distributionof the furthest vehicles 𝐶

0and 𝑆 in that 𝐶

0has maximum

opportunity to escape from 𝑆. So the cumulative distributionfunction (cdf) of distance distribution between 𝐶

0and 𝑆 can

be specified in the form

𝐹𝑐0(𝐿) = ∫

𝜋/2

0

𝑃 {𝑁 (𝐵0) = 0 | Θ = 𝜃} 𝑓

Θ(𝜃) 𝑑𝜃, (22)

where 𝐵0is the region without a vehicle, as marked with red

dotted lines in Figure 1. The angle 𝜃 is uniformly distributedin [0, 𝜋/2]; that is, 𝑓

Θ(𝜃) = 2/𝜋. Generally, the conditional

probability that 𝑖 vehicles locate in region 𝐵0conditioned on

a specific angle 𝜃 is

𝑃 (𝑁 (𝐵0) = 𝑖 | Θ = 𝜃) =

(𝜆𝐵0)𝑖

𝑖!𝑒−𝜆𝐵0 . (23)

Particularly, the area of 𝐵0can be calculated approxi-

mately with

𝐵0≈ 𝑤 (𝑅 − 𝐿 cos 𝜃) , (24)

where 𝑤 is the width of the road; hence (22) is rewritten as

𝐹𝑐0(𝑙) =

2

𝜋∫

𝜋/2

0

𝑒−𝜆𝑤(𝑅−𝐿 cos 𝜃)

𝑑𝜃. (25)

By differentiating (25), we obtain the pdf of distancebetween 𝐶

0and 𝑆:

𝑓𝑐0(𝑙) =

𝑑

𝐹𝐹𝑐0(𝑙) = −

2𝜆

𝜋∫

𝜋/2

0

𝑒−𝜆𝐵0

𝜕𝐵0

𝜕𝑙𝑑𝜃. (26)

(3) It is a reasonable assumption that the time 𝑆 sendspackets follows uniform distribution; thus,

𝑓 (𝑡) =1

𝑡𝑏

. (27)

C0

S

R

L

B0

𝜃

m

Figure 1: Vehicle distribution on the road.

5 10 15 20 250

0.01

0.02

0.03

0.04

0.05

0.06

0.07

Vehicle velocity (m/s)

Theo

retic

al lo

ss ra

tio

Beacon = 1 s : 10 vBeacon = 2 s : 10 v

Figure 2: Packet loss ratio of 10 vehicles.

Substituting (21) and (26)-(27) into (19), then the the-oretical probability that the vehicle 𝐶

0can escape from 𝑆

(i.e., packet loss probability 𝑝𝑖(𝑡, V, 𝑙)) can be calculated easily.

Figures 2 and 3 show the results under different conditions.It is observed that the packet loss ratio goes up with anincrease in beacon period, which is the direct result of theintegral of (27), (19). Moreover, note that increasing thevehicle density also incurs an increase in the packet loss ratioslightly, which is due to the fact that the more vehicles locatein the forwarding area of 𝑆, the smaller the area of 𝐵

0would

be, which results in an increase in the integral of (26).Next, we derive the success broadcast probability (𝑃succ)

[28] to justify the function of the multi-candidate strategy tomeet the challenge of the rapid movements of vehicles. Firstof all, recall that the vehicle number in the forwarding areaof 𝑆 follows Poisson distribution. Therefore, the probabilitythat there are 𝑗 potential vehicles that can take part inrebroadcasting the received packet can be given by

𝑃 {𝑁𝐶= 𝑗} =

(𝜆𝑅𝑤)𝑗

𝑗!𝑒−𝜆𝑅𝑤

. (28)


4 6 8 10 12 140

0.020.040.060.08

0.10.120.140.160.18

0.2

Vehicle velocity (m/s)

Theo

retic

al lo

ss ra

tio

Beacon = 1 s/40 vBeacon = 2 s/40 v

Figure 3: Packet loss ratio of 40 vehicles.

Further, assuming that at most 𝑀 candidates participate inforwarding a packet, the success probability can be foundfrom

𝑝(𝑀)

succ =𝑀−1

∑

𝑗=0

𝑃 {𝑁𝐶= 𝑗}(1 −

𝑗

∏

𝑖=0

𝑝𝑖(𝑡, V, 𝑙))

+ 𝑃 {𝑁𝐶≥ 𝑀}(1 −

𝑀

∏

𝑖=0

𝑝𝑖(𝑡, V, 𝑙)) .

(29)

The former part corresponds to the case that the number ofpotential candidate vehicles is less than 𝑀, while the latterpart reveals that the number of available forwarding vehiclesis more than 𝑀. However, only 𝑀 vehicles are involved inrebroadcasting the packet. By employing the complementaryrelationship, the probability that there is more than 𝑀

potential candidate vehicles can be easily given:

𝑃 {𝑁𝐶≥ 𝑀} = 1 −

𝑀−1

∑

𝑖=0

(𝜆𝑅𝑤)𝑖

𝑖!𝑒−𝜆𝑅𝑤

. (30)

Finally, the success broadcast probability can be expressedin two different forms as

𝑃succ = {𝑃 {𝑁

𝑐≥ 1} (1 − 𝑝

𝑖(𝑡, V, 𝑙)) 𝑀 = 0

𝑃 {𝑁𝑐≥ 1} 𝑝

(𝑀)

succ 𝑀 ≥ 1,(31)

where𝑀 = 0 implies that the packet is successfully transmit-ted without the help of the candidates, while 𝑀 ≥ 1 meansthat multiple candidate vehicles participate in forwarding thepacket. Obviously, the probability that at least one vehicle inthe forwarding area of 𝑆 is

𝑝 {𝑁𝑐≥ 1} = 1 − 𝑝 {𝑁

𝑐= 0} = 1 − 𝑒

−𝜆𝑤𝑅

. (32)

Combining (28)–(30) and (32) with (31), we can obtainthe success transmission probability easily. Given the param-eter values in Table 3, the analytical results are shown inFigures 4 and 5. As can be seen, the success transmission

5 10 15 20 250.8

0.820.840.860.88

0.90.920.940.960.98

1

Node velocity (m/s)

Theo

retic

al d

eliv

ery

ratio

Beacon = 1 s: no candidateBeacon = 1 s: one candidate


Figure 4: Delivery ratio of 10 vehicles.

4 6 8 10 12 140.7

0.75

0.8

0.85

0.9

0.95

1

Node velocity (m/s)

Theo

retic

al d

eliv

ery

ratio




Table 3: Parameter values used in analysis.

Parameter Value𝑅 200mRoad width 40mVehicle length 5m𝑇 20msMax velocity 25m/sMaximum process time 60ms

probability without the auxiliary vehicle decreases as thevehicle velocity arises. In comparison, the success probabilityincreases significantly while one auxiliary vehicle is involvedin the packet relay, which exactly justifies the necessity andeffectiveness of our strategy. In particular, with an increasein the vehicle density, the success transmission probabilitydeclines slightly for the same reason as the packet loss ratio.

4.3. Retransmission Ratio. Due to the relative movements ofvehicles, the vehicle 𝑆 may not receive the message broad-casted by 𝐶

0(𝑆 takes this message as an acknowledgment


message) and it will regard thatmessage transmitting fails andthen rebroadcast the packet, which leads to the result that𝑆 retransmits the packet. However, the theoretical analysisresult (Figure 6) shows that retransmission probability isacceptable under different vehicle densities. Apparently, theanalysis procedure is similar to that in Section 4.2.1 except theintegral upper limit of 𝑓(𝑡) in (19) is replaced with Ψ, and Ψ

is defined as

Ψ = 𝑡𝑠+ 𝜏

𝑐+ 𝛿

𝑡, (33)

𝑡𝑠= (𝑅 − 𝑙 cos 𝜃) ∗ 𝑇

𝑅, (34)

where 𝑡𝑠denotes the wait time before the packet is rebroad-

casted by 𝐶0, 𝜏𝑐represents the maximum process time in

𝐶0and 𝛿

𝑡is the one-hop wireless signal transmission time.

Equation (34) comes from line 19 of Algorithm 1; therefore,the retransmission probability can be obtained by rewriting(19) as follows:

𝑝𝑟𝑡(𝑡, V, 𝑙)

= ∫

𝑅

0

𝑓𝐿 𝑖(𝑙) ∫

𝜏𝑐+𝛿𝑡

0

𝑓 (𝑡)

× ∫

2Vmax

0

𝑓V𝑟 (V)V𝑟𝑡

𝑅 − 𝑙 cos 𝜃𝑑𝑙 𝑑𝑡 𝑑V

+ ∫

𝑅

0

𝑓𝐿 𝑖(𝑙) ∫

2Vmax

0

𝑓V𝑟 (V)V𝑟𝑡𝑠

𝑅 − 𝑙 cos 𝜃𝑑𝑙 𝑑V,

(35)

where the former integral indicates that the retransmissionprobability is mainly result of process time and the wirelesssignal transmission time given that the other parts of integralare known. The 𝑓(𝑡) can be depicted with a random variablewith uniform distribution; thus,

𝑓 (𝑡) =1

𝛿𝑡+ 𝜏

𝑐

. (36)

According to (34), the defer time of timer in 𝐶0(𝑡𝑠)

is an implicit random variable determined by the distancedistribution 𝑓(𝑙).

As can be seen from Figure 6, the retransmission prob-ability caused by vehicle movements is less than 1.2%, sothe message overheads incurred are very limited. Meanwhile,our theoretical results are expected to provide upper boundin that the parameter values configured correspond to therelative extreme case.

4.4. Failures of RSUs. The RSUs play important roles in pin-pointing the forwarding paths, so it is a necessary capabilityfor Dubhe to tolerate the failures of the RSUs. In Dubhe, themessages of arrival ratios from adjacent RSUs act as the heartbeating to maintain the neighboring relationship betweenRSUs. If the messages of arrival rates of a neighboring RSUhave not been received within a specified interval (in unitsof an exchanging cycle) on condition that there are vehiclescoming from the RSU, then the RSU is regarded as failed.When a vehicle finds that the approaching RSU is failed, it

4 6 8 10 12 14 16 18 200

0.2

0.4

0.6

0.8

1

1.2

1.4

Node velocity (m/s)

Theo

retic

al re

tran

smiss

ion

ratio

(%)

Vehicles = 10

Vehicles = 40

Figure 6: Theoretical retransmission ratio.

Figure 7: Road topology used in the experiments.

tries to forward the packet to the vehicle travelling in the samedirection. Otherwise, it selects a random path and tries toforward the packet.

5. Evaluation

We conduct simulation experiments to investigate the datadelivery performance and message overheads over threeindependent factors (i.e., the vehicle density, the trafficsignal cycle and the connectivity estimation period). Theexperimental results show that our mechanism can achievereliable and low-latency data delivery in comparison withthe epidemic-based data dissemination protocol (EBDD) [8]and the static-node-assisted adaptive data delivery protocol(SADV) [12].

5.1. Experimental Setup. The experiments are performed ona 6000 × 4000m area, where the street layout is derivedand normalized from a real map in Google Maps. In theexperiments, roads are two-lane and 35 RSUs are deployed atmajor intersections. Figure 7 shows the road topology, wherethe black dots indicate the deployed RSUs.

We use NCTUns 6.0 [29] to conduct experiments, andthe MAC protocol follows IEEE 802.11b. In order to simulateroadside buildings in urban environments, barrier walls areset along the roadsides. The Intelligent Driver Model (IDM)is chosen for vehicles as their mobility model to reflect thefeatures of vehicle movements in urban scenarios.


Table 4: Experimental parameters.

Parameter ValuePropagation model Two-ray ground refection modelExpected transmission radius 200mVehicle velocity 0∼72KM/hAverage length of road segments 1 KmNumber of vehicle 100–1035Traffic signal period 40–120 sConnectivity estimation period 5–20 sBeacon message interval 1 sTTL (time to live) of a packet 1200 sLength of beacon messages 76 bytesLength of data packets 68–418 bytes

In the experiments, the initial positions of all vehicles(100–1035 vehicles) are randomly placed, and vehicles maketurns at the intersections with a certain probability (probabil-ities of going straight, turning left and turning right are 0.6,0.2, and 0.2, resp.).The green phase of traffic signal equals thered phase, with 50 s as default. Each simulation lasts for onehour.

Data are generated as follows: 10 percent of vehiclesgenerate a data request with a random destination (SADVand Dubhe use RSUs as destinations while EBDD usesvehicles as destinations) per second. Table 4 outlines theexperimental parameters.

5.2. Results and Analyses. We conduct three groups ofexperiments to evaluate the influence of the traffic densi-ties, the traffic signal cycle and the connectivity estimationperiod/delay update period on the performance and over-heads.

(1) Vehicle Density. The vehicle density is one of the mostimportant factors affecting the connectivity of a VANET.Thefirst group of experiments examines the performance andoverheads under different traffic densities. Here, the numberof vehicles is employed to represent the vehicle density, whichis increased from 100 to 1035 in experiments. As shown inFigure 8, the delivery ratios grow substantially as the numberof vehicles increases. Note that EBDD is more sensitiveto the vehicle density than Dubhe and SADV. The packetdelivery ratio of EBDD reaches its peak at 635 vehicles andthen declines as the vehicle density further increases. Thebehaviors of EBDD are rooted in the fact that it works ina manner of best effort, that is, the packets are exchangedwhen the vehicles enter the transmission range of each other,which leads to severe channel conflicts under heavy traffic. Incontrast, the delivery ratios of SADV and Dubhe show steadyimprovement as the vehicle density increases. Furthermore,Dubhe shows a significant advantage over SADV in terms oftransmission delay and delivery ratio. It is worth mentioningthat the packet loss in Dubhe is mainly due to expiration ofTTL values of packets.

135 235 335 435 535 635 735 835 935 103550

55

60

65

70

75

80

85

90

95

100

Number of nodes

Deli

very

ratio

(%)

EBDDSADVDubhe

Figure 8: Data delivery ratio.

135 235 335 435 535 635 735 835 935 1035200

250

300

350

400

450

500

550

600

650

700

Number of nodes

Del

ay (s

)

EBDDSADVDubhe

Figure 9: Transmission delay.

Figure 9 indicates that the delay of EBDD is lower thanthat of SADVandDubhewhen the vehicle density is relativelylow (less than 635 vehicles) but grows gradually as the vehiclenumber continues to increase. Benefiting from exchangingthe transmitting delay between adjacent intersections, thetransmission delay of SADV and Dubhe cuts down rapidlywith the increase of the vehicle density. But the delay ofDubhe is shorter than that of SADV in that Dubhe employs amore accurate delaymodel than SADV for packet forwardingdecisions.

Figure 10 shows the experimental results of messageoverheads. Here, as for a vehicle, its message overhead is


135 235 335 435 535 635 735 835 935 10350

0.5

1

1.5

2

2.5

3

Number of nodes

Mes

sage

ove

rhea

d (103M

B)

EBDDSADVDubhe

Figure 10: Message overhead.

defined as the message traffic that is received and sent exceptpayload of data packets. It can be observed that SADVconsumes fewermessages thanDubhe and EBDD because nobeacon messages are sent in SADV. The message overheadsof EBDD increase dramatically as the growth of vehiclenumber, which results from the exchanging of packet digestlists when the vehicles can communicate with each other,and the lengths of the lists rise sharply as the vehicle densityincreases. However, the overheads of Dubhe remain relativelystable, and the majority of messages of Dubhe come fromthe beacon messages. Considering that the beacon is anessential mechanism for safety-related applications, we claimthat Dubhe does not increase substantial message overheadsin comparison with SADV.

(2) Traffic Signal Cycle Length. The traffic signals at inter-sections interrupt the traffic flow and affect the VANETconnectivity significantly. Therefore, the second group ofexperiments investigates the influence of traffic signal cyclelength on the performance of data delivery. Figures 11–16show the results of the delivery ratios, the transmission delay,and the overheads under distinct traffic signal cycle lengthsfor the scenarios of different vehicle number, (200 vehiclesand 1000 vehicles, resp.).

As shown in Figure 11, the delivery ratio of SADV andDubhe decreases about 5% and the delay rises about 10% asthe length of the traffic signal cycle increases from40 s to 120 sin the 200-vehicle scenario, while the delivery ratio of EBDDboosts about 10%,whichmay arise from the fact that in EBDDthe long traffic signal cycle providesmore chances for vehiclesqueuing at intersections to infect mutually.

From Figures 12 and 14, in the 1000-vehicle scenario,EBDD is significantly inferior to SADV and Dubhe in terms

40 60 80 100 12040

50

60

70

80

90

100

Traffic light cycle (s)

Deli

very

ratio

(%)

Low-density flow-200

EBDDSADVDubhe


40 60 80 100 12040

50

60

70

80

90

100


Deli

very

ratio

(%)

High-density flow-1000

EBDDSADVDubhe


of the packet delivery ratio and the transmission delay.Meanwhile, the transmission delay of Dubhe is minimumwhen the traffic signal cycle lasts 80 seconds, which is 8%higher than the delay with 40 seconds as the cycle length.Moreover, the transmission delay of SADV is 20% longerthan that of Dubhe as a whole. The reason behind this is thatthe estimation of transmitting delay in SADV is concernedlittle about the interference of traffic signals while the delaymodel of Dubhe distinguishes between the required time forestablishing awireless link during the red phase and the green


40 60 80 100 120300

350

400

450

500

550

600

650

700


Del

ay (s

)


EBDDSADVDubhe

Figure 13: Transmission delay of 200 vehicles.

40 60 80 100 120200

250

300

350

400

450


Del

ay (s

)

High-density flow-1000

EBDDSADVDubhe


phase, togetherwith the consideration of the changing vehicledensity.

What is worth noting is that short traffic signal cyclescorrespond to high transmission delay when the vehicledensity is high. This phenomenon can be interpreted asthe multi-hop wireless link between adjacent intersectionscannot be established during the short green phase, whichincreases the proportion of the transmission time of thepacket carried by vehicles.

In addition, the data in Figures 15 and 16 imply thatthe traffic signal cycles have little impact on the message

40 60 80 100 1200

0.1

0.2

0.3

0.4

0.5



Mes

sage

ove

rhea

d (103M

B)

EBDDSADVDubhe

Figure 15: Overhead of 200 vehicles.

40 60 80 100 1200

0.5

1

1.5

2

2.5

3High-density flow-1000


Mes

sage

ove

rhea

d (103M

B)

EBDDSADVDubhe


overheads of SADV and Dubhe. However, the messagesconsumed by EBDD are more sensitive to the length of thetraffic signal cycle especially in the 1000-vehicle scenariowhere the number of messages rises about 10% at most withthe increase of the length of the traffic signal cycle.

(3) Connectivity Estimation Period/Delay Update Period. TheRSUs deployed at intersections contribute to store the packetswhen no suitable vehicles can relay the packets. In Dubhe,an RSU estimates the change of queue lengths of vehicles atintersections with the period 𝜏, here 𝜏; is called a connectivity


5 10 15 2050

55

60

65

70

75

80

Period of estimation/update (s)

Deli

very

ratio

(%)

SADV-200Dubhe-200


5 10 15 2070

75

80

85

90

95

100


Deli

very

ratio

(%)

SADV-1000Dubhe-1000


estimation period. In SADV, the static nodes at adjacent inter-sections periodically exchange the data transmitting delay onroads, which is utilized to decide the packet forwarding paths.Because connectivity estimation period 𝜏 and delay updateperiod 𝜂 of static nodes have similar function, we design agroup of experiments to examine their effects on the datadelivery performance.

As shown in Figures 17 and 18, with the increase of 𝜏 and𝜂, both delivery ratios decrease slightly, that is, 1% for Dubheand 2% for SADV. Meanwhile, we find that the parameter𝜏 in Dubhe shows more impact on the transmission delaythan on the delivery ratio. That is, from Figures 19 and 20,

5 10 15 20550

600

650

700

Del

ay (s

)

SADV-200Dubhe-200



5 10 15 20200

220

240

260

280

300

320

340

360

Del

ay (s

)

SADV-1000Dubhe-1000



the transmission delay increases as 𝜏 gets longer. In detail,when 𝜏 changes from 5 s to 20 s, the transmission delay growsabout 5.6% in the 1000-vehicle scenario while being 2% or soin the 200-vehicle scenario. The increased delay is due to thedeteriorated accuracy of connectivity estimation.

Similar to 𝜏, as 𝜂 becomes longer, the transmission delayof SADV goes up about 9% in the 1000-vehicle scenariowhile being 4% in the low-density case. In comparison withDubhe, the parameter 𝜂 has more effect on performance ofSADV than 𝜏 on Dubhe. The reasons are twofold: on the onehand, the error rates of the actual delay information go up as𝜂 increases. On the other hand, Dubhe can forecast the lost


5 10 15 200

0.05

0.1

SADV-200Dubhe-200

Mes

sage

ove

rhea

d (103M

B)



5 10 15 200

0.5

1

1.5

2

2.5

3

SADV-1000Dubhe-1000


Mes

sage

ove

rhea

d (103M

B)


arrival ratios of vehicles at neighboring intersections fromhistorical data while SADV has no such remedy strategy.

No significant changes in the message overheads areobserved in the experiments of different traffic densities, asshown in Figures 21 and 22. Since the examined parameters𝜏 and 𝜂 are only related to the frequency of exchanging thearrival ratios and exchanging the actual delay information,the messages incurred are negligible to the total overheads.

6. Conclusions

The efficient and reliable data delivery lays a foundationfor many VANET applications. We present Dubhe, whichreduces the data transmission delay and improves the data

delivery reliability of long-distance data transmission inurban scenarios. In Dubhe, the RSUs deployed at mainintersections are exploited to collect the traffic densitiesand estimate the real-time transmitting delay between theRSUs and then choose a minimum delay path for thepackets.Meanwhile, an improved greedy broadcast algorithmis proposed to enhance the reliability of one-hop datatransmission; then we theoretically analyze the reliability andthe retransmission ratio. The experimental results show thatthe packet delivery ratio of Dubhe is 10% higher than ofSADV and transmitting delay is 5–12% lower than that ofSADV.Moreover, in comparisonwith EBDD, ourmechanismexhibits stable performance, low message overheads, andhigh scalability.

Acknowledgment

This work was supported by the National Natural ScienceFoundation of China under Grants nos. 60970027 and91124001.

References

[1] D. Abusch-Magder, P. Bosch, T. E. Klein, P. A. Polakos, L. G.Samuel, and H. Viswanathan, “911 -NOW: a network on wheelsfor emergency response and disaster recovery operations,” BellLabs Technical Journal, vol. 11, no. 4, pp. 113–133, 2007.

[2] N. Wisitpongphan, F. Bai, P. Mudalige, V. Sadekar, and O.Tonguz, “Routing in sparse vehicular ad hoc wireless networks,”IEEE Journal on Selected Areas in Communications, vol. 25, no.8, pp. 1538–1556, 2007.

[3] N. Wisitpongphan, O. K. Tonguz, J. S. Parikh, P. Mudalige, F.Bai, and V. Sadekar, “Broadcast storm mitigation techniques invehicular ad hoc networks,” IEEEWireless Communications, vol.14, no. 6, pp. 84–94, 2007.

[4] J. Nzouonta, N. Rajgure, G. Wang, and C. Borcea, “VANETrouting on city roads using real-time vehicular traffic informa-tion,” IEEE Transactions on Vehicular Technology, vol. 58, no. 7,pp. 3609–3626, 2009.

[5] K. Shafiee and V. C. M. Leung, “Connectivity-aware minimum-delay geographic routing with vehicle tracking in VANETs,” AdHoc Networks, vol. 9, no. 2, pp. 131–141, 2011.

[6] J. Zhao and G. Cao, “VADD: vehicle-assisted data delivery invehicular Ad hoc networks,” IEEE Transactions on VehicularTechnology, vol. 57, no. 3, pp. 1910–1922, 2008.

[7] J. J. Blum, A. Eskandarian, and L. J. Huffman, “Challenges ofintervehicle Ad Hoc networks,” IEEE Transactions on IntelligentTransportation Systems, vol. 5, no. 4, pp. 347–351, 2004.

[8] A. Vahdat and B. David, “Epidemic routing for partiallyconnected Ad Hoc networks,” Duke Tech Report CS-2000-06,2000.

[9] P. Ruiz, B. Dorronsoro, P. Bouvry, and L. Tardon, “Informationdissemination in VANETs based upon a tree topology,” Ad HocNetworks, vol. 10, no. 1, pp. 111–127, 2012.

[10] F. J. Ros, P. M. Ruiz, and I. Stojmenovic, “Acknowledgment-based broadcast protocol for reliable and efficient data dissem-ination in vehicular ad hoc networks,” IEEE Transactions onMobile Computing, vol. 11, no. 1, pp. 33–46, 2012.

[11] H. Xu and M. Barth, “An adaptive dissemination mechanismfor inter-vehicle communication-based decentralized traffic


information systems,” in Proceedings of the IEEE IntelligentTransportation Systems Conference (ITSC ’06), pp. 1207–1213,Toronto, Canada, September 2006.

[12] Y. Ding and L. Xiao, “SADV: static-node-assisted adaptive datadissemination in vehicular networks,” IEEE Transactions onVehicular Technology, vol. 59, no. 5, pp. 2445–2455, 2010.

[13] H. Zhao and J. Q. Zhu, “Efficient data dissemination in urbanVANETs: parked vehicles are natural infrastructures,” Interna-tional Journal of Distributed Sensor Networks, vol. 2012, ArticleID 151795, 11 pages, 2012.

[14] M. Morales, R. Haw, E. J. Cho, C. S. Hong, and S. Lee,“An adaptable destination-based dissemination algorithmusinga publish/subscribe model in vehicular network,” Journal ofComputing Science and Engineering, vol. 6, no. 3, pp. 227–242,2012.

[15] P. Fazio, F. D. Rango, and C. Sottile, “An on Demand Inter-ference Aware Routing Protocol for VANETS,” Journal ofNetworks, vol. 7, pp. 1728–1738, 2012.

[16] F. de Rango, F. Veltri, P. Fazio, and S. Marano, “Two-leveltrajectory-based routing protocol for vehicular ad hoc networksin freeway and Manhattan environments,” Journal of Networks,vol. 4, no. 9, pp. 866–880, 2009.

[17] G. Korkmaz, F. Ozguner, E. Ekici, and U. Ozguner, “Urbanmulti-hop broadcast protocol for inter-vehicle communicationsystems,” in Proceedings of the 1st ACM International Workshopon Vehicular Ad Hoc Networks, Held in Conjunction withMOBICOM (VANET ’04), pp. 76–85, October 2004.

[18] C. Lochert, B. Scheuermann, M. Caliskan, and M. Mauve,“The feasibility of information dissemination in vehicular ad-hoc networks,” in Proceedings of the 4th Annual Conference onWireless on Demand Network Systems and Services (WONS ’07),pp. 92–99, Obergurgl, Austria, January 2007.

[19] W. Viriyasitavat, F. Bai, and O. K. Tonguz, “Dynamics of net-work connectivity in urban vehicular networks,” IEEE Journalon Selected Areas in Communications, vol. 29, no. 3, pp. 515–533,2011.

[20] M. Khabazian and M. K. M. Ali, “A performance modeling ofconnectivity in vehicular Ad Hoc networks,” IEEE Transactionson Vehicular Technology, vol. 57, no. 4, pp. 2440–2450, 2008.

[21] J. Nzouonta, M. K. Nakayama, and C. Borcea, “On derivingand incorporatingmultihop path duration estimates in VANETprotocols,” ACM Transactions on Modeling and Computer Sim-ulation, vol. 21, no. 2, article 14, 2011.

[22] I.W.-H.Ho, K. K. Leung, and J.W. Polak, “Stochasticmodel andconnectivity dynamics for VANETs in signalized road systems,”IEEE/ACM Transactions on Networking, vol. 19, no. 1, pp. 195–208, 2011.

[23] M. Fiore and J. Harri, “The networking shape of vehicularmobility,” in Proceedings of the 9th ACM International Sympo-sium on Mobile Ad Hoc Networking and Computing (MobiHoc’08), pp. 261–271, Hong Kong, May 2008.

[24] B. L. Smith, B. M. Williams, and R. Keith Oswald, “Compar-ison of parametric and nonparametric models for traffic flowforecasting,” Transportation Research Part C, vol. 10, no. 4, pp.303–321, 2002.

[25] B.M.Williams andL.A.Hoel, “Modeling and forecasting vehic-ular traffic flow as a seasonal ARIMA process: theoretical basisand empirical results,” Journal of Transportation Engineering,vol. 129, no. 6, pp. 664–672, 2003.

[26] P. S. OLszewski, “Modelling of queue probability distribution attraffic signals,” in Proceedings of the 11th International Sympo-sium, Transportation and TrafficTheory, pp. 569–588, 1990.

[27] T. Fujiki, M. Kirimura, T. Umedu, and T. Higashino, “Efficientacquisition of local traffic information using inter-vehicle com-munication with queries,” in Proceedings of the 10th Interna-tional IEEE Conference on Intelligent Transportation Systems(ITSC ’07), pp. 241–246, Seattle, Wash, USA, October 2007.

[28] S. Yang, C. K. Yeo, and B.-S. Lee, “Toward reliable data deliveryfor highly dynamicmobile ad hoc networks,” IEEE Transactionson Mobile Computing, vol. 11, no. 1, pp. 111–124, 2012.

[29] S. Y. Wang and C. C. Lin, “NCTUns 6. 0: a simulator foradvanced wireless vehicular network research,” in Proceedingsof the IEEE 71st Vehicular Technology Conference (VTC ’10), vol.2010, pp. 1–2, 2010.

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttp://www.hindawi.com Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014


Active and Passive Electronic Components

Control Scienceand Engineering

Journal of



RotatingMachinery


Hindawi Publishing Corporation http://www.hindawi.com

Journal ofEngineeringVolume 2014

Submit your manuscripts athttp://www.hindawi.com

VLSI Design



Shock and Vibration


Civil EngineeringAdvances in

Acoustics and VibrationAdvances in



Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation http://www.hindawi.com

Volume 2014

The Scientific World JournalHindawi Publishing Corporation http://www.hindawi.com Volume 2014

SensorsJournal of


Modelling & Simulation in EngineeringHindawi Publishing Corporation http://www.hindawi.com Volume 2014


Chemical EngineeringInternational Journal of Antennas and

Propagation




Navigation and Observation



DistributedSensor Networks


Documents

Research Article Dubhe: A Reliable and Low-Latency Data ...downloads.hindawi.com/journals/ijdsn/2013/581821.pdf · Research Article Dubhe: A Reliable and Low-Latency Data Dissemination