TDMA Slot Reservation in Cluster- Based VANET

Ph.D. Proposal

DEPARTMENT OF COMPUTER SCIENCE

OLD DOMINION UNIVERSITY

2011

By

Mohammad Salem Almalag

malmalag@cs.odu.edu

Advisor

Dr. Michele C. Weigle

Committee Members

Dr. Hussein Abdel-Wahab

Dr. Stephan Olariu

Dr. Tamer Nadeem

Dr. Dimitrie C. Popescu

CONTENTS

ABSTRACT

1. INTRODUCTION 1.1. Motivation

1.2. Objective

2. BACKGROUND and RELATED WORK 2.1. TDMA

2.2. Clustering

2.3. Cluster Formation in MANETs

2.4. Cluster Formation in VANETs

2.5. Cluster Maintenance algorithms 2.6. Cluster-based MAC

3. ARCHITECTURE 3.1 Cluster head election and cluster formation

3.2 TDMA assignment

3.3 Intra-cluster communication

3.4 Inter-cluster communication

3.5 Cluster maintenance

4. EVALUATION

5. SUMMARY

6. PLAN

7. REFERENCES

Abstract

A vehicular ad-hoc network (VANET) is an autonomous and self-configuring network of

mobile hosts connected by wireless links. It is a highly dynamic network and has harsh channel

conditions. Cluster-based MAC scheme in VANET is needed to overcome the lack of specialized

hardware for infrastructure and the mobility to support network stability and channel utilization.

This work presents a MAC algorithm for vehicular ad-hoc networks using a new method for

TDMA slot reservation under a lane-based clustering formation algorithm. Our algorithm aims to

decrease collisions and packet drops in the channel. It also will make clusters stable which will

reduce the overhead of re-clustering and lead to an efficient hierarchical network topology.

Cluster performance will be measured in terms of average clusterhead duration, average cluster

member duration, and average rate of cluster head change. The proposed algorithm is also robust

to channel error and exhibits reasonable overhead.

1. INTRODUCTION

According on the Texas Transportation Institute [35], in 2009 the cost of traffic congestion in

the US was about $115 billion. This cost based on the wasted time and fuel. The total hours

wasted in traffic congestion in the US only is about 4.8 billion hours. Moreover, about 3.9 billion

gallons of fuel is wasted in the US. Besides the economic cost, traffic congestion leads to more

pollution in our cities.

The impact of traffic congestion on the economy and the environment motivated the research

and development of Intelligent Transportation Systems (ITS). Vehicular Ad-hoc Networks

(VANETs) are important component of ITS and useful for a wide variety of applications,

including both safety applications and non-safety applications [24]. The emergence of vehicular

networking has encouraged researchers to study how such communications could be used to

enhance drivers’ comfort. In the past several years, government agencies have partnered with car

manufacturers to design and prototype different types of non-safety-related vehicular

applications. Most of these applications rely on communication in the vehicular environment.

In the US, VANETs use 75 MHz of spectrum in the 5.850 to 5.925 GHz band specially

allocated by the U.S. Federal Communications Commission for Vehicle-to-Vehicle

communication (V2V) and Vehicle-to-Infrastructure communication (V2I) using Dedicated

Short Range Communication (DSRC) technology [19]. The main goal of VANETs is to share

information such as traffic data and road conditions to increase the safety of drivers.

The spectrum band is divided into seven 10 MHz channels, Figure 1. One channel is control

channel (CCH) and the other six channels are service channels (SCH). The CCH is assigned for

safety and control messages, while the SCHs are for both safety and non-safety messages.

According to IEEE 1609.4 standard [21], there is a Sync Interval that is divided into a CCH

Interval and a SCH Interval, Figure 2. Vehicles will be switching between the intervals to listen

to the channels. So, all vehicles will switch to the CCH during CCH Interval and to SCHs during

the SCHs Interval, even if there is no transmission in any channel.

Figure 1: US DSRC spectrum allocation

Figure 2: Division of time into CCH intervals and SCH intervals, IEEE 1609.4 standard.

1.1. Motivation

As an ad-hoc network, a VANET cannot rely on specialized infrastructure, such as Access

Point, to support network stability. Each node in a VANET is required to maintain its own

connectivity to other nodes in the network. With the large number of nodes and the lack of

routers, a flat routing scheme, where each node acts as a router, may cause serious scalability and

hidden terminal problems. One possible solution to these problems is hierarchical clustering. In

addition, using clustering can lead to more node coordination and fewer nodes interfering with

each other.

A cluster is a group of nodes that can communicate without disconnection and that identify

themselves to be part of a cluster. These nodes select a clusterhead to coordinate the

communication among themselves. Clustering in VANETs requires selecting a clusterhead that

produce a stable cluster. Cluster stability is also affected by the dynamics of vehicles in the

cluster. New vehicles joining the cluster and other vehicles leaving the cluster are going to

change the topology of the cluster. Having a simple clustering scheme in formation and

maintaining clusters will save a significant amount of time and channel bandwidth needed to

complete this process.

Since safety applications of vehicular communication have stringent reliability and delay

requirements, giving each vehicle the time to send safety messages without interfering with other

vehicles is required. TDMA is the technique that can be used to assign a unique time slots to

each vehicle in the cluster. The goal of any assignment scheme is to make the process of

assigning slots easy and straightforward. Also, as important are the safety messages, non-safety

messages need to be delivered even if we have a lot of safety messages.

1.2. Objective

The objective of this work is to propose new media access techniques that can be used for

clustering management and communications. This protocol integrates the centralization approach

of cluster management and a new scheme for slots reservations, where cluster members are

assigned by the clusterhead to local IDs. In this technique, all vehicles are able to tune to the

Control Channel (CCH) or the Service Channels (SCHs) if needed during the time cycle. In other

words, the time cycle is not divided into two different intervals, CCH Interval and SCH Interval.

The main objective of this work is to allow vehicles to send and receive non-safety messages

without any impact on the reliability of sending and receiving safety messages even if the traffic

density is high. Interestingly, in high density traffic, non-safety activities may have to be

completely shut down [39].

In this work, we propose a dynamic TDMA slot assignment scheme for cluster-based

VANETs. In this scheme, the collision-free intra-cluster communications are managed by the

clusterhead using a TDMA scheme. As a result, we encounter three important problems. These

problems are cluster formation, cluster maintenance, and slot assignment. In this work, we

propose three algorithms to solve the addressed problems.

Thesis statement: TDMA slot reservation in cluster-based VANET will increase the

performance of delivering non-safety messages with a little impact on safety messages delivery.

This performance will measured on the type of delivery delay and reception probability of safety

and non-safety messages over the current DSRC channel switching. We also measure the

overhead by counting the extra control messages required channel assignment and cluster

maintenance.

2. BACKGROUND and RELATED WORK:

2.1. TDMA

Time Division Multiple Access (TDMA) is a method used to enable multiple nodes to

transmit on the same frequency channel. It divides the signal into different time frames. Each

frame is divided into several time slots, where each node is assigned to a time slot to transmit

[40]. The length of the time slot may vary, based on the needs of the node assigned to it. For

example, if node i needs to transmit on the channel, it will use its time slot to transmit. The nodes

will transmit in rapid succession, each using its own time slot, Figure 3.

Figure 3TDMA frame structure showing a data stream divided into frames and the frames divided into time slots

The main advantage of TDMA is reducing interference between nodes. However, it adds slot

allocation complexity.

2.2. Clustering

A cluster is conceptual structure where a group of nodes, called cluster members, identify

themselves to be part of cluster. The node designated to be the clusterhead will broadcast packets

to other nodes in the cluster. All cluster members are within one-hop or multi-hop

communications range of the clusterhead, thus the potential cluster size increases with the

transmission range. Each clusterhead may act as a relay point for communication between the

clusters members [7]. Some clustering schemes assign another node to act as a relay node for

communication with other clusters [20], as shown in Figure 4. Also, the clusterhead will be in

charge of managing the TDMA slots assignment.

Figure 4: Two clusters communicating with each other through relay nodes (RN) that are assigned by the

clusterhead (CH) of each cluster.

2.3. Cluster Formation in MANETs

A Mobile Ad hoc NETwork (MANET) is a self-organizing wireless network of mobile

nodes, which can communicate without pre-existing infrastructure. MANETs can have either a

flat topology or a hierarchical clustered topology. In large networks, a flat topology faces

scalability issues [24]. Routing in MANETs requires flooding to find routes and in large

networks this flooding leads to severe congestion. This issue gets more severe as the network is

more mobile and the topology changes frequently.

One of the solutions for the scalability issue in MANETs is using cluster-based approach

[25]. Clustering provide scalability by creating a backbone network of nodes. It also provides

stability for dynamic networks.

There are several algorithms for cluster formation in MANET. The main task of the cluster

formation process is to elect a clusterhead. One of the clustering algorithms is Lowest-ID. The

Lowest-ID clustering algorithm [11] is based on selecting as the clusterhead the member with the

lowest ID, assuming each node has a fixed ID. Simply, each node broadcasts its ID to other

nodes in range. When a node receives the messages from other nodes, it determines the

clusterhead as the node with the lowest ID. This algorithm is very simple and stable for general

MANET applications.

Another clustering algorithm is Highest-Degree. The Highest-Degree algorithm [16] selects

the clusterhead based on the node connectivity to the other nodes in the same cluster. Each node

knows the number of other nodes in range and then broadcasts this knowledge to the others. The

node with the maximum number of neighbors is selected as the clusterhead. This algorithm is

one of the basic techniques for clusterhead formation in MANETs.

2.4. Cluster Formation in VANETs

Several algorithms have been introduced for cluster formation in VANETs. Some of these

algorithms were first developed for MANETs. For example, Lowest-ID algorithm seems to be

simple and stable for MANET; however, this algorithm is not always stable for VANET because

the movement of the vehicles is not considered. The Highest-degree algorithm that is also not

ideal for VANETs. It is not stable for VANETs due to the nature of the nodes movement. If the

clusterhead changes its behavior at any moment, the connectivity level could change

dramatically.

To have a good VANET clustering algorithm, we have to consider the characteristics of

VANET. The Utility Function algorithm [6] is a VANET clustering algorithm that performs

better than the previous two algorithms, Lowest-ID and Highest-Degree. This algorithm is based

on a multiple-metric weighting algorithm, considering speed, velocity and position. In the

process of the clusterhead selection, the closest position to the average and the closest velocity to

the average of all proximal vehicles are calculated along with connectivity level to determine the

most stable clusterhead. Periodically, each node broadcasts its status to other nodes in range.

When the node receives this information, it starts to evaluate each node by using the utility

function. The node with the highest value is chosen to be the clusterhead. In a highway

environment, this algorithm has been shown to provide better results than the classic MANET

algorithms. It puts the position and velocity, which are major VANET characteristics, into

consideration. However, it still ignores the traffic flow on the road. For example, in an urban

scenario where are many intersections, if the clusterhead is located on the leftmost lane, it has to

turn left even if most of the vehicles are going straight. In this case, the vehicles will need to

perform the process of clusterhead selection again.

Zaydoun and Mahmud [34] proposed grouping vehicles based on the mobility patterns.

Vehicles with close speeds will be grouped together in one cluster. But this might lead to having

clusters overlap.

We introduce a new cluster formation algorithm. This scheme aims to extend the life of the

clusterhead. We take advantage of knowing the exact lane of vehicles on the road and then

broadcast this knowledge to other nearby vehicles to determine the optimal clusterhead. Our

method of selecting the clusterhead is the key to achieving a more stable cluster.

2.5. Cluster Maintenance Algorithms

Based on dynamic nature of vehicles in VANETs and the changes in the network topology,

clusters must be updated frequently to maintain the stability of the network. Cluster maintenance

is a very important process in any clustering algorithm. It will be performed more often than

cluster formation. The cluster maintenance process should be done in a manner that it does not

add much communication overhead.

A significant amount of research on cluster maintenance has been done in MANETs [36-37].

VANETs, however, have different mobility characteristics than MANET. So, applying MANTE

algorithms in VANET is not always successful. Venkataraman et al. [36] proposed a clustering

algorithm that performs formation and maintenance in MANET. In the process of maintenance,

if the clusterhead is leaving the cluster, the clusterhead selection process will be performed

again. This will consume a lot of time and channel bandwidth. It is better to do the clusterhead

changes in a way that does not require a lot communications.

In our cluster maintenance algorithm, we introduce a new cluster maintenance algorithm that

solves the issues of the network topology changes. It addresses such as new cluster member(s)

joining, current cluster member(s) leaving the cluster, clusters merging, and clusterhead

changing. All the problems mentioned are going to be solved with least amount of overhead. We

aim to design an algorithm that makes changes in the topology unseen by most of the cluster

members. For example, when the clusterhead is about to leave the cluster, it will find a stable

clusterhead candidate. Then, the current clusterhead will switch local IDs with clusterhead

candidate. After switching, a new more stable clusterhead will take over with the need of

performing the clusterhead selection process again.

2.6. Cluster-based MAC Algorithms

A significant amount of research has been spent in developing new cluster-based MAC

protocols. Cluster-based MAC protocols are presented in [26-32]. Gunter et al. [27] proposed

schemes where the clusterhead takes on a managerial role and facilitates intra-cluster

communication by providing a TDMA schedule to its cluster members. Su and Zhang [32]

proposed a scheme where adjacent clusters are assigned different CDMA codes to avoid

interference between clusters. This work shows a substantial reduction in probability of message

delivery failure, when compared to the traditional 802.11 MAC. The disadvantages of this work

are that it uses two transceivers. It also reserve channel for specific tasks; so if there is no

activities on these channels, the channels will be wasted.

Much of the recent VANET research discussing cluster-based MACs and routing

schemes also present a low-maintenance clustering algorithm. Each of these algorithms works

essentially the same way, whereby nodes periodically transmit HELLO beacons to indicate their

present state. States can be one of the following: Undecided, Clusterhead, Cluster Member, and

sometimes Gateway. An undecided node will join the first clusterhead that it hears a HELLO

beacon from (or joins all clusterheads if Gateway nodes are allowed). If the node does not hear

from a clusterhead within a given time period, it will become a clusterhead itself. In addition,

protocols are introduced to deal with colliding clusters, which occurs when two clusterheads

come within range of one another. During a cluster collision, one clusterhead decides to give up

its status to the other. This technique is used by Su and Zhang [32] and Santos and Seed [33]

without regard for mobility. Gunter et al. [27] proposed a scheme where mobility is addressed

during cluster collision, whereby the winning clusterhead is the one with both lower relative

mobility and closer proximity to its members. Alternatively, Kayis and Acarman [31] address

mobility by first classifying nodes into speed groups, such that nodes will only join a CH of

similar velocity.

3. ARCHITECTURE

3.1. Clusterhead Election and Cluster Formation

We present a lane-based clustering algorithm designed to provide stability in cluster lifetime

for vehicular ad-hoc networks (VANETs). Stable clustering methods reduce the overhead of re-

clustering and lead to an efficient hierarchical network topology. During the creation of VANET

clusters, cluster members select one member to be the clusterhead. Fewer clusterhead changes

result in a more stable cluster. To achieve this goal, cluster members must select a member that

has the potential to be a clusterhead longer than other cluster members. Our method aims to

select a clusterhead based on the lane where most of the traffic will flow.

Our proposed algorithm is based on the assumption that each vehicle knows its exact lane on

the road via a lane detection system and an in-depth digital street map that includes lane

information, such as NAVTEQs NAVSTREETS [14]. A lane detection system is an important

element of many applications in VANETs, such the Extended Emergency Brake Light system

[9].

The Global Positioning System (GPS) is the primary system that is used for vehicle

localization. However, GPS has weaknesses when it comes to updating the positioning data and

when there is no signal. GPS has a 5 m error which is larger than the distance between lanes.

There has been much research on detecting and localizing lanes on roads. Several algorithms

have been proposed using different techniques. Some methods use GPS combined with a wheel

odometer [4], which provides relative localization as it detects changes in pose relative to the

previous pose. The advantage of the wheel odometer is that it is high resolution and simple to

use. It can typically detect movements on the order of tenths of millimeters. Other algorithms do

not use GPS, and instead use techniques such as vision [2], [3], [15], LIDAR (Light Detection

and Ranging) [10] and a beacon network using infrastructure to triangulate vehicle position [13].

Here we present the three clusterhead selection algorithms that we will be comparing our

work against:

Lowest-ID: As we explained in Section2, this algorithm is very simple and stable for general

MANET applications. However, in VANETs, the lowest ID clusterhead is not always the ideal

selection because the movement of the vehicles is not considered.

Highest-Degree: As we explained in Section 2, this algorithm is one of the basic techniques

for clusterhead formation in MANETs. However, it is not stable for VANETs due to the nature

of the nodes movement. If the clusterhead changes its behavior at any moment, the connectivity

level could change dramatically.

Utility Function: As we explained in Section 2, it puts the position and velocity, which are

major VANET characteristics, into consideration. However, it still ignores the traffic flow on the

road.

In our approach, we considered urban scenarios with intersections in the design of our

proposed algorithm. The clusterhead will be selected based on the flow of the majority of traffic.

For example, if the road has four lanes and three of them are going straight, the clusterhead

should be selected from the lanes that are going straight. This research applies the knowledge of

each vehicles lane and the flow direction of each lane, Figure 5.

In urban scenarios, traffic flow splits at each intersection. There are three main traffic flows at

an intersection: Left Turn (LT), Right Turn (RT), and No Turn (NT). The intersection may have

all three types of traffic flows or only some of them. LT is applied to the leftmost lane(s) if it

splits the traffic to the left, RT is applied to the rightmost lane(s) if it splits the traffic to right,

while NT is applied to the lane(s) in the middle if traffic goes straight.

Figure 5: Showing how the traffic flow is dividing vehicles ahead in different directions.

Our proposal follows the same general idea as the Utility Function [6], but employs a

different set of rules. We considered the effect of traffic flow, using lane information, on the

process of clusterhead selection. Each vehicle computes and broadcasts its Clusterhead Level

(CHL) along with its speed, position, etc. The vehicle with the highest CHL will be selected as

the clusterhead. If the selected clusterhead is a member of another cluster, the nodes will choose

the second highest, etc. CHL is defined as

iiii AVLADLtNCLCHL )( (1)

where NCL is the network connectivity level, ADL is average distance level, and AVL is

average velocity level. The computation of each of these metrics is described below.

A. Lane Weight: The key to our approach is to consider the lane a vehicle belongs to. We

apply to each metric a lane weight (LW) for each traffic flow (LT, RT and NT). The weight is

determined based on the total number of lanes on the roadway (TNL) and the number of lanes

for each traffic flow (NLTF). If the road has three different traffic flows, we will have three

different LWs. LW is defined as

kk NLTFTNL

LW 1

(2)

where k is the lane number. For example, if we have a road of four lanes where one lane is

LT, one lane is RT, and two lanes are NT, then the LW for each traffic flow will be LTLW = RTLW

= 0.25 and NTLW = 0.50. If a vehicle is on a lane with traffic flow LT, then it will use LTLW . In

the equations that follow, TFLW represents the LW for the traffic flow of the vehicle performing

the computation.

B. Network Connectivity Level: To compute the Network Connectivity Level (NCL), we

need to calculate the overall NCL and the NCL for each traffic flow. The overall NCL, , is the

maximum number of vehicles that are directly connected to vehicle i. This is defined as

j

i tjiAt ),,( (3)

where j is a potential neighboring vehicle. A(I, j, t) is equal to 1 if a connection between i and

j exists at time t, and is equal to 0 otherwise. At this point, we have calculated the connectivity

level between a vehicle and all other vehicles on the road. Now, we calculate the connectivity

level for a vehicle and the vehicles in the traffic flow it belongs to. The traffic flow connectivity

level β for vehicle i is defined as

TFj

TFi tjiAt ),,( (4)

where TFj is a vehicle in the same traffic flow as vehicle i.

After calculating both levels of network connectivity, we define the total connectivity level

for vehicle i on a lane belonging to traffic flow TF as

TFiii LWtttNCL )( (5)

where TFLW is the lane weight for the lane that vehicle i occupies.

C. Average Distance Level: To calculate the Average Distance Level (ADL), we calculate

the overall average absolute distance, i , between vehicles that are directly connected to vehicle

i. This is defined as

NV

yyxxj

ijij

i

22 )()(

(6)

where j is any vehicle connected to i, and NV is the total number of vehicles that are directly

connected to i in any lane.

Next, we calculate the average absolute distance, i , between vehicle i and other vehicles in

the same traffic flow, TF. This is defined as

TF

j

ijij

iNV

yyxxTF

22 )()(

(7)

where j is any vehicle in the same traffic flow and connected directly to i, and TFNV is the

total number of vehicles that are directly connected to i and in the same traffic flow.

The ADL for vehicle i in traffic flow TF is defined as

TFiii LWADL (8)

D. Average Velocity Level: We calculated the overall Average Velocity Level (AVL) as the

difference between the average velocities of all vehicles in range and the candidate clusterhead

velocity. Then, we add this to the product of LW and the average velocity for the traffic flow. The

overall AVL, i , for vehicle i is defined as

j

jii VelVel (9)

where j is a potential neighboring vehicle, and iVel is the velocity of vehicle i.

Now, we calculate the AVL, i , for vehicle i and the traffic flow it belongs to. This is

defined as

TFj

jii VelVel (10)

where TFj is a vehicle in the same traffic flow as vehicle i.

The AVL for vehicle i in traffic flow TF is defined as

TFiii LWAVL (11)

3.2. TDMA assignment

The presentation of our scheme involves several aspects of intra- and inter-cluster

communication. In turns, each of these communication regimens is partitioned into cases

depending on whether or not the cluster is single-hop. The clustering scheme we are using is

clusterhead based, where the consensus is dictated by the clusterhead (CH). We also assume that

all vehicles are equipped with GPS to ensure that vehicles have synchronized clocks.

To explain our scheme, we assume an N-car cluster where the cars are synchronous. The

transmission time is partitioned into consecutive, non-overlapping logical TDMA frames. We

assume the existence of k slotted service frequency channels (channels, for short) numbered from

0 through k-1. In each channel, the logical TDMA frames are aligned, i.e. begin and end at the

same time. Each logical frame contains 𝑁

𝑘 + 1 slots numbered from 0 through

𝑁

𝑘 . Also, all

slots are the same size; the slot size τ, is known to all vehicles in the cluster. We also assume one

narrow-band control channel − channel k − used by the vehicles and CH for disseminating status

and/or control messages; we anticipate the size of a mini-slot to be sufficient to allow individual

vehicles to communicate their status information. By virtue of synchronization, the vehicles

should know frame and slot boundaries. The number of vehicles (N) may change dynamically;

and the CH is responsible for updating N and for informing all vehicles in the cluster of the new

value of N. N, k and τ are technology-dependent. We anticipate:

o N to be at most 64

o k to be around 6

o τ to be about 6ms, the time to send a non-safety message (i.e., about 1500 bytes)

o a mini-slot on the control channel to contain at least 64 bytes; (i.e. assuming k=6,

this implies a slot size of 384 bytes)

Each vehicle in the cluster will receive a local ID. This local ID is a number from 0 to N-1.

The CH will have always ID 1, while ID 0 is reserved for a “virtual vehicle”. Other than

exceptionally, we do not expect all N vehicles in the cluster to be communicating

simultaneously; a vehicle is said to be active if it uses one of the communication channels. The

CH keeps a list of all the currently-active vehicles and disseminates this list to all the members of

the cluster using one of the mechanisms discussed below.

In each logical frame vehicle j, 0 ≤ 𝑗 ≤ 𝑁 − 1 , owns:

o channel j mod k during time slot 𝑗

𝑘 ; we also say that vehicle j owns the ordered

pair 𝑗 mod 𝑘, 𝑗

𝑘

o the j-th mini-slot of slot ( 𝑗

𝑘 -1) mod

𝑁

𝑘 , on channel k, as illustrated in Figure 6

below; we use the convention that (-1 mod 𝑁

𝑘 ) is the

𝑁

𝑘 -th slot of the previous

logical frame (see table 1);

Figure 6: Illustrating mini-slots on channel k; car j owns a mini-slot in the slot preceding its own slot

The basic idea is that in each logical frame, while idle, vehicle j listens to channel j mod k in

slot 𝑗

𝑘 and sets the corresponding byte in. Notice that the Integer Division Theorem guarantees

that if i ≠ j then either:

o i

k ≠

j

k or

o i mod k ≠ j mod k, or both

This confirms that no two cars own the same ordered pair. For an illustration, let N=61 and

k=6; as illustrated in the table 1, car 39 owns channel (39 mod 6) =3 during slot 39

6 =6, as well

as the 4-th mini-slot on the control channel in slot 6-1=5.

Table 1: Show the logical frame

channel\slot 0 1 2 3 4 5 6 7 8 9 10

5 5 11 17 23 29 35 41 47 53 59 unused

4 4 10 16 22 28 34 40 46 52 58 unused

3 3 9 15 21 27 33 39 45 51 57 unused

2 2 8 14 20 26 32 38 44 50 56 unused

1 1 7 13 19 25 31 37 43 49 55 61

0 reserved 6 12 18 24 30 36 42 48 54 60

mini-slot ownership

[6-11] [12-17]

[18-23]

[24-29]

[30-35]

[36-41]

[42-47]

[48-53]

[54-59]

[60-65]

[1-5]

We note that for any given N, there are:

)mod(1mod11*1*1 kNkNkNkNNkkk

NNk

k

N

unused channels in slot 𝑁

𝑘 ; in the previous example, there were 6 − 1 − 61 mod 6 =

5 − 1 = 4 unused channels in slot 10. These unused channels are

𝑁 mod 𝑘, 𝑁 mod 𝑘 + 1, 𝑁 mod 𝑘 + 2 , … , 𝑘 − 1

and will serve other purposes, as will be discussed later. Note also that in each logical frame

in addition to the 𝑘 − 1 − (𝑁 mod 𝑘) unused channels in slot 𝑁

𝑘 , channel 0 in slot 0 is reserved

for the CH to broadcast safety-messages. Consequently, on the control channel, mini-slot 0 is

always unused in slot 𝑁

𝑘 and mini-slots 𝑁 mod 𝑘, 𝑁 𝑚𝑜𝑑 𝑘 + 1, 𝑁 𝑚𝑜𝑑 𝑘 + 2 , … , 𝑘 − 1 of

slot 𝑁

𝑘 − 1 are unused. These unused slots/mini-slots will be put to work in various ways that

depend on the specific clustering regimen under investigation.

3.3. Intra-cluster communication

In the intra-cluster communication, we are looking at the single-hop cluster case first; multi-

hop cluster intra-cluster communication will be investigated later. Our goal is to design

lightweight communication protocols that avoid, to the largest extent possible, the involvement

of the CH in setting up connections between vehicles. As a single-hop cluster, all vehicles in the

cluster can communicate directly. Consequently, the cars do not need to discover their neighbors.

Each vehicle uses its own mini-slot to disseminate status information. The first byte of the

mini-slot can be used to encode 28 = 128 different situations; a few of them are listed below:

0 indicates that the vehicle is not communicating at the moment.

1 indicates that the vehicle is involved in communicating with some other vehicle in

the cluster; the binary encoding of the ID of the interlocutor follows in the second

byte.

2 indicates that the vehicle is involved in communicating with some other vehicle in

the cluster; the binary encodings of the IDs of the members of the multicast group

follows in the next 63 bytes.

125 confirmation of “Hello” message.

126 indicates that the vehicle will transmit a request during its upcoming slot (i.e.,

next slot).

127 indicates that the car will use its upcoming slot to transmit.

Different messages need to be transmitted inside the cluster. These messages are safety,

governance and non-safety messages. Also, the messages could be broadcasted or unicasted. We

will explain our scheme below.

Disseminating intra-cluster safety /governance messages: In this type of messages, the CH is

responsible for disseminating control messages. When a safety message needs to be broadcast to

the cluster the Ch interrupt the transmission/reception of non-safety data. Then, the CH will

change the status of its mini-slot to 127. Using the CH’s receiving slot, the CH will broadcast the

safety message to other vehicles in the cluster. This will be repeated on other available slots of

the next logical frame to achieve the effect of broadcasting to the entire cluster. Other vehicles,

in each logical frame, will be tuned in to one of these channels to pick up potential safety

messages. The CH may decide to disseminate safety messages to a subset of the vehicles in

which case it will also broadcast an N-bit vector, indicating which vehicles are targeted by the

message; if all bits are set, the message is a cluster-wide broadcast.

In addition to safety messages the previously-described mechanism is employed for cluster

governance messages including:

The updated value of N and multicast group setup requests.

Channels and slot times during which it has “office hours” and will listen to

individual requests.

Disseminating intra-cluster non-safety: For the non-safety messages, the CH uses its own

slot for non-safety data exchanges behaving as a usual vehicle.

Setting up an intra-cluster unicast communication: Unicast (a.k.a. point-to-point)

communications are set up without CH intervention. Suppose vehicle i wishes to talk with

vehicle j; setting up a connection between them is done as follows:

By tuning in to vehicle j’s own mini-slot, vehicle i determines whether or not car j is

available.

If so, vehicle i transmits a handshake packet on channel j mod k during time slot 𝑗

𝑘 .

Assuming no collision (i.e. some other vehicle may also want to talk to j), j will pick up the

handshake packet and will negotiate with vehicle i the parameters of the data exchange by

replying on channel i mod k during time slot 𝑖

𝑘 ; again, assuming no collision, the connection

between cars i and j has been set up. Now, both vehicles set up the first byte of their mini-slots to

indicate the status change. Once the connection has been set up, the two vehicles can

communicate either in i’s own slot or in j’s own slot or both, if possible. If vehicles i and j need

more than the basic amount of bandwidth, they need to seek permission from the CH to use one

or several extra unused time slots.

Setting up an intra-cluster multicast communication: Multicast (a.k.a. point-to-multipoint)

communications may be set up with or without CH intervention. Suppose vehicle j wishes to

establish a multicast group involving vehicles j, 𝑖1, 𝑖2, … , 𝑖𝑝 . If the multicast group is small,

vehicle j will attempt to send a handshake message to each of the remaining vehicles in the

multicast group. Once the group has been set up, vehicle j will transmit on channel j mod k

during time slot time 𝑗

𝑘 and all the other vehicles will listen to the channel. If the size of the

multicast group is large, vehicle j will send the CH a multicast group request consisting of its

own ID along with an N-bit vector with the bits corresponding to the multicast group set. Once

received by the CH, this multicast group setup request will be disseminated by the CH in the next

available logical frame, by all the modalities discussed above. Once the multicast group has been

set up, vehicle j will transmit to the group on channel j mod k during slot 𝑗

𝑘 .

3.4. Inter-cluster communication

The concept of clusters in VANET is used to avoid the interference between vehicles.

Vehicles in the same cluster should not suffer from this problem, as explained above. However,

interference could happen with other vehicles from different clusters, Figure 7. So, managing the

channels and the communications among the vehicles that are susceptible to interference must be

done for avoidance of collision and for passing messages to other clusters. This issue will be

studied and a solution will be developed in the future work.

Another type of inter-cluster communication is communicating with Road Side Units (RSU).

Communications in VANET are Vehicle-to-Vehicle (V2V) communication or Vehicle-to-

Infrastructure (V2I) communication. All we explained above is V2V, which does not cover RSU.

Communication with RSU is a V2I communication. Once the cluster gets in the range of a RSU,

vehicles in the cluster should able to communicate with RSU. That can be done by assigning a

slot to the RSU by giving it a local ID by the clusterhead. Once the local ID is assigned to the

RSU, vehicles can communicate with the RSU in the same fashion as intra-cluster

communication.

3.5. Cluster maintenance

Due to the movement of vehicles, the cluster will not stay the same for long time. Many

vehicles’ behavior may change the topology of the cluster. For example, a clusterhead leaving

the cluster, a new vehicle joining the cluster, a cluster member leaving the cluster, two clusters

get in range of each other, and a multi-hop cluster shrinks to a one-hop cluster. All these changes

in the cluster topology are going to be a dressed.

Figure 7: demonstrates the scenario where cluster members of two neighbor clusters might interfere with each other.

Clusterhead leaving the cluster: When a clusterhead is elected using our clusterhead

selection algorithm, the clusterhead will be locally assigned to ID 1. If the clusterhead noticed

changes in its behavior that might lead to being an unstable clusterhead, it will prepare for giving

up its responsibility as a clusterhead. This process can be done by choosing another cluster

member that is willing to serve as a stable clusterhead. Form the Figure 8, we have two-hop

cluster where the clusterhead is assigned to a local ID of 1. In this case, the clusterhead is ready

to resign. Before this happens, the clusterhead will pick one of the vehicles in area C to be a

clusterhead. The reason for that is to have a two-hop cluster even after the current clusterhead

leaves. The process will be done by switching the local ID between the current clusterhead, ID 1,

and the new clusterhead. If there are no vehicles in C besides the clusterhead, a new three-hop

cluster will formed.

From Figure 4, C includes CH and vehicles x and y. Vehicle with ID z is in between x and y

but not part of C. we want to proof that z is part of C and can be a clusterhead.

V – Set of vehicles in an area of 600m or less.

CH – is a vehicle that can communicate directly with all vehicles in V.

C – Set of vehicles that can do the job of CH.

C={x, y}, x- The farthest vehicle of C behind CH.

y- The farthest vehicle of C ahead of CH.

L – Set of vehicles that are behind CH and not part of C.

R – Set of vehicles that are ahead of the CH and not part of C.

Given:

∀ 𝑐 ∈ 𝐶 Adjust to all other vehicles in V

R.T.P:

Any vehicle z between x and y is part of C and can be a CH.

1. 𝑧 − 𝑙 ≤ 300 𝑚, where 𝑙 ∈ 𝐿

2. 𝑧 − 𝑟 ≤ 300 𝑚, where 𝑟 ∈ 𝑅

Proof 1:

∵ 𝑥 ∈ 𝐶 (1)

∵ 𝑥 ∈ [𝐶𝑠𝑡𝑎𝑟𝑡 , 𝐶𝑒𝑛𝑑 ] (2)

Where 𝐶𝑠𝑡𝑎𝑟𝑡 is the start point of C and 𝐶𝑒𝑛𝑑 is the end point of C.

∵ 𝐶 − 𝑅 ≤ 300 𝑚 (3)

Let’s assume we have r that belongs to R where:

𝐶 − 𝑟 ≤ 300 𝑚

From 1, we can replace C with x:

∴ 𝑥 − 𝑟 ≤ 300 𝑚 (4)

∵ 𝑥 − 𝐶𝑒𝑛𝑑 ≥ 𝑧 − 𝐶𝑒𝑛𝑑 (5)

∴ 𝑧 − 𝑟 ≤ 300 𝑚 I

Proof 2:

∵ 𝑦 ∈ 𝐶 (1)

∵ 𝑦 ∈ [𝐶𝑠𝑡𝑎𝑟𝑡 , 𝐶𝑒𝑛𝑑 ] (2)

Where 𝐶𝑠𝑡𝑎𝑟𝑡 is the start point of C and 𝐶𝑒𝑛𝑑 is the end point of C.

∵ 𝐶 − 𝐿 ≤ 300 𝑚 (3)

Let’s assume we have l that belongs to L where:

𝐶 − 𝑙 ≤ 300 𝑚

From 1, we can replace C with y:

∴ 𝑦 − 𝑙 ≤ 300 𝑚 (4)

∵ 𝑦 − 𝐶𝑠𝑡𝑎𝑟𝑡 ≥ 𝑧 − 𝐶𝑠𝑡𝑎𝑟𝑡 (5)

∴ 𝑧 − 𝑙 ≤ 300 𝑚 II

From I and II, z can be a cluster head.

Figure 8: clusterhead, ID 1, is about to resign. so, it will choose vehicle x, y, or z to a new clusterhead.

If there are no vehicles in C besides the clusterhead, a new three-hop cluster will be formed.

This process is done by selecting a vehicle from L and a vehicle from R that are in range of each

other. The selection of these two vehicles is done by the current clusterhead. After the selection

of the vehicles is completed, one of them will have local ID 1, after swapping local IDs with the

current clusterhead. If there are no vehicles in L and R that are in range, the clusterhead will

inform L and R that they will be two single-hop clusters. And the clusterhead selection algorithm

can be performed in both new clusters, L and R.

Once the cluster is created, it will not stay the same

In our scheme, the cluster is considered to be a single-hop. However, because of the dynamic

of vehicles in the road, two clusters may get in the range of each other. In this case, two clusters

will merge into one cluster; which may lead to three-hop cluster.

A new vehicle joining the cluster: We now describe the process of accepting a new car in the

cluster. Importantly, this operation is preempted by the broadcast of safety messages. When a

new car wishes to join the cluster, it will attempt to get the CH’s attention by transmitting in the

mini-slot of virtual car 0. Assuming that it was successful in getting the clusterhead’s attention,

the clusterhead will broadcast a “new Vehicle” message as discussed above and tentatively

assign the new-comer ID number 0. The new-comer will then transmit a “Hello” message in both

the clusterhead’s slot of the current frame as well as the virtual vehicle’s slot in the next frame;

this has the effect of broadcasting to the entire cluster. To confirm that the single-hop structure of

the cluster is preserved, each vehicle in the cluster will have to confirm receipt of the “Hello“. To

achieve this, each vehicle sets the first byte of its mini-slot to 125 and the clusterhead will read

all the mini-slots. If all the vehicles have confirmed receipt of the “Hello” message, the new

vehicles is accepted in the cluster and will be allocated the lowest available ID number. If

necessary, N has to be adjusted and the clusterhead will broadcast the updated information to the

cluster.

A cluster member leaving the cluster: When a vehicle is leaving the cluster, the clusterhead

places the local ID of the departing vehicle in the list of available IDs. Also, the clusterhead will

inform the other cluster members of the departure. These changes in the cluster may lead to

reassignment of the local IDs, which may lead to change the length of the transmission cycle.

Tow clusters get in range of each other: When to single-hop clusters are in range of each

other and in the same direction, these two clusters will merge creating one single-hop or multi-

hop cluster, Figure 9. The merging process will be performed by vehicles advertise themselves to

other vehicles in different clusters by sending a message that includes their IDs and the time their

TDMA frame start. This procedure will be performed every once in awhile, e.g. 1 seconds. If

there is a vehicle in a different cluster receives this message, this vehicle will try to communicate

with the sender based on the received message. During this communication, both cars will share

information about the number of vehicles in each cluster and the TDMA frame starting times.

Figure 9: two single hop clusters in range of each other, before merging.

The vehicle in the larger cluster will be the cluster head. Then, the old cluster head in the

large cluster will resign by swapping the IDs, ID 1, with the new cluster head. And the old

cluster head will be pushed into a stack for future needs, Figure 9. The sender vehicle in the other

cluster will swap its ID with the cluster head of its cluster. And the old cluster head will be

pushed into a stack for future needs. The reason of pushing the old clusterheads in a stack is that

they will be valuable if the two single-hop clusters split again. Once the whole large cluster is

formed, the vehicles of the old large single-hop cluster will have the same IDs, while the vehicles

of the old small single-hop cluster will receive new IDs. The new IDs will be the old IDs plus the

size of the old larger one-hop cluster. So, the low IDs will be in the old large one-hop cluster and

the high IDs will be in old the smaller one-hop cluster, Figure 10.

Figure 10: The new clusterhead for the new cluster that resulted of the merge.

When the vehicle with ID 6 in the old small cluster, high ID, wants to communicate with the

vehicle with ID 2 in the old large cluster, low ID, it will recognize that this vehicle belongs to the

large old cluster and is not in its communication range. So, Vehicle 6 will communicate to

vehicle 2 via the farthest vehicle from itself and the closest vehicle to vehicle 2, which is the

cluster head of the old small cluster, the vehicle with the ID 4.

A multi-hop cluster shrinks to a one-hop cluster: when vehicles in a multi-hop cluster, Figure

10, moving closer to each other, there is a chance of forming a single-hop cluster. In this case,

vehicles will keep the same IDs as the clusterhead remains a clusterhead. Also, vehicles in the

cluster will be able to communicate directly with other vehicles in the cluster, Figure 11.

Figure 11: One single-hop cluster, after shrinking of a multi-hop cluster

4. EVALUATION

We will design several simulations to test our techniques. We will use the ns-3 network

simulator [8], which is a follow-on to the popular ns-2 simulator. For VANETs, we will use

modules [1] that added well-known traffic mobility models, the Intelligent Driver Model (IDM)

[18] and the MOBIL lane change model [17].

We will be focusing on intra-cluster communications and the cluster formation. Inter-cluster

communication will covered as part of the cluster maintenance when two clusters are merging.

Here we list the evaluation plan:

1. Simulation environment:

We will test our algorithm in a highway environment with exits in the road.

Different levels of density (high, mid, low).

Different number of lanes.

2. Scenarios:

Formation of the cluster at the beginning of the simulation.

When new vehicles joining the cluster.

When cluster members leaving the cluster.

When the clusterhead is changing.

When the two clusters merging.

When a cluster is dividing.

3. Metrics:

Number of clusters: It is defined as the number of clusters that will be formed

after performing the cluster formation algorithm.

Number of times clusterhead changes.

Waiting time for packet transmission: This is to measure the average time

each non-safety packet needs to be transmitted.

Throughput.

Control overhead: This is to count the extra control messages required for

channel assignment and cluster maintenance.

We also need to evaluate the impact on non-safety applications and show that our technique

can support both safety and non-safety applications better than the current MAC proposed for

DSRC/802.11p.

5. SUMMARY

In this work, we designed a TDMA slot reservation scheme for cluster-based VANET, in

which the collision-free intra-cluster communications were organized by the clusterheads using a

TDMA scheme. We started by introducing a new stable cluster formation scheme in VANET.

With the availability of lane detection, lane direction and map matching, we were able to select

the most stable clusterhead. Then, we proposed some solutions for network topology changes.

This scheme should make the changes in the network topology harmless to the cluster. We also

explain a light weight slot reservation algorithm. Our work is based on making vehicles

guarantee receiving non-safety messages without affecting the safety messages. We also changed

the concept of having two intervals by having vehicles listening to the control channel and the

service channels during the same time cycle. This scheme should be easy and fast to maintain.

6. PLAN

1. Design and evaluate a cluster formation algorithm in VANET:

We designed a lane-based clustering algorithm to provide stability in cluster lifetime for

vehicular ad-hoc networks. The results show that the clusterhead stays with the majority of

the cluster members.

Status: Completed and published.

Mohammad S. Almalag and Michele C. Weigle, "Using Traffic Flow for Cluster Formation

in Vehicular Ad-hoc Networks," In Proceedings of the Workshop On User MObility and

VEhicular Networks (ON-MOVE). Denver, CO, October 2010, pp. 631-636.

2. Design and evaluate a TDMA slot assignment algorithm:

We aim to test and evaluate our algorithm for TDMA slot assignment. We will use

simulation to test the message exchange between cluster members in a high way

environment. We are going to see how much non-safety messages are going to be

transmitted without any impact on the safety messages delivery.

a) Analytical. (Task 1)

b) Experimental. (Task 2)

3. Design and evaluate a cluster maintenance algorithm:

We aim to test and evaluate our cluster maintenance algorithm. We also are going to study

the dynamic of the cluster members and the effect of changes in network topology. Part of

our algorithm is studying how the clusterhead changing is handled.

a) Analytical. (Task 3)

b) Experimental. (Task 4)

4. Evaluate the system:

We aim to evaluate our system, including the clustering algorithm and the TDMA

assignment algorithm, in realistic simulation environment. We will be running safety and

non-safety applications and measuring the throughput, and then compare it with

DSRC/802.11p where the channel switching is applied. (Task 5)

The timeline for completing the remaining tasks is shown in Table 2:

Table 2: The Tasks' Time Frame

Months

2011 2012

Apr. May Jun. Jul. Aug. Sep. Oct. Nov. Dec. Jan. Feb. Mar.

Task 1, 2

Task 3

Task 4

Task 5

Dissertation Writing

7. REFERENCES

[1] H. Arbabi and M. C. Weigle. Highway mobility and vehicular ad-hoc networks in ns-3. In

Proceedings of the Winter Simulation Conference, Baltimore, MD, Dec 2010.

[2] R. Aufrere, R. Chapuis, and F. Chausse. A model-driven approach for real-time road

recognition. Machine Vision and Applications, 13(2 2001):95–107, 2001.

[3] M. Bertozzi and A. Broggi. Gold: A parallel real-time stereo vision system for generic

obstacle and lane detection. IEEE Transactions on Image Processing, 7(1):62–81, January

1998.

[4] A. J. Dean and S. N. Brennan. Terrain-based road vehicle localization on multi-lane

highways. In Proceedings of the American Control Conference, pages 707–712, Piscataway,

NJ, USA, 2009. IEEE Press.

[5] P. Fan, J. G. Haran, J. Dillenburg, and P. C. Nelson. Traffic model for clustering

algorithm in vehicular ad-hoc network. In Proceedings of the Consumer Communications

and Networking Conference, volume 1, pages 168–172, 2006.

[6] P. Fan, J. G. Haran, J. F. Dillenburg, and P. C. Nelson. Cluster-based framework in

vehicular ad-hoc networks. In ADHOC-NOW, pages 32–42, 2005.

[7] M. Garg and R. K. Shyamasundar. A distributed clustering framework in mobile ad hoc

networks. In Proceedings of the International Conference on Wireless Networks, pages 32–

38, 2004.

[8] T. R. Henderson, S. Roy, S. Floyd, and G. F. Riley. ns-3 project goals. In Proceeding

from the 2006 workshop on ns-2: the IP network simulator, 2006.

[9] W. Holfelder. Vehicle safety communications in the US. In Proceedings of the 2007 ITS

World Congress, London, U.K., 2006.

[10] M. Jabbour, P. Bonnifait, and V. Cherfaoui. Enhanced local maps in a GIS for a precise

localisation in urban areas. In Proceedings of the 9th

IEEE Conference on Intelligent

Transportation Systems (ITSC 2006), 2006.

[11] M. Jiang, J. Li, and Y. Tay. Cluster based routing protocol. Internet Draft. draft-ietf-

manet-cbrp-spec-01, August 1999.

[12] S. Kuklinski and G. Wolny. Density based clustering algorithm for VANETs. Testbeds

and Research Infrastructures for the Development of Networks and Communities,

International Conference on, 0:1–6, 2009.

[13] C. D. McGillem and T. S. Rappaport. A beacon navigation method for autonomous

vehicles. IEEE Transactions on Vehicular Technology, 38(3):132–139, 1989.

[14] NAVTEQ. NAVTEQs NAVSTREETS Street Data Reference Manual, 3.4 edition,

October 2009.

[15] T. Oskiper, Z. Zhu, S. Samarasekera, and R. Kumar. Visual odometry system using

multiple stereo cameras and inertial measurement unit. In Proceedings Computer Vision and

Pattern Recognition, 2007.

[16] A. Ramalingam, S. Subramani, and K. Perumalsamy. Associativity based cluster

formation and cluster management in ad hoc networks. In Proceedings of the 9th

International

Conference On High Performance Computing, 2002.

[17] M. Treiber and D. Helbing. Realistische mikrosimulation von straenverkehr mit einem

einfachen modell. In Proceedings of the 16th

Symposium ”Simulationstechnik ASIM 2002”

Rostock, pages 514–520, September 2002.

[18] M. Treiber, A. Hennecke, and D. Helbing. Congested traffic states in empirical

observations and microscopic simulations. Phys Rev E Stat Phys Plasmas Fluids Relat

Interdiscip Topics, 62(2 Pt A):1805–24, 2000.

[19] US Department of Transportation. Standard Specification for Telecommunications and

Information Exchange Between Roadside and Vehicle Systems - 5 GHz Band Dedicated

Short Range Communications (DSRC) Medium Access Control (MAC) and Physical Layer

(PHY) Specifications. ASTM E2213-03, 2003.

[20] G. Wolny. Modified DMAC clustering algorithm for VANETs. In ICSNC ’08:

Proceedings of the 2008 Third International Conference on Systems and Networks

Communications, pages 268–273, Washington, DC, USA, 2008. IEEE Computer Society

[21] IEEE 1609.4 standard, - 2007

[22] Save Lives, Save Dollars. Prevent motor vehicle–related injuries

http://www.cdc.gov/injury/pdfs/cost-MV-a.pdf

[23] “The economic impact of motor vehicle crashes 2000,” National Highway Traffic Safety

Administration, May 2002.

[24] Mohammad Almalag, “Safety-Related Vehicular Applications,” In Vehicular Networks:

From Theory to Practice, Stephan Olariu and Michele C. Weigle, Eds. Chapman &

Hall/CRC, 2009.

[24] X. Hong, K. Xu, and M. Gerla, “Scalable routing protocols for mobile ad hoc net-

works,” Network, IEEE, vol. 16, no. 4, pp. 11–21, Jul/Aug 2002.

[25] Y. Gunter, B. Wiegel, and H. Grossmann, “Cluster-based medium access scheme for

vanets,” Intelligent Transportation Systems Conference, 2007. ITSC 2007. IEEE, pp. 343–

348, 30 2007-Oct. 3 2007.

[26] L. Bononi and M. Di Felice, “A cross layered mac and clustering scheme for efficient

broadcast in vanets,” Mobile Adhoc and Sensor Systems, 2007. MASS 2007. IEEE

Internatonal Conference on, pp. 1–8, Oct. 2007.

[27] Y. Gunter, B. Wiegel, and H. Grossmann, “Cluster-based medium access scheme for

vanets,” Intelligent Transportation Systems Conference, 2007. ITSC 2007. IEEE, pp. 343–

348, 30 2007-Oct. 3 2007.

[28] Z. Rawashdeh and S. Mahmud, “Media access technique for cluster-based vehicular ad

hoc networks,” Vehicular Technology Conference, 2008. VTC 2008-Fall. IEEE 68th, pp. 1–

5, Sept. 2008.

[29] F. Farnoud and S. Valaee, “Repetition-based broadcast in vehicular ad hoc networks in

rician channel with capture,” in INFOCOM 2009. The 28th Conference on Computer

Communications. IEEE, April 2008, pp. 1–6.

[30] B. Hassanabadi, L. Zhang, and S. Valaee, “Index coded repetition-based mac in

vehicular ad-hoc networks,” in Consumer Communications and Networking Conference,

2009. CCNC 2009. 6th IEEE, Jan. 2009, pp. 1–6.

[31] O. Kayis and T. Acarman, “Clustering formation for inter-vehicle communication,”

Intelligent Transportation Systems Conference, 2007. ITSC 2007. IEEE, pp. 636–641, 30

2007-Oct. 3 2007.

[32] H. Su and X. Zhang, “Clustering-based multichannel mac protocols for qos

provisionings over vehicular ad hoc networks,” Vehicular Technology, IEEE Transactions

on, vol. 56, no. 6, pp. 3309–3323, Nov. 2007.

[33] R. E. R.A. Santos and N. Seed, “Inter vehicular data exchange between fast moving road

traffic using ad-hoc cluster based location algorithm and 802.11b direct sequence spread

spectrum radio,” PostGraduate Networking Conference, 2003.

[34] Z.Y. Rawashdeh and S.M. Mahmud, "Toward Strongley Connected Clustering Structure

in Vehicular Ad hoc Networks", in Proceedings of VTC Fall, 2009

[35] D. Schrank, T. Lomax and S. Turner. “TTI’s 2010 URBAN MOBILITY REPORT”,

Texas Transportation Institute The Texas A&M University System. December 2010.

[36] G. Venkataraman, S. Emmanuel and S. Thambipillai. “Size-restricted cluster formation

and cluster maintenance technique for mobile ad hoc networks”, INTERNATIONAL

JOURNAL OF NETWORK MANAGEMENT, 2007; 17: 171–194

[37] L. Wang and S. Olariu. “Cluster Maintenance in Mobile Ad-hoc Networks”, Cluster

Computing, Vol. 8, Iss. 2-3, pp. 111, 2005

[38] J. A. Torkestani and M. R. Meybodi. “An efficient cluster-based CDMA/TDMA scheme

for wireless mobile ad-hoc network: A learning automata approach”, Journal of Network and

Computer Applications, 2010.

[39] Z. Wang and M. Hassan. “How Much of DSRC is Available for Non-Safety Use?”,

Proceedings of the fifth ACM international workshop on VehiculAr Inter-NETworking,

September 15, 2008, San Francisco, California, USA.

[40] P. Djukic and P. Mohapatra. “Soft-TDMAC: A Software TDMA-based MAC over

Commodity 802.11 hardware,” INFOCOM 2009, IEEE, pp. 1836-1844, April 2009