8/12/2019 Hassan TMC 12

1/14

1

VeMAC: A TDMA-based MAC Protocol forReliable Broadcast in VANETs

Hassan Aboubakr Omar,Student Member, IEEE,Weihua Zhuang,Fellow, IEEE,and Li Li, Member, IEEE

AbstractThe need of a medium access control (MAC) protocol for an efficient broadcast ser vice is of great importance to support

the high priority safety applications in vehicular ad hoc networks (VANETs). This paper introduces VeMAC, a novel multichannel TDMA

MAC protocol proposed specifically for a VANET scenario. The VeMAC supports efficient one-hop and multi-hop broadcast services on

the control channel by using implicit acknowledgments and eliminating the hidden terminal problem. The protocol reduces transmission

collisions due to node mobility on the control channel by assigning disjoint sets of time slots to vehicles moving in opposite directions

and to road side units. Analysis and simulation results in highway and city scenarios are presented to evaluate the performance of

VeMAC and compare it with ADHOC MAC, an existing TDMA MAC protocol for VANETs. It is shown that, due to its ability to decrease

the rate of transmission collisions, the VeMAC protocol can provide significantly higher throughput on the control channel than ADHOC

MAC.

Index TermsTDMA, medium access control, reliable broadcast, and vehicular ad hoc networks.

1 INTRODUCTION

AN ad hoc network is defined as a collection ofnodes dynamically forming a network withoutany existing infrastructure or centralized administration.One special type of mobile ad hoc networks is thenetwork among moving vehicles, which is known asvehicular ad hoc network (VANET). A VANET consistsof a set of vehicles equipped with a communicationdevice, called on-board unit (OBU), and a set of sta-tionary units along the roads, called road side units(RSUs), which can be connected together and/or to the

Internet via wireless or wireline links. Each OBU hasa radio interface to connect to other OBUs and RSUs,as well as a wireless or wired interface to which anapplication unit can be attached. The main objectivesof VANETs are to provide efficient vehicle-to-vehicle(V2V) and vehicle-to-RSU (V2R) communications. Basedon these two kinds of communications, VANETs cansupport many applications in safety, entertainment, andvehicle traffic optimization [2], [3]. Motivated by theimportance of vehicular communications, the UnitedStates Federal Communication Commission (FCC) hasallocated 75MHz radio spectrum in the 5.9GHz band forDedicated Short Range Communications (DSRC) to beexclusively used by V2V and V2R communications. TheDSRC spectrum is divided into seven 10MHz channels:

H. A. Omar and W. Zhuang are with the Center for Wireless Com-munications, Department of Electrical and Computer Engineering, Uni-versity of Waterloo, 200 University Avenue West, Waterloo, Ontario,Canada, N2L 3G1.E-mail: {h3omar,wzhuang}@bbcr.uwaterloo.ca

L. Li is with the Communications Research Center, Ottawa, Ontario,Canada, K2H 8S2.E-mail: li.li@crc.ca.

This work is presented in part at the 2011 IEEE INFOCOM Workshop [1].This work was supported by a research grant from the Natural Science andEngineering Research Council (NSERC) of Canada.

six service channels for safety and non-safety relatedapplications, and one control channel for transmissionof control information and high priority short safetymessages.

Most (if not all) of the high priority safety applicationsproposed for VANETs are based on one-hop broadcastof information. For instance, for V2V communication-

based applications such as the pre-crash sensing, blindspot warning, emergency electronic brake light, andcooperative forward collision avoidance, each vehicleperiodically broadcasts information about its position,

speed, heading, acceleration, turn signal status, etc, toall the vehicles within its one-hop neighbourhood [2].Similarly, for V2R communication-based applications,such as the curve speed warning and traffic signal viola-tion warning, an RSU periodically broadcasts to all theapproaching vehicles information related to the trafficsignal status and timing, road surface type, weather con-ditions, etc [2]. As the precision of the safety applicationsis directly related to the safety of people on road, theneed of a medium access control (MAC) protocol whichprovides an efficient broadcast1 service is crucial forVANETs.

Various MAC protocols have been proposed for

VANETs, based either on IEEE 802.11 or on channeliza-tion such as time division multiple access (TDMA), spacedivision multiple access (SDMA), and code division mul-tiple access (CDMA). In SDMA schemes, each vehicledecides whether or not it is allowed to access the channel

based on its location on the road [4], [5]. An SDMAscheme consists of three main parts: adiscretization proce-durewhich divides the road into small areas called cells,a mapping function which assigns to each of the cells a

1. When broadcast is mentioned solely, it refers to one-hop broad-cast.

8/12/2019 Hassan TMC 12

2/14

2

unique time slot, and an assignment rule which specifieswhich time slots a vehicle is allowed to access based onthe cell where it is currently located. Similarly, CDMAis proposed for MAC in VANETs due to its robustnessagainst interference and noise [6], [7]. The main problemwhich arises with CDMA in VANETs is how to allocatethe pseudo noise (PN) codes to different vehicles. Due toa large number of vehicles, if every vehicle is assigned aunique PN code, the length of these codes will becomeextremely long, and the required bit rates for VANETapplications may not be attained. Consequently, it ismandatory that the PN codes be shared among differentvehicles in a dynamic and fully distributed way [7]. Onthe other hand, the IEEE 802.11p is a recently proposedMAC standard for VANETs [8]. The protocol is basedon the legacy IEEE 802.11 standard [9] which is widelyimplemented, but does not provide an efficient broadcastservice. The reason is that, for broadcast frames, noRTS/CTS exchange is used and no acknowledgementis transmitted from any of the recipient of the frame

[9]. This lack of RTS/CTS exchange results in a hiddenterminal problem which reduces the frame delivery ratioof the broadcast service, especially with the absenceof acknowledgement frames [10]. Another limitation isthat, in a VANET scenario, by employing the enhanceddistributed channel access (EDCA) scheme defined in theIEEE 802.11 standard, the high priority safety messageswill be assigned to the high priority access categories(ACs) which contend for the wireless channel using asmall contention window size [9]. Although this smallcontention window size allows the high priority safetyframes to be transmitted with small delays, it increasesthe probability of transmission collisions when multi-

ple nodes within the same communication range aresimultaneously trying to broadcast their safety messages[11]. Moreover, unlike the unicast case, the size of thecontention window is not doubled when a collisionhappens among the broadcasted safety messages sincethere is no collision detection for the broadcast servicewithout CTS and acknowledgment frames [9]. Differ-ent from the contention-based IEEE 802.11p standard,the ADHOC MAC protocol is based on TDMA and isproposed for inter-vehicle communication networks [12].The ADHOC MAC protocol operates in a time slottedstructure, where time slots are grouped into virtualframes, i.e. no frame alignment is needed. By letting each

node report the status of all the time slots in the previous(sliding) virtual frame, the ADHOC MAC can support areliable2 broadcast service without the hidden terminalproblem [12]. As well, the ADHOC MAC provides amulti-hop broadcast service which can cover the wholenetwork using a significantly smaller number of relayingnodes than that using a flooding procedure. Moreover,in ADHOC MAC, each node is guaranteed to accessthe channel at least once in each virtual frame, whichis suitable for non delay-tolerant applications. However,

2. In this paper, the term reliable means acknowledged.

simulation results show that, due to node mobility, thethroughput reduction can reach 30% for an averagevehicle speed of 50km/h [13]. Another major limitationof ADHOC MAC is that it is a single channel protocol,not suitable for the seven DSRC channels.

This paper presents VeMAC, a novel multichannelTDMA protocol developed based on ADHOC MAC [12]and designed specifically for VANETs. On the controlchannel, the protocol provides a reliable2 one-hop broad-cast service without the hidden terminal problem as wellas an efficient multi-hop broadcast service to disseminateinformation all over the network. The VeMAC assignsdisjoint sets of time slots to vehicles moving in oppositedirections and to RSUs, and hence can decrease the rateof transmission collision on the control channel caused

by node mobility. As well, the VeMAC employs newtechniques for the nodes to access the available time slotsand to detect transmission collisions. These techniquesare different from the ones used by ADHOC MAC,which have some limitations as to be discussed in details.

It is shown that the proposed VeMAC protocol providessignificantly higher throughput on the control channelthan that of ADHOC MAC and ADHOC-enhanced (anenhanced version of ADHOC MAC introduced in thispaper).

The rest of this paper is organized as follows. Section 2describes the system model and Section 3 presents theVeMAC protocol. The performance of the VeMAC pro-tocol on the control channel is analyzed in Section 4.Section 5 presents the simulation results and explainssome limitations of the ADHOC MAC protocol. A dis-cussion on the main features of the VeMAC protocolin comparison with ADHOC MAC and some possible

VeMAC extensions are presented in Section 6. Finally,Section 7 concludes this research and suggests somefuture works.

2 SYSTEMM ODEL

The VANET under consideration consists of a set ofRSUs and a set of vehicles moving in opposite directionson two-way vehicle traffic roads. A vehicle is said to bemoving in a left (right) direction if it is currently headingto any direction from north/south to west (east), asshown in Fig.1. Based on this definition, if two vehiclesare moving in opposite directions on a two-way road,

it is guaranteed that one vehicle is moving in a leftdirection while the other vehicle is moving in a rightone.

The VANET has one control channel, denoted byc0, and M service channels, denoted by c1, c2, . . . , cM.Channel c0 is used for transmission of two kinds ofinformation: high priority short applications (such asperiodic or event driven safety messages), and controlinformation required for the nodes to determine whichtime slots they should access on channel ci, i= 0, . . . , M .The M service channels are used for transmission ofsafety or non-safety related application messages. A

8/12/2019 Hassan TMC 12

3/14

3

Fig. 1: Right and left directions of vehicle movement.

Fig. 2: Partitioning of each frame on channel c0 into L,Rand F sets.

provider is a node which announces on channel c0 fora service offered on a specific service channel, while auser is a node which receives the announcement for aservice and decides to make use of this service. Eachnode has two transceivers: transceiver1 is always tunedto channel c0, while transceiver2 can be tuned to anyservice channel ci, i = 1, . . . , M . It is assumed that thetransmission power levels on all channels are fixed andknown to all nodes. All channels are symmetric, in thesense that nodex is in the communication range of nodey if and only if node y is in the communication rangeof node x. Each node is identified by a MAC addressas well as a short identifier (ID). The ID is chosen byeach node at random, included in the header of eachpacket transmitted on channel c0, and changed if thenode detects that its ID is already in use by anothernode [12].

Time is partitioned to frames consisting of a constantnumber of fixed duration time slots. The number oftime slots per frame on channel cm is denoted by sm,m= 0, . . . , M , and a time slot on channelcmis identified

by the index of this time slot within a frame on channelcm3. On channel c0, each frame is partitioned into threesets of time slots: L, R, and F, as shown in Fig. 2.

The F set is associated with RSUs, while the L andR sets are associated with vehicles moving in left andright directions respectively. Every node (i.e. vehicle orRSU) is equipped with a global positioning system (GPS)receiver and can accurately determine its position andmoving direction using GPS. The current position ofeach node is included in the header of each packettransmitted on channel c0, and synchronization amongnodes is performed using the 1PPS signal provided byany GPS receiver. The rising edge of this 1PPS is aligned

3. Note that, the same time slot can have different indices onchannels ci and cj , i =j , since si is not necessarily equal to sj .

with the start of every GPS second with accuracy within100ns even for inexpensive GPS receivers. Consequently,this accurate 1PPS signal can be used as a commontime reference among all the nodes. All the channelsare slot synchronized and, on each channel, each secondcontains an integer number of frames as shown in Fig. 2for channel c0. Hence, at any instant, each node candetermine the index of the current slot within a frame onany channelcm,m= 0, . . . , M , and whether it belongs totheL, R, orFset on channel c0. In case of a temporaryloss of GPS signal, the synchronization among differentnodes can still be maintained within a certain accuracyfor a time duration which depends on the stability of theGPS receivers local oscillator at each node [14]. If theGPS signal is lost in a certain area for a long duration(longer than a specified threshold), a distributed syn-chronization scheme, such as the one presented in [14],should be employed until the GPS signal is recovered.Details of such a back up synchronization scheme areout of scope of this paper.

For a certain node x, the following sets are defined: N(x): the set of IDs of the one-hop neighbours

of node x on channel c0, from which node x hasreceived packets on channel c0 in the previous s0slots;

Tm(x): the set of time slots that node x mustnot use on channel cm in the next sm timeslots, m = 0, . . . , M .

Set Tm(x) is used by node x to determine which timeslots it can access on channel cm without causing anyhidden terminal problem. Constructing and updating setT0(x) is different from sets Tm(x), m = 1, . . . , M , asdescribed in Section 3.

3 VEMAC PROTOCOL

3.1 VeMAC Preliminaries

In the VeMAC protocol, each node must acquire exactlyone time slot in a frame on channel c0. Once a nodeacquires a time slot, it keeps accessing the same slot in allsubsequent frames on channel c0 unless a transmissioncollision is detected (as to be explained in details inSubsection 3.2). Each packet4 transmitted on channel c0is divided into four main fields: header, announcementof services(AnS), acceptance of services (AcS), and highpriority short applications, as shown in Fig. 3.

Header AnS AcS High priority short applications

Fig. 3: Format of each packet transmitted on channel c0.

Each node must transmit a packet during its time sloteven if the node has no data to include in the highpriority short applications field. The reason is that, infor-mation in the header,AnSandAcSfields, is necessary for

4. In the rest of the paper, the term packet is used instead of frameto refer to MAC layer Protocol Data Unit (MPDU), i.e. the unit of dataexchanged between two peer MAC entities. It is to avoid confusionwith the frame which is a collection of time slots.

8/12/2019 Hassan TMC 12

4/14

4

Fig. 4: Merging collision caused by node mobility.

other nodes to decide which time slots they can accesson the control channel and service channels as to bedescribed in Subsections 3.2 and 3.3.

Two types of transmission collision on time slots canhappen on channel c0 [13]: access collision and mergingcollision. An access collision happens when two or morenodes within two hops of each other attempt to acquirethe same available time slot. On the other hand, amerging collision happens when two or more nodesacquiring the same time slot become members of the

same two-hop set5

(THS) due to node activation or nodemobility. The difference between the two types of colli-sion is that access collisions occur among nodes whichare trying to acquire a time slot, while merging collisionsoccur among nodes which have successfully acquired atime slot. In VANETs, although merging collisions canhappen among vehicles moving in the same directiondue to acceleration or deceleration, it is more likelyto occur among vehicles moving in opposite directions(approaching each other) or between a vehicle and astationary RSU since they approach each other with amuch higher relative velocity as compared to vehiclesmoving in the same direction. For example, in Fig. 4, if

vehiclex moves to THS2 and ifx is using the same timeslot as z, then collision will occur at y. Upon detectionof a merging collision on channel c0, each colliding nodeshould release its time slot and acquire a new one, whichmay generate more access collisions.

3.2 Accessing Slots on the Control Channel

For the purpose of time slot assignment on channel c0,in the header of each packet transmitted on channel c0,the transmitting node y should include setN(y)and thetime slot used by each node z N(y). The short IDs inset N(y)serve to decrease the overhead as compared to

including the MAC address of each one-hop neighbourin the header of each transmitted packet. Suppose nodex is just powered on and needs to acquire a time slot.It starts listening to channel c0 for s0 successive timeslots (not necessarily in the same frame). At the end ofthe s0 slots, node x can determine N(x) and the timeslot used by each node i N(x). In addition, since eachnodei N(x)announcesN(i)and the time slot used byeach node j N(i), node x can determine the time slotused by each of its two-hop neighbours, j N(i), j /

5. A two-hop set is a set of nodes in which each node can reach anyother node in two hops at most.

N(x), i N(x). Accordingly, node x sets T0(x) to theset of time slots used by all nodes within its two-hopneighbourhood. Then, sets N(x) and T0(x) are updated

by node x at the end of each time slot (always based oninformation received in the previous s0 slots).

Given T0(x), node x determines the set of accessibletime slots A(x) (to be discussed) and then attempts toacquire a time slot by randomly accessing any time slotin A(x), say time slot k. If no other node in the two-hop neighbourhood of node x simultaneously attemptsto acquire time slot k, then no access collision happens.In this case, the attempt of node x is successful and eachone-hop neighbour i of node x adds node x to the setN(i) and records that node x is using time slot k. Onthe other hand, if at least one node within the two-hopneighbourhood of node x accesses time slot k, then allthe transmissions in the slot fail and time slot k is notacquired by any of the contending nodes. Node x willdetermine whether its attempt was successful or not byobserving thes01time slots followingk. The attempt of

node x is considered successful iff the packets receivedfrom all i N(x) indicate that x N(i). Otherwise,node x re-accesses one of the time slots in A(x) until itsuccessfully acquires a time slot. Once node x acquires atime slot, it keeps using the same slot in all subsequentframes unless a merging collision happens. Similar to anaccess collision, a merging collision is detected by nodex as soon as it receives a packet from a node i N(x)indicating that x /N(i).

At the end of each time slot, the collision detectionby a certain node x should be done before updatingthe set N(x). Upon receiving a packet from a nodey indicating that x / N(y), we stress on that, node

x should approve this collision detection and releaseits time slot iff the transmitting node y N(x). Thiscondition is referred to as the slot release prevention(SRP) condition, and its main objective is to prevent nodex from unnecessarily releasing its time slot when it justenters the communication range of another node y. Toillustrate that, consider the time slot assignment shownin Fig. 5 for the two nodes x and y . When nodex entersthe communication range of nodey , even if no collisionhappens, the first packet received by node x from nodey will indicate that x / N(y). The reason is that, bythe time node y transmits its packet, node y has not yetreceived any packet from node x to include it in N(y).

By applying the SRP condition, when node xreceives thefirst packet from node y, node x determines that nodey / N(x) and does not release its time slot (rememberthat collision detection by node x is done before updat-ing N(x)). After node xs transmission, the subsequentpackets transmitted by nodeywill indicate thatx N(y)and, hence, the unnecessary release of node xs timeslot is prevented. Note that, without the SRP condition,when two nodes enter the communication range of eachother, one of them will eventually release its time sloteven if no merging collision happens. This behaviourcan significantly decrease the performance of a TDMA

8/12/2019 Hassan TMC 12

5/14

5

Fig. 5: The SRP condition preventing node x fromunnecessarily releasing its time slot.

protocol as discussed in Subsection 5.3.Consider that nodex is moving in one of the right di-

rections. Initially, node x limits the set A(x)to the avail-able time slots associated with the right directions, i.e.A(x) = T0(x)R . If after a certain number of frames, say frames, nodex cannot acquire a time slot, then node xaugmentsA(x)by adding the time slots associated withthe opposite direction, i.e. A(x) = T0(x) (R L).If, after more frames, node x still cannot acquire atime slot, node x will start to access any available timeslot, i.e. A(x) = T0(x). The same procedure applies fora vehicle moving in a left direction by replacing R with

L. Similarly, if node x is an RSU, for the first framesA(x) = T0(x) F, and then A(x) = T0(x). The parameter is referred to as split up parameter, and the choice ofthe value can affect the rates of access collision andmerging collision. For example, when = 0, all thevehicles and RSUs are accessing the same set of timeslots. Hence, a merging collision is possible betweenany two nodes. However, when a merging collisionhappens, each colliding nodex is free to access any timeslot in T0(x), which can decrease the probability of anaccess collision. On the other extreme, when =, thevehicles moving in opposite directions and the RSUs areaccessing disjoint sets of time slots. However, when a

merging collision happens, for example among vehiclesmoving in a right direction, there is a higher probabilityof an access collision (compared with the = 0 case)since the choice of each colliding node x is limited totime slots in T0(x) R. A performance comparison be-tween theses two extreme cases is provided in Section 5.

Using the proposed scheme, a reliable broadcast ser-vice can be provided on channel c0. That is, if node xtransmits a broadcast packet on time slot k , by listeningto thes0 1time slots followingk, nodex can determinethe set of one-hop neighbors which have successfullyreceived the packet, where = {i N(x) : x N(i) }.

In other words, when node i indicates that x N(i), itis considered as an implicit acknowledgement by nodei of receiving the packet broadcasted by node x.

3.3 Accessing Slots on the Service Channels

Consider that a node x has a MAC layer service dataunit (MSDU)6 to be delivered to a certain destination (as-

suming unicast) on service channel cm. In what follows,the term service refers to the delivery of an MSDU ona certain service channel. By using Tm(x) (how node xconstructsTm(x) will be explained), node x determinesthe set of time slots that it will access on channelcm to offer the service, denoted by m(x), such thatm(x) Tm(x) =. Accordingly, node x announces thefollowing information in the AnS field of its next packettransmitted on channel c0: a) priority of the service,

b) reliability of the service (i.e. acknowledged or not),c) MAC address of the intended destination y, d) thenumbermof the service channel, and e) m(x). Once theprovider x announces for the service, no further action

is needed unless the destination accepts the service asdescribed below.Based on the information announced by provider x on

channelc0, the destination y determines whether or notto make use of the announced service. If node y decidesto use the service by provider xon channelcm, it acceptsthe service by includingm(x)in theAcSfield of its nextpacket transmitted on channel c0. The announcementof m(x) by the user y is for the surrounding nodesto update their Tm sets as to be discussed. Also, for areliable service, node y should include in the AnSfieldthe time slot that will be used by node y to transmitthe acknowledgement packet, denoted by km(y). Node

y determines km(y) such that km(y) / Tm(y). Whenprovider xreceives the acceptance of the service, it tunesits Transceiver2 to channel cm and starts offering theservice on the time slots announced in m(x). As well,if the service is reliable, node x should include km(y)inthe AcSfield of its next packet transmitted on channelc0. Again, the announcement ofkm(y) by provider x isto avoid the collision of the acknowledgement packet byproperly updating theTmsets of the surrounding nodes.Nodey should transmit the acknowledgement only afternode x announces km(y) on channel c0.

Each node updates sets Tm, m = 1, . . . , M , as fol-lows. When node x receives a packet on channel c0

from another node y, based on the position of nodey which is included in the header of the packet, andthe position of node x obtained from the GPS receiver,nodex can estimate its distance to node y . Based on thisestimated distance and on the fixed transmission poweron all channels which is known to node x, node x candetermine whether or not nodey is in its communicationrange on channel cm, m= 1, . . . , M

7. If node x decides

6. An MSDU is a unit of data arriving to the MAC layer entity fromthe layer above.

7. It is assumed that each node has a path loss model for each servicechannel cm,m = 1, . . . ,M .

8/12/2019 Hassan TMC 12

6/14

6

that it can reach node y on a certain channel cj , nodex adds to set Tj(x) the time slots indicated by each jset and kj slot included in the AcS field of the packettransmitted by y. The reason is that, each j representsa set of time slots over which nodeywill receive a packeton channelcj from a certain provider in the next sj slots.Similarly, each kj indicates a time slot over which nodey will receive an acknowledgement packet on channelcj from a certain user in the next sj slots. Consequently,

by updatingTj in the way described, collision at nodey can be prevented since all the one-hop neighbourhoodof node y will avoid using the time slots over whichnode y will receive packets. At the end of time slot imon channel cm, if im Tm, im is removed from Tm,m = 1, . . . , M and im = 1, . . . , sm. Note that updatingtheTm,m = 1 . . . , M , sets is based on information in the

AcS(not in theAnS) field, which eliminates any exposedterminal problem. The following example illustrates howthe nodes access the service channels.

Consider the THS configuration shown in Fig. 6, node

x has a reliable service to offer to node z on time slotsnumbered 2, 3, and 5 on channel c1. Fig. 6 shows thesequence of actions taken by providerx, userz , and thesurrounding nodesy and w. First, nodex announces forthe service and includes1(x) = {2, 3, 5}in theAnSfieldof its packet transmitted on channel c0. Following thisannouncement, no action is taken by both surroundingnodes w and y. Once node z accepts the service andannounces 1(x), node x starts offering the service onchannelc1 on time slots {2, 3, 5}as announced in 1(x).When node y receives the packet transmitted by nodez on channel c0, it adds 1(x) to T1(y) to avoid usingthe upcoming time slots {2, 3, 5}over which node z will

receive packets from node x (assume that node y canreach node z on channel c1). Note that, node w is freeto use the time slots in 1(x) ={2, 3, 5} since it did notreceive the acceptance of service transmitted by node zon channel c0; hence, simultaneous transmissions fromnodew to vand from node xtoz are allowed on channelc1, i.e. no exposed terminal problem. However, in theabsence of the exposed terminal problem, it is possiblethat node w announces a service to node y on timeslots {2, 3, 5} after node x did the same announcementto node z (note that simultaneous transmissions fromnode w to y and from node x to z result in a collisionat node y). In this case, if node y accepts the service

and includes 1(w) = {2, 3, 5} in the AcS field of itspacket transmitted on channel c0 (on time slot {7}),node x will receive this packet transmitted by node y,includes1(w)to T1(x), and avoids using the upcomingtime slots {2, 3, 5} on channel c1 to prevent collision atnode y (recall the definition of T1(x)), although nodex was supposed to transmit a packet to node z on thetime slot {3} following node ys acceptance of service.This missing packet, together with the other packetsincorrectly received by node z, are (re)transmitted bynodexafter it receives the acknowledgment packet fromnode z. The acknowledgement packet is transmitted

Fig. 6: Node x offering a service to node zon channel c1.

using the same procedure as illustrated in Fig. 6.

3.4 Multi-hop Broadcast Service

This subsection shows that the efficient multi-hop broad-cast service presented in [12] for ADHOC MAC can bedirectly supported by VeMAC on channel c0. Supposenode x transmits a broadcast packet on channel c0, andconsider that this packet needs to propagate throughoutthe whole network. For each node i which receivesthe broadcast packet, define Z(i) as the set of one-hopneighbours of node i which did not receive the packet

broadcast by node x. Node i does not relay the packetif one of the following holds:

Z(i) = ; j N(i)\Z(i) such that Z(i) N(j) and |N(j)|>

|N(i)|, where | | denotes the cardinality of a set; j N(i)\Z(i) such that Z(i) N(j), |N(j)| =

|N(i)|, and ID(j)> ID(i).

When nodei receives the broadcast packet from nodex, it listens to channelc0 fors0 1successive time slots.At the end of this duration, node ican determine the setsN(j), j N(i), and Z(i) = {j N(i) : x / N(j)}.

Accordingly, node i relays the packet if none of the pre-vious three conditions is satisfied. By using this relayingprocedure, it is shown in [12] and [15] that, in most cases,the minimum set of relaying nodes needed to cover thewhole network is selected.

4 PERFORMANCE A NALYSIS

4.1 Time Slot Acquisition

The objective of the analysis in this subsection is toinvestigate how fast the contending nodes can acquirea time slot on channel c0 by using the VeMAC protocol.

8/12/2019 Hassan TMC 12

7/14

7

Let Kdenote the number of contending nodes, each ofwhich needs to acquire a time slot on channel c0. Wewant to determine the average number of nodes whichacquire a time slot within n frames, the probability thata specific node acquires a time slot within n frames, andthe probability that all the nodes acquire a time slotwithin n frames. To simplify the analysis, the follow-ing assumptions are made: a) all the contending nodes

belong to the same set of THSs, with the same T0 andA sets, e.g. node w and node x in its final position inFig. 4; b) the set of THSs to which the contending nodes

belong does not change; c) the set A is not augmentedwhen a node fails to acquire a time slot after frames,i.e.= 0; d) at the end of each frame, each node is awareof all acquired time slots during the frame, and updatesthe sets T0 and A accordingly, i.e. all nodes are withinthe communication range of each other; e) at the end ofeach frame, all contending nodes are informed whetheror not their attempts to access a time slot during thisframe were successful. Based on this information, each

colliding node randomly chooses an available time slotfrom the updatedA set, and attempts to access this slotduring the coming frame.

Let N be the number of initially available time slotsin a frame, and Xn be the total number of nodes whichacquired a time slot within n frames. Under the assump-tions,Xnis a stationary discrete-time Markov chain withthe following transition probabilities.IfK N,

pij =

W(ji,Ki,Ni)(Ni)Ki

, 0 i K 1,

i j K

1, i= j = K

0, elsewhere

whereW(l ,u,v)is the number of ways by which l nodescan acquire a time slot given that there are u contendingnodes each randomly choosing a time slot among vavailable time slots. A node acquires a time slot if noother nodes choose to access the same slot. The Markovchain is illustrated in Fig. 7.

Fig. 7: Markov chain for Xn when K N.

IfK > N,

pij =

W(ji,Ki,Ni)(Ni)Ki

, 0 i N 1,

i j N 1

1, i= j, Ni K

0, elsewhere.

The Markov chain is illustrated in Fig. 8. To calculate

Fig. 8: Markov chain for Xn when K > N.

W(l ,u,v), considering u different balls randomly dis-tributed in v different boxes with equal probabilities,W(l ,u,v) is the number of ways of having l boxeseach containing exactly one ball. This special occupancyproblem is solved in a recursive way as follows [16].

Ifu v,

W(l ,u,v) =

Cul Avl

(v l)ul

uli=1W(i, u l, v l)

, 0 l < u

Avl , l= u

0, l > u

where Avl = v!(vl)! and C

ul =

Aul

l! .

Ifu > v ,

W(l ,u,v) =

Cul Avl

(v l)ul

vli=1 W(i, u l, v l)

, 0 l < v

0, l v.

Let Pbe the one-step transition probability matrix, andPn the n-step transition probability matrix. Given thatinitially all nodes are contending for time slots, i.e.X0 = 0with probability 1, the unconditional probabilitydistribution ofXn is represented by the first row ofP

n.That is,

p(Xn= i) = Pn1,i+1, i= 0, . . . , K .

The probability that all nodes acquire a time slot withinn frames is

Falln =p(Xn= K) = Pn1,K+1.

The average number of nodes which acquire a time slotwithin n frames is

n=Ki=0

iPn1,i+1.

The probability that a specific node, say node x, acquiresa time slot within n frames is

Fn =

Ki=0

p(E|Xn= i)p(Xn= i)

=Ki=1

CK1i1CKi

Pn1,i+1 =n

K

where E is the event that node x acquires a time slot

8/12/2019 Hassan TMC 12

8/14

8

0 1 2 3 4 5 6 7 8 9 100

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

n (frame)

Fna

ll

N=15 K=5

N=15 K=10

N=15 K=15

N=20 K=7

N=20 K=12

N=20 K=17

N=20 K=20

Fig. 9: Probability that all nodes acquire a time slot within nframes.

within n frames and p(E|Xn = i) = CK1

i1

CKi

= iK since

all nodes have equal chances of acquiring a time slot.

Note that, since the VeMAC assumes a fixed numberof constant duration time slots in a frame on channelc0, the choice of the s0 value should always ensure thatK N. However, the analysis of the protocol for thecase K > Ncan be useful in the future to determine anoptimal value fors0. This analysis gives an indication ofhow the protocol will behave if the number of nodes ina THS becomes larger than s0.Fig. 9 illustrates Falln for different values of N andK. As shown in Fig. 9, in a dense scenario such as(N= 15, K= 15), there is a probability greater than 0.9that all the contending nodes acquire a time slot within8 frames. Hence, given a frame duration of 35 ms (as

discussed in Subsection 4.2), the simplifying assumptionof invariant THSs (assumption b) is acceptable, sinceit is reasonable to assume that the THS configurationremains constant for a sufficiently large time after allthe contending nodes acquire a time slot. The analysispresented in this subsection is verified in [1] via Matlabsimulations.

4.2 Protocol Overhead and Packet Delay

As discussed in Subsection 3.1, a VeMAC packet trans-mitted on channel c0 consists of header, Ans, Acs, andhigh priority short applications fields. The size of each

field in a packet transmitted by a node, x, is estimatedas follows. The main part of the header consists ofannouncing the setN(x)and the time slot used by eachnode inN(x)on channelc0. On the other hand, the mainparts of the Ans and Acs fields consist of the sets m(x)and n(x) of time slots over which node x is offeringservice or will receive packets on service channels cmand cn respectively. If the maximum number of nodeswhich can exist in a THS is Nmax, at least log2Nmax

bits are required to represent a node ID, where denotes the ceiling function. Similarly, log2si bits aresufficient to identify a time slot in a frame on channel

ci, i = 0, . . . , M . Therefore, the total VeMAC packet sizeS(in bits) is

S=|N(x)|

bID + log2s0

+ |m(x)|log2sm+

|n(x)|log2sn + bapp+ bextra

where bID is the number of bits to represent a nodeID, bapp is the number of bits for the high priority

application field, and bextra is the number of bits forall other information in the packet such as MAC ad-dresses, node x position, priority fields, error correctingcodes, etc. Assuming Nmax = 100 (as assumed in [12]),bID = 1 byte, si = 100, i = 0, . . . , M , bapp = 200 bytes,bextra = 30 bytes, |m(x)| = |n(x)| = 10, and |N(x)| =100, the estimated VeMAC packet size in this case isS= 3480 bits 435 byte. While the information in theheader, AnS and AcS fields, represents an overhead ofthe VeMAC protocol on channel c0, the communicationsover the service channels ci, i= 1, . . . , M , are overhead-free.The delay that a high priority safety packet experiences

on channel c0 depends on the value of s0 as well asthe duration of a time slot. Considering a maximumVeMAC packet size of 450 byte and a transmission rateof 12 Mbps8, the packet requires a transmission time of0.3 ms. By adding guard periods and taking accountof the physical layer overhead, such as the preambleand the physical layer header, a 0.35 ms slot durationcan be assumed. In terms of synchronization, this slotduration is suitable as it is much larger than the jitter ofthe 1PPS of GPS receivers which is usually in the order ofnanoseconds. Assuming thats0 = 100slots, the durationof one complete frame on channel c0 is 35 ms. Hence,when a node is successfully acquiring a time slot, it can

transmit its high priority safety messages once every35 ms, which is compliant with the 100 ms maximumdelay requirements for most of the safety applicationsin [2]. However, due to packet queueing, the total delaythat a safety packet encounters on channel c0 can belarger than the duration of one complete frame. For thisreason, a detailed packet delay analysis on channel c0will be considered in the future.

5 SIMULATIONS

This section presents Matlab simulation results to eval-uate the performance of VeMAC as compared with

ADHOC MAC in accessing channel c0 in highway andcity scenarios.

5.1 Simulation Scenarios and Performance Metrics

The first scenario under consideration is a segment ofa two-way vehicle traffic highway. A vehicle can com-municate with all the vehicles within its communicationrange, i.e. no obstacles. Each vehicle moves with aconstant speed drawn from a normal distribution, and

8. The 12 Mbps is one of the mandatory supported bit rates by theIEEE 802.11p OFDM physical layer for the 5GHz band.

8/12/2019 Hassan TMC 12

9/14

9

TABLE 1: Simulation parameters

Parameter Highway CityHighway length 1km

# horizontal streets 3# vertical streets 3

City street length 430m# city blocks 4

City block edge length 200m# lanes/direction 2 1

Lane width 5m 5m

Speed mean value 100km/h 50km/hSpeed standard deviation 20km/h 10km/hTransmission range 150m 150m

# slots/frame 100 100# slots for left directions 50 50

# slots for right directions 50 50# slots for RSUs 0 0

Slot duration 1ms 1ms

Simulation time10 2 min. 2 min.

# vehicles 80 to 280

(step= 20)150 to 600

(step= 50)

THSO 0.24 to 0.96

(step= 0.06)0.17 to 0.70

(step= 0.06)

the number of vehicles on the highway segment remainsconstant during the simulation time. When a vehiclereaches one end of the highway segment, it re-enters thesegment from the other end. For this reason, to preventthe unrealistic merging collisions caused by vehicleswhich jump from one end to the other end, if a vehicleis located at a distance d R (R is the communicationrange) from one end of the highway segment, it cancommunicate with vehicles located within a distanceR d from the other end of the segment. In this way,for each traffic direction, the vehicles at the end of thesegment act as if they are following the vehicles at thestart of the segment.The second scenario is a city grid layout consistingof three horizontal and three vertical two-way vehicle

traffic streets. All the streets have the same dimensions,and the horizontal and vertical streets are evenly spacedresulting in four identical square city blocks9. The areaof intersection of a horizontal street with a vertical oneis referred to as a junction area. Each vehicle moveswith a constant speed drawn from a normal distribution.When a vehicle reaches a junction area, it chooses oneof all possible moving directions with equal probability(vehicles are not allowed to leave the simulation areaduring the simulation time). A vehicle located at a

junction area can communicate with vehicles within itscommunication range located on both streets intersectingat the junction area. On the other hand, a vehicle located

at a street but not at a junction area cannot communicatewith vehicles located on other streets due to the existenceof city blocks which obstruct the wireless signal.For both scenarios under consideration, all the transmit-ted packets are broadcast packets, the wireless channelis ideal, and the only source of packet errors is thetransmission collision. Table 1 summarizes the simula-tion parameters and Figs. 10 and 11 show snap shots ofthe simulated highway and city scenarios respectively,

9. A city block is the smallest area that is surrounded by streets.10. Each minute is simulated with a unique initial distribution of the

vehicles.

0 200 400 600 800 10000

5

10

15

20

(m)

Fig. 10: A snap shot of the simulated highway segment.

0 100 200 300 4000

100

200

300

400

(m)

Fig. 11: A snap shot of the simulated area of the city.

where the black and white triangles represent vehiclesmoving in opposite directions.We define a parameter, called the THS occupancy(THSO), equal to Nv

2RLh

1s0 or NvNs

2RLs 1s0

in

the highway and city scenarios respectively, where N

vis the total number of vehicles, Ns is the total number ofstreets in the city, R is the communication range, Lh isthe length of the highway segment, Ls is the length ofa city street, and s0 is the number of slots per frame onchannelc0. Note that, the ratio

NvNs

approximately equalsthe number of vehicles on a city street, the number s0represents the maximum number of time slots availablefor a THS, and the length 2R is the maximum lengththat a THS can occupy on the highway segment or on acity street. Consequently, the THSO indicates the ratio ofthe number of time slots required by a THS to the totalnumber of time slots available for a THS. However, the

THSO is not guaranteed for each THS in the simulations.The reason is that, if there are Nv moving vehicles, thisdoes not mean that at each instant, each THS on thehighway consists ofNv

2RLh

vehicles or each THS in the

city consists of NvNs 2RLs

vehicles. Also, in the city scenariounder consideration, a THS located near a junction areacan occupy a length on the streets up to4R(2Ron eachof the horizontal and vertical street intersecting at the

junction area).The following performance metrics are considered:

a) Rate of merging collisions: the average number ofmerging collisions per frame per THS;

8/12/2019 Hassan TMC 12

10/14

10

b) Rate of access collisions: the average number of accesscollisions per slot per THS;

c) Tx throughput: the average number of successfultransmissions per slot per THS. A transmission by avehiclexin a certain time slot is considered successfuliff no other vehicles in the two-hop neighbourhood ofx transmits in the same slot;

d) Rx throughput: the average number of successfullyreceived packets per slot per THS. As mentioned,packet errors only happen due to transmission col-lision.

Each of the metrics is calculated first for the wholesimulation area, and then multiplied by 2RLh or

1Ns

2RLsfor the highway and city scenarios respectively. Notethat, unlike the other three metrics, the rate of mergingcollisions is calculated per frame not per slot. The reasonis that, merging collisions happen due to the movementof the vehicles, which is negligible in the duration ofone time slot. The metrics are obtained for each ofthe MAC protocols mentioned in Subsection 5.2. At the

beginning of the simulations, the vehicles are randomly(uniformly) placed on the highway segment and on allstreets of the city. The vehicles remain stationary and tryto acquire a time slot by using the MAC protocol underconsideration. Once no more vehicle can acquire a timeslot, the vehicles begin moving and the simulation timerstarts. The objective of this process is to quickly bringthe system to a steady state where most of the vehicleshave acquired a time slot.

5.2 Simulated Protocols

Two versions of the VeMAC protocol are considered:

VeMAC with = 0 and = . As will be shownin Subsection 5.3, both versions of the VeMAC protocolsignificantly outperform the ADHOC MAC protocol in[12]. The poor performance of ADHOC MAC is caused

by the following two main reasons. First, due to thelack of a condition similar to the SRP condition inVeMAC, when two vehicles having acquired a time slotenter the communication range of each other, one ofthem releases its time slot even if no merging collisionhappens. Second, as mentioned in [12], a node whichneeds to acquire a time slot should attempt transmissionin the next available time slot with probability p. For acertain time slot, the optimal probability popt = 1/Nc,

whereNc is the number of contending nodes attemptingto acquire this time slot [12]. However, since Nc is notknown to any of the contending nodes, each contendingnode x sets Nc = Nmax Nsucc(x), where Nmax isthe maximum number of nodes which can exist in aTHS and Nsucc(x) is the number of nodes in the two-hop neighbourhood of node x which have successfullyacquired a time slot as derived from the framing in-formation received by node x [12]. This estimation ofNc is far from accurate. The reason is that, if a node xdetects that Nsucc(x) nodes have successfully acquireda time slot, this does not mean at all that there are

TABLE 2: The simulated protocols

Protocol AbbreviationVeMAC with = V-infVeMAC with = 0 V0

ADHOC MAC as in [12] ADHOCADHOC-enhanced AE

ADHOC-optimal A-opt

Nmax Nsucc(x)nodes which need to acquire a time slotin the two-hop neighbourhood of node x. Also, even ifthere are exactlyNmaxNsucc(x)contending nodes, theydo not necessarily contend for the same time slots sinceeach of the nodes may belong to a different set of THSs.Additionally, Nmax is not constant since it depends onparameters such as the inter-vehicle distance and thenumber of lanes which considerably vary based on thescenario (i.e. highway, city, urban, sub-urban, or ruralareas).

Based on the two limitations of the typical ADHOCMAC protocol [12], two more versions of ADHOC MACare considered in the simulations: the ADHOC-enhancedand the ADHOC-optimal. The ADHOC-enhanced elim-inates the first limitation of ADHOC MAC by using acondition similar to the SRP condition of VeMAC. Moreprecisely, a nodex does not release its time slot based ona packet received from a nodey unless nodex has previ-ously received a packet from nodey, i.e. unless nodey isincluded in the framing information [12] constructed bynodex. For both ADHOC MAC and ADHOC-enhanced,the probability of accessing an available time slot by acontending node x is p = 1s0Nsucc(x) . Note that, Nmaxis replaced by s0 (i.e. the maximum number of slotsavailable for a THS) as it is not mentioned in [12] how

to determineNmax. To evaluate the second limitation ofADHOC MAC, the ADHOC-optimal protocol is imple-mented. The ADHOC-optimal is similar to the ADHOC-enhanced protocol with the difference that, for each timeslot, each contending node is aware of the number ofcontending nodesNc within its two-hop neighbourhoodand sets p = popt =

1Nc

. Note that this awareness ofNc is provided by the simulator and cannot be achievedin reality. Hence, the ADHOC-optimal is not a realisticprotocol, it just represents an upper bound on the perfor-mance of ADHOC MAC. The five MAC protocols underconsideration are summarized in Table 2.

To demonstrate the difference among the three ADHOC

MAC versions, Fig. 12 shows the number of vehiclessuccessfully acquiring a time slot, denoted by Nq, inthe first five seconds of the simulation in the highwayscenario. For the ADHOC protocol, due to the lack ofthe SRP condition, Nq drops from 60 to 20 vehicle/THSin the first second of the simulation. Also, each vehiclewhich releases its time slot in the first second cannotquickly acquire a new one due to the inexact probabilityof accessing an available time slot. For this reason, Nqremains below 20 vehicle/THS at the end of the fiveseconds. Unlike ADHOC MAC, in the AE protocol, thesudden decrease in Nq is eliminated thanks to the SRP

8/12/2019 Hassan TMC 12

11/14

11

0 1 2 3 4 510

20

30

40

50

60

Time (s)

Nq

(

vehicle/THS)

ADHOC

AE Aopt

Fig. 12: The number of vehicles acquiring a time slot for thethree ADHOC MAC versions in the highway scenario, at

THSO = 0.6(i.e. 60 vehicle/THS).

0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

0.25

0.5

0.75

1

1.25

1.5

1.75

2

2.25

THSO

Rate

of

merg

ing

co

llis

ions

(co

llis

ion

/frame

/THS

)

ADHOC

AE

V0

Vinf

Aopt

Fig. 13: The rate of merging collisions in highway.

condition. For this protocol, Nq decreases gradually andreaches 54 vehicle/THS at the end of the five seconds.On the other hand, the A-opt protocol does not showany decrease inNq at the end of the five seconds since itcan control the access collisions by using the optimalprobability popt for accessing the available time slots.Similar behaviours of the three ADHOC MAC versionswere seen in the city scenario (results are omitted).

5.3 Simulation Results

5.3.1 Highway Scenario

Fig. 13 shows the rate of merging collisions for all the

MAC protocols under consideration. The V-inf protocolachieves a low rate of merging collisions since it assignsdisjoint sets of time slots to vehicles moving in oppositedirections. The V0 and A-opt protocols have almostthe same rate of merging collisions for different THSOvalues. Note that for a high THSO, the ADHOC protocolprovides a low rate of merging collision, even less thanthe V-inf protocol, due to a small number of nodes whichsuccessfully acquire a time slot as compared to otherprotocols (recall that, by definition, a merging collisionhappens only among the nodes which are successfullyacquiring a time slot).

0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

0.5

1

1.5

2

2.5

3

3.5

THSO

Rate

of

a

ccess

co

llis

ions

(co

llis

ion

/s

lot

/THS

)

ADHOC

AE

V0

Vinf

Aopt

Fig. 14: The rate of access collisions in highway.

The rate of access collisions is shown in Fig. 14 for allthe protocols. As expected, the A-opt protocol shows aconsiderably smaller rate of access collisions than bothADHOC and AE protocols, which verifies the ineffi-ciency of both protocols in determining the probabil-ity of accessing an available slot. Due to the abilityof the V-inf protocol to decrease the rate of mergingcollisions, as shown in Fig. 13, it also achieves a lessrate of access collisions than that of the V0 protocol. Thereason is that, each merging collision generates accesscollisions, especially for a high THSO, until each nodewhich released its time slot reacquires a new one. BothVeMAC protocols (V-inf and V0) provide a rate of accesscollisions which is slightly higher than that of the A-optprotocol but significantly lower than the rates provided

by the ADHOC and AE protocols especially for a high

THSO.Fig. 15 shows the Tx throughput for all the protocols.Because of the limitations discussed in Subsection 5.2,the performance of the ADHOC protocol is the lowestamong all the MAC protocols for all the THSO values.The AE protocol has better performance than the AD-HOC protocol, but its Tx throughput decreases for a highTHSO due to its inability to handle the access collisions.For a THSO < 0.7, the V-inf and V0 protocols havealmost the same Tx throughput, while for a THSO> 0.7,the V-inf protocol starts to perform better than the V0protocol. Both protocols outperform the AE and ADHOCprotocols for all the THSO values, and the Tx throughput

of the V-inf is slightly less than the unrealistic A-optprotocol for a THSO > 0.7.

The Rx throughput is shown in Fig. 16. It is clearthat, the V-inf and V0 protocols achieve a higherRx throughput than both of the AE and ADHOCprotocols for all the THSO values. For instance, atTHSO= 0.78, the V-inf protocol provides an Rx through-put of 51 packet/slot/THS as opposed to only 21packet/slot/THS in the case of the ADHOC protocol (i.e.a 143% increase in the Rx throughput). Note that, for ahigh THSO, even if the Tx throughput remains constantor slightly decreases, the Rx throughput continues in-

8/12/2019 Hassan TMC 12

12/14

12

0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10.1

0.2

0.3

0.4

0.5

0.6

0.7

THSO

Txthroughput

(packet/slot/THS)

ADHOC

AE

V0

Vinf

Aopt

Fig. 15: The Tx throughput in highway.

0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

10

20

30

40

50

60

70

THSO

Rx

throug

hput

(pac

ke

t/s

lot

/THS

)

ADHOC

AE

V0

Vinf

Aopt

Fig. 16: The Rx throughput in highway.

creasing. The reason is that, for the same Tx throughput,

when the number of vehicles on the highway segmentincreases (i.e. when the THSO increases), more vehiclescan receive packets since all the packets transmitted areof broadcast type. Similar to the Tx throughput, the V-inf protocol provides a slightly less Rx throughput thanthe A-opt protocol. For the range of THSO considered inthe highway, the maximum relative difference11 betweenthe Rx throughput of the V-inf and A-opt protocols isapproximately 3.9% (achieved at THSO= 0.72).

5.3.2 City Scenario

The rate of merging collision in the city scenario isshown in Fig. 17 for all the protocols. It is noted that, therelative difference between the rate of merging collisionprovided by the V-inf protocol and that provided bythe V0 protocol is reduced as compared to the highwayscenario. For instance, at a THSO = 0.7 in the highwayscenario, the V0 protocol shows approximately 150%higher rate of merging collision than the V-inf protocol,as opposed to only an 8% increase in the city scenario atthe same THSO. The reason is that, in the city scenario,the V-inf protocol suffers from the merging collisions

11. The relative difference between two values x and y is defined as|xy|

min(x,y)

0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.70

1

2

3

4

5

6

THSO

Rate

of

m

erg

ing

co

llis

ions

(co

llis

ion

/frame

/THS

)

ADHOC

AE

V0

Vinf

Aopt

Fig. 17: The rate of merging collisions in city.

0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.70

0.3

0.6

0.9

1.2

1.5

1.8

THSO

Rate

of

access

col

lis

ions

(co

llis

ion

/s

lot

/

THS

)

ADHOC

AE

V0

Vinf

Aopt

Fig. 18: The rate of access collisions in city.

near the junction areas due to vehicles which changetheir moving direction. This kind of merging collisiondoes not exist with the V-inf protocol when employedin the highway scenario (the merging collisions onlyhappens among vehicles moving in the same direction).The close rate of merging collisions of both V-inf and V0protocols also results in a close rate of access collisions,as shown in Fig. 18. Similar to the highway scenario,

both V-inf and V0 protocols provide a rate of accesscollision which is higher than that of the A-opt protocol

but lower than those provided by the AE and ADHOCprotocols.

The Tx throughput and Rx throughput are shown in

Figs. 19 and 20 respectively. The V-inf and V0 protocolshave the same performance for a THSO< 0.5, while theV-inf protocol performs slightly better for a THSO > 0.5.Unlike the highway scenario, where the A-opt and V-infprotocols have very close Tx and Rx throughputs, in thecity scenario the A-opt outperforms the V-inf protocol.This outperforming is a result of the excess mergingcollisions that the V-inf protocol experiences in the cityscenario due to vehicles which change their movingdirections. However, similar to the highway scenario,

both V-inf and V0 protocols provide higher Tx and Rxthroughputs than the AE and ADHOC protocols.

8/12/2019 Hassan TMC 12

13/14

13

0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.70.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5

THSO

Txthroughput

(packet/slot/THS)

ADHOC

AE

V0

Vinf

Aopt

Fig. 19: The Tx throughput in city.

6 DISCUSSION

In terms of communication over the control channel, themain similarities and differences between the VeMACand ADHOC MAC protocols can be summarized as fol-lows. Both protocols are based on TDMA, work over thephysical layer of different standards (such as the IEEE802.11), and achieve an efficient multi-hop broadcastservice as well as a reliable one-hop broadcast servicewithout the hidden terminal problem. Also, they bothrequire each node to periodically announce the time slotsused by all nodes within its one-hop neighbourhood.However, the VeMAC protocol significantly outperformsthe ADHOC MAC protocol, thanks to the followingthree main features: the reduction of the access collisionrate by using fixed time frames (versus sliding framesin ADHOC MAC) and a new method for the nodes

to access the available time slots, the reduction of themerging collision rate by assigning disjoint sets of timeslots to vehicles moving in opposite direction and toRSUs, and the SRP condition which prevents the nodesfrom unnecessarily releasing their time slots when they

just enter the communication range of each other. Theseadvantages of VeMAC come in addition to being a multi-channel protocol more suitable for the DSRC spectrum ascompared to the single channel ADHOC MAC protocol.

Since the VeMAC protocol assumes constant frame andslot durations on channel c0, the value ofs0 should belarge enough to accommodate all the nodes which canexist in a THS and prevent any node starvation. In a low

density scenario, when the number of nodes in a THSis much less than s0, or in case of an event driven highpriority safety message, it is possible that a node accessesadditional time slots in a frame on channel c0by virtually

behaving as multiple nodes. That is, the node employsthe same procedure as described in Subsection 3.2 byusing a different node ID to access each additional timeslot. However, the conditions based on which a node isallowed to access more than one time slot per frame needmore investigation to prevent any unfair scenario, wheresome nodes are accessing more than one time slot, whileother nodes cannot successfully acquire a time slot.

0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.70

5

10

15

20

25

30

35

THSO

Rx

throug

hput

(pac

ket

/s

lot

/THS

)

ADHOC

AE

V0

Vinf

Aopt

Fig. 20: The Rx throughput in city.

7 CONCLUSIONS ANDFUTURE WOR K

In this paper, a novel multichannel TDMA MAC proto-col, called VeMAC, is proposed for VANETs based on the

ADHOC MAC protocol. Each node is ensured to accessthe control channel once per frame, and hence nodeshave equal opportunities to announce for services pro-vided on the service channels and to transmit their highpriority application messages. The nodes access the timeslots on the control channel and service channels in dis-tributed ways which are designed to avoid any hiddenterminal problem. On the control channel, the VeMACprovides a reliable one-hop broadcast service, which iscrucial for high priority safety applications supported onthis channel. As well, the efficient multi-hop broadcastservice of ADHOC MAC can be directly supported byVeMAC on the control channel. Simulation results in

highway and city scenarios show that, compared withthe ADHOC MAC and ADHOC-enhanced protocols, theVeMAC provides a smaller rate of transmission collisions(access collisions and merging collisions), which resultsin a significantly higher throughput on the control chan-nel.In the future, the effect of the existence of RSUs onthe performance of the VeMAC protocol will be in-vestigated in both highway and city scenarios usingrealistic mobility models. As well, the performance ofVeMAC using different values of the split up parameter, other than = 0 and = , will be considered.Also, the mechanisms and the conditions which allow

each node to access more than one time slot per frameon the control channel will be investigated. A detailedpacket delay analysis will be conducted to calculate thetotal delay that a safety message experiences on thecontrol channel. Another issue which will be examinedis the effect of asymmetric wireless channels amongthe nodes as well as the effect of packet errors caused

by the wireless channel impairments such as noise,fading, and shadowing. Since each node interprets anypacket error as a transmission collision, the packet errorsdue to a poor wireless channel may result in nodesunnecessarily releasing their time slots on the control

8/12/2019 Hassan TMC 12

14/14

14

channel. Concerning the service channels, we plan toevaluate the proposed scheme for unicast via analysisand simulations, and extend this scheme to support areliable broadcast on the service channels. Finally, theperformance of VeMAC on the control channel andservice channels will be compared with that of the IEEE802.11p operating under the IEEE 1609.4 standard [17]for a multichannel operation.

ACKNOWLEDGMENTS

We would like to thank the reviewers for their commentswhich really helped to improve the quality of this paper.

REFERENCES

[1] H. A.Omar, W.Zhuang, and L.Li, VeMAC: a novel multichannelMAC protocol for vehicular ad hoc networks, in Proc. IEEEINFOCOM Workshop on Mobility Management in the Networks ofthe Future World (MobiWorld 2011), Apr. 2011, pp. 413418.

[2] Vehicle safety communications project task 3 final report, TheCAMP Vehicle Safety Communications Consortium, Tech. Rep.DOT HS 809 859, Mar. 2005.

[3] R.Baldessari et al., Car-2-car communication consortium mani-festo, Tech. Rep. Version 1.1, Aug. 2007.

[4] J. J.Blum and A.Eskandarian, A reliable link-layer protocol forrobust and scalable intervehicle communications, IEEE Trans.Intell. Transp. Syst., vol. 8, no. 1, pp. 413, Mar. 2007.

[5] R.Mangharam, R.Rajkumar, M.Hamilton, P.Mudalige, and F.Bai,Bounded-latency alerts in vehicular networks, in Proc. MobileNetworking for Vehicular Environments (MOVE 2007), May 2007,pp. 5560.

[6] F.Watanabe, M.Fujii, M.Itami, and K.Itoh, An analysis of inci-dent information transmission performance using MCS/CDMAscheme, inProc. IEEE Intelligent Vehicles Symposium (IV05), Jun.2005, pp. 249254.

[7] H.Nakata, T.Inoue, M.Itami, and K.Itoh, A study of inter vehiclecommunication scheme allocating PN codes to the location on theroad, in Proc. IEEE Intelligent Transportation Systems Conference(ITSC 2003), vol. 2, Oct. 2003, pp. 15271532.

[8] IEEE standard for information technologytelecommunicationsand information exchange between systemslocal and metropoli-tan area networksspecific requirements Part 11: Wireless lanmedium access control (MAC) and physical layer (PHY) specifica-tions Amendment 6: Wireless access in vehicular environments,IEEE Std 802.11p-2010, pp. 151, Jul. 15, 2010.

[9] IEEE standard for information technology-telecommunicationsand information exchange between systems-local and metropoli-tan area networks-specific requirements - Part 11: Wireless lanmedium access control (MAC) and physical layer (PHY) specifi-cations, IEEE Std 802.11-2007 (Revision of IEEE Std 802.11-1999),pp. 11184, Jun. 2007.

[10] M.Hassan, H.Vu, and T.Sakurai, Performance analysis of theIEEE 802.11 MAC protocol for DSRC safety applications, IEEETrans. Veh. Technol., vol. 60, no. 8, pp. 3882 3896, Oct. 2011.

[11] S.Eichler, Performance evaluation of the IEEE 802.11p WAVEcommunication standard, inProc. IEEE 66th Vehicular Technology

Conference (VTC 2007-Fall), Oct. 2007, pp. 21992203.[12] F.Borgonovo, A.Capone, M.Cesana, and L.Fratta, ADHOC MAC:New MAC architecture for ad hoc networks providing efficientand reliable point-to-point and broadcast services, Wireless Net-works, vol. 10, pp. 359366, Jul. 2004.

[13] F.Borgonovo, L.Campelli, M.Cesana, and L.Fratta, Impact of usermobility on the broadcast service efficiency of the ADHOC MACprotocol, in Proc. IEEE 61st Vehicular Technology Conference (VTC2005-Spring), vol. 4, Jun. 2005, pp. 23102314.

[14] W.Franz, H.Hartenstein, and M.Mauve, Inter-vehicle communica-tions based on ad hoc networking principles: the FleetNet project.Universitatsverlag Karlsruhe, 2005.

[15] F.Borgonovo, L.Campelli, M.Cesana, and L.Coletti, MAC for ad-hoc inter-vehicle network: services and performance, in Proc.IEEE 58th Vehicular Technology Conference (VTC 2003-Fall), vol. 5,Oct. 2003, pp. 27892793.

[16] A recursive solution to an occupancy problem resulting fromTDM radio communication application,Applied Mathematics andComputation, vol. 101, no. 1, pp. 13, Jun. 1999.

[17] IEEE standard for wireless access in vehicular environments(WAVE) - multi-channel operation, IEEE Std 1609.4-2010 (Revisionof IEEE Std 1609.4-2006), pp. 189, Feb. 2011.

Hassan Aboubakr Omar (S11) received the

M.Sc. (2009) degree in Engineering Mathemat-ics and the B.Sc. degree (2005) in Electron-ics and Communications Engineering, both fromCairo University, Egypt. From 2005 to 2008, hewas a Research Assistant at the Science andTechnology Research Center, at the AmericanUniversity in Cairo. Since 2010, he has beenworking toward a Ph.D. degree at the Depart-ment of Electrical and Computer Engineering,University of Waterloo, Canada. Mr. Omars cur-

rent research interest includes medium access control, routing, and roadside unit placement in vehicular ad hoc networks.

Weihua Zhuang (M93-SM01-F08) has beenwith the Department of Electrical and Computer

Engineering, University of Waterloo, Canada,since 1993, where she is a Professor and a Tier ICanada Research Chair in Wireless Communi-cation Networks. Her current research focuseson resource allocation and QoS provisioning inwireless networks. She is a co-recipient of theBest Paper Awards from the IEEE MultimediaCommunications Technical Committee in 2011,IEEE Vehicular Technology Conference (VTC)

Fall 2010, IEEE Wireless Communications and Networking Conference(WCNC) 2007 and 2010, IEEE International Conference on Communi-cations (ICC) 2007, and the International Conference on HeterogeneousNetworking for Quality, Reliability, Security and Robustness (QShine)2007 and 2008. She received the Outstanding Performance Award 4times since 2005 from the University of Waterloo, and the PremiersResearch Excellence Award in 2001 from the Ontario Government.Dr. Zhuang is the Editor-in-Chief of IEEE Transactions on Vehicular

Technology, and the Technical Program Symposia Chair of the IEEEGlobecom 2011. She is a Fellow of the IEEE, a Fellow of the CanadianAcademy of Engineering (CAE), a Fellow of the Engineering Institute ofCanada (EIC), and an elected member in the Board of Governors of theIEEE Vehicular Technology Society. She was an IEEE CommunicationsSociety Distinguished Lecturer (2008-2011).

Li Li (M03) received the M.Sc. (1990) degreein Electrical Engineering from Southeast Uni-versity, Nanjing, China and Ph.D (1993) degreein Electrical Engineering from University of Ot-tawa, Ottawa, ON, Canada. From 1993 to 1999,she was with Nortel Networks Ltd. as a sys-tem architect and then product manager. From1999 to 2003, she was the chief architecture at

SS8 Networks Inc. Since 2003, she has beenwith Communications Research Centre (CRC),Ottawa, ON, Canada, where she is a research

scientist. She has contributed previously to ITU-T and IETF standardworking groups, co-authored IETF RFC and has been awarded withseveral US patents. Her current research focuses on mobile tacticalradio networking and adaptive networks, with particular interests inmobile network optimization, networking protocols and algorithms, andperformance modeling of mobile networks. She has served as anAssociated Editor for Springers Journal on Peer-to-peer Networkingand Applications. Dr. Li has served as Co-chair for the IEEE PERCOMMP2P06-08 workshops, track co-chair for IEEE VTC2010-Fall, andsession co-chair for IEEE MILCOM08-11, and track co-chair for IEEEMILCOM 2012. She is an Associate Editor for IEEE Transactions onVehicular Technology.