Reliable Broadcast of Safety Messages in Vehicular Ad hoc ......tions, that the proposed protocol...

Reliable Broadcast of Safety Messagesin Vehicular Ad hoc Networks

by

Farzad Hassanzadeh

A thesis submitted in conformity with the requirementsfor the degree of Master of Applied Science

Graduate Department of Electrical and Computer EngineeringUniversity of Toronto

Copyright c© 2008 by Farzad Hassanzadeh

Abstract

Reliable Broadcast of Safety Messages in Vehicular Ad hoc Networks

Farzad Hassanzadeh

Master of Applied Science

Graduate Department of Electrical and Computer Engineering

University of Toronto

2008

Broadcast communications is critically important in vehicular networks. Many safety

applications need safety warning messages to be broadcast to all vehicles present in an

area. In this thesis, we propose a novel repetition-based broadcast protocol based on

“optical orthogonal codes.” Optical orthogonal codes are used because of their ability

to reduce the possibility of collision. We present a detailed mathematical analysis for

obtaining the probability of success and the average delay. Furthermore, we propose to

use coding to increase network throughput, and “adaptive elimination” of potentially

colliding transmissions to further increase reliability. We show, by analysis and simula-

tions, that the proposed protocol outperforms existing repetition-based ones and provides

reliable broadcast communications and can reliably deliver safety messages under load

conditions deemed to be common in vehicular environments. We also show that the

proposed protocol is able to provide different levels of quality of service.

ii

Dedication

To my beloved sister,

Farnaz

iii

Acknowledgements

I wish to express my sincerest thanks to my supervisor, Prof. Shahrokh Valaee, whose

support made this work possible. His knowledge, guidance, and encouragement has been

of invaluable help to me in producing this work.

I would also like to thank my colleagues at Wireless and Internet Research Laboratory

(WIRLab) for their constructive comments.

This work was supported by AUTO21 Network of Centres of Excellence and in-kind

contributions of Mark IV Industries.

iv

Contents

1 Introduction 1

1.1 Motivation and Purpose . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.2 Infrastructure Networks and Ad Hoc Networks . . . . . . . . . . . . . . . 5

1.3 Collision Avoidance Operation: A Map of the Neighbourhood . . . . . . 8

1.4 Delay Requirement of Vehicular Communication . . . . . . . . . . . . . . 10

1.5 Overview of Vehicular Communications Standards . . . . . . . . . . . . . 11

1.6 Scope and Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

1.7 Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

2 Background 16

2.1 CSMA/CA-based Broadcast Protocols . . . . . . . . . . . . . . . . . . . 18

2.1.1 IEEE 802.11 Wireless LAN . . . . . . . . . . . . . . . . . . . . . 18

2.1.2 Broadcast Support Multiple Access (BSMA) . . . . . . . . . . . . 19

2.1.3 Broadcast Medium Window (BMW) . . . . . . . . . . . . . . . . 19

2.1.4 Batch Mode Multicast MAC (BMMM) . . . . . . . . . . . . . . . 21

2.1.5 Location Aware Multicast MAC (LAMM) . . . . . . . . . . . . . 23

2.1.6 TRAcking DEtection (TRADE) . . . . . . . . . . . . . . . . . . . 23

2.1.7 Distance Defer Transfer (DDT) . . . . . . . . . . . . . . . . . . . 24

2.1.8 Urban Multihop Broadcast (UMB) . . . . . . . . . . . . . . . . . 25

2.2 Reliable Reservation ALOHA (RR-ALOHA) . . . . . . . . . . . . . . . . 26

v

2.3 Repetition-based Broadcast Protocols . . . . . . . . . . . . . . . . . . . . 28

2.3.1 Synchronous p-Persistent Repetition (SPR) . . . . . . . . . . . . 29

2.3.2 Synchronous Fixed Repetition (SFR) . . . . . . . . . . . . . . . . 29

2.4 Other Protocols and Chapter Summary . . . . . . . . . . . . . . . . . . . 30

3 OOC-based Broadcast Protocol 31

3.1 Broadcast using Optical Orthogonal Codes . . . . . . . . . . . . . . . . . 31

3.1.1 CDMA, Positive Optical Systems, and OOC . . . . . . . . . . . . 33

3.2 Distributed Code Assignment . . . . . . . . . . . . . . . . . . . . . . . . 35

3.2.1 Code Information Response Window . . . . . . . . . . . . . . . . 36

3.3 Frame-Synchronous and Frame-Asynchronous . . . . . . . . . . . . . . . 39

3.4 OOC Code Generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

4 Analytical Performance Study 45

4.1 Probability of Success . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

4.1.1 Probability of Success for SPR . . . . . . . . . . . . . . . . . . . . 47

4.1.2 Probability of Success for SFR and OOC . . . . . . . . . . . . . . 47

4.1.3 Probability of Success and Interference Probabilities . . . . . . . . 49

4.2 Average Delay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

4.2.1 SPR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

4.2.2 SFR and OOC . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

4.3 Numerical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

5 Adaptive Elimination, Coding, and QoS Provisioning 68

5.1 Adaptive Elimination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

5.2 Coding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

5.2.1 Probability of Success with Coding . . . . . . . . . . . . . . . . . 71

5.3 Quality of Service . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

5.4 Appendix: Proof of Proposition 5.2.1 . . . . . . . . . . . . . . . . . . . . 75

vi

6 Simulation Results 78

6.1 Channel Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

6.2 Protocol Performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

6.2.1 Simulation Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

6.2.2 Probability of Success . . . . . . . . . . . . . . . . . . . . . . . . 81

6.3 Performance versus Distance . . . . . . . . . . . . . . . . . . . . . . . . . 88

6.3.1 Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

6.3.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

6.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93

7 Conclusions and Future Work 95

7.1 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96

Bibliography 97

vii

List of Tables

4.1 Possible codewords and their probabilities for one interfering user . . . . 53

4.2 Sample values of p1 obtained by generating OOC codes for various L and w 56

viii

List of Figures

1.1 Fatalities per 100 million vehicle-miles . . . . . . . . . . . . . . . . . . . 3

1.2 US Department of Transportation ITS architecture . . . . . . . . . . . . 4

1.3 Obstacle causes abnormal road situation . . . . . . . . . . . . . . . . . . 6

1.4 Chain collision avoidance . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

1.5 Intersection collision avoidance . . . . . . . . . . . . . . . . . . . . . . . . 7

2.1 Broadcast ad hoc medium access control protocols . . . . . . . . . . . . . 17

2.2 BMW protocol under ideal conditions for transmitting one data packet . 21

2.3 BMMM protocol under ideal conditions for transmitting one data packet 21

2.4 Blocking: N7 becomes blocked because N6 is not able to transmit CTS . 22

2.5 Far nodes are in yield state, only node B receives the broadcast message 26

2.6 Division of time into frames and timeslots for repetition-based broadcast 28

2.7 A transmission pattern and its binary representation. . . . . . . . . . . . 29

3.1 CDMA, OOC-based optical CDMA, and, OOC-based broadcast . . . . . 35

3.2 Network association phase . . . . . . . . . . . . . . . . . . . . . . . . . . 37

3.3 CIQ and CIR transmissions . . . . . . . . . . . . . . . . . . . . . . . . . 38

3.4 Timeline of CIQ and CIR transmissions of Fig. 3.3 . . . . . . . . . . . . 38

3.5 Circular and linear representation of a transmission pattern . . . . . . . . 40

3.6 Frame-synchronous and frame-asynchronous transmissions . . . . . . . . 40

3.7 Distribution of ‘1’s in a sample OOC . . . . . . . . . . . . . . . . . . . . 44

ix

4.1 Approximate and sample values of p1 for OOC . . . . . . . . . . . . . . . 57

4.2 Probability of failure of SPR . . . . . . . . . . . . . . . . . . . . . . . . . 63

4.3 Optimum average number of transmissions, w, for SPR . . . . . . . . . . 64

4.4 Probability of failure of SFR . . . . . . . . . . . . . . . . . . . . . . . . . 64

4.5 Probability of failure of OOC . . . . . . . . . . . . . . . . . . . . . . . . 65

4.6 Optimum number of transmissions, w, for SPR and OOC . . . . . . . . . 65

4.7 Optimum probability of failure for OOC, SFR, and, SPR . . . . . . . . . 66

4.8 Delay of successful transmissions for OOC, SFR, and, SPR . . . . . . . . 67

5.1 Both u1 and u2 are scheduled to transmit in the third timeslot . . . . . . 69

5.2 u2 disables its transmission in the third timeslot . . . . . . . . . . . . . . 70

5.3 Probability of success of SPR for different r’s, L = 64, N = 31, w = 6 . . 73

5.4 Probability of success of SFR for different r’s, L = 64, N = 31, w = 6 . . 74

5.5 Probability of success of OOC for different r’s, L = 64, N = 31, w = 6 . . 74

5.6 High and low priority messages with different number of transmissions . . 75

6.1 Map of roadway and cars. . . . . . . . . . . . . . . . . . . . . . . . . . . 81

6.2 Probability of failure versus w, for µp = 0.2. . . . . . . . . . . . . . . . . 82

6.3 Probability of failure for r = 2 versus w, for µp = 0.2. . . . . . . . . . . . 83

6.4 Probability of failure for r = 3 versus w, for µp = 0.2. . . . . . . . . . . . 83

6.5 Average delay versus w, for µp = 0.2. . . . . . . . . . . . . . . . . . . . . 84

6.6 Probability failure versus average delay . . . . . . . . . . . . . . . . . . . 84

6.7 Probability of success versus average load, for w = 7. . . . . . . . . . . . 86

6.8 Throughput versus average load, for w = 7. . . . . . . . . . . . . . . . . . 86

6.9 Providing different QoS levels with OOC. . . . . . . . . . . . . . . . . . . 87

6.10 Effect of Adaptive Elimination. . . . . . . . . . . . . . . . . . . . . . . . 88

6.11 Comparison of analytical and simulation results: Ps . . . . . . . . . . . . 88

6.12 Comparison of analytical and simulation results: Ps with coding . . . . . 89

x

6.13 Comparison of analytical and simulation results: Delay . . . . . . . . . . 90

6.14 Probability of success vs. distance from the receiver for L = 64 . . . . . . 91

6.15 Probability of success vs. distance from the receiver for L = 94 . . . . . . 92

6.16 Probability of success vs. distance from the receiver for different QoS levels 92

6.17 Simulation with and without capture . . . . . . . . . . . . . . . . . . . . 93

xi

Chapter 1

Introduction

1.1 Motivation and Purpose

According to the World Health Organization (WHO), road accidents annually cause

approximately 1.2 million deaths and 50 million injuries worldwide [1]. If preventive

measures are not taken, traffic accident death is likely to become the third cause of the

loss of disability-adjusted life years (DALY) 1 in 2020 from ninth place in 1990 [2].

However, fatalities caused by car crashes are, in principle, avoidable. 21,000 of the

annual 43,000 road accident deaths in the US are caused by roadway departures and

intersection related incidents [3]. This number can be significantly lowered by deploy-

ing local warning systems enabled by vehicular communications. Departing vehicles can

inform other vehicles of their intention to exit the highway and arriving cars at intersec-

tions can send warning messages to other cars traversing that intersection. Studies show

that in western Europe a mere 5km/hr decrease in average vehicle speeds could result in

25% decrease in deaths [1]. Policing speed limits will be notably easier and more efficient

using wireless communication technologies.

Governments and manufacturers have been increasingly investing to find new ways

1DALYs are the sum of the years of life lost due to premature mortality and the years lost due todisability.

1

Chapter 1. Introduction 2

to improve the safety of drivers, occupants, and pedestrians. Until recently, these efforts

have generally been following a passive and non-cooperative approach. A passive safety

system tries to minimize the casualties and cost of a collision by using devices such as air

bags and shock absorbers but it is not able to prevent collisions. Many manufacturers

have been deploying non-cooperative digital technologies into their vehicles wherein each

vehicle tries to reach maximum possible safety by making decisions based on information

that it has obtained individually. These systems include advanced braking systems and

cruise control systems.

In the United States, from 1960 to 2005, the rate of fatalities per vehicle-miles de-

creased from 5.1 fatalities per 100 million vehicle-miles (MVM) 2 to 1.5 [4]. As observed

from Fig. 1.1, in spite of a sharp decline form 1965 to 1995, this rate has been roughly

constant from 2000 to 2005. This may indicate that, although passive/non-cooperative

safety systems have been effective in decreasing fatalities, these approaches alone are no

longer capable of significantly reducing fatalities beyond the current state. Therefore,

we are urged to move from passive/non-cooperative safety systems to active/cooperative

safety systems, which can prevent accidents.

Active/cooperative safety systems are part of a broad range of emerging communica-

tions, electronics, and informatics technologies, unified under Intelligent Transportation

Systems (ITS), being developed to fundamentally enhance safety and productivity in sur-

face transportation. ITS technologies are designed to significantly improve road travel by

preventing accidents, decreasing congestion and gridlock, and enhancing traffic manage-

ment and enforcement. ITS development relies, at its core, on a communication platform

enabling fast and reliable communication in vehicular environments. Dedicated Short

Range Communication (DSRC) standard, adopted by IEEE and ASTM International3

(ASTM E 2213-03 [5]), provides the communication platform required by ITS [6]. The

2Vehicle-miles: Miles of travel by all types of motor vehicles as determined by the states on the basisof actual traffic counts and established estimating procedures.

3Originally known as the American Society for Testing and Materials


1960 1965 1970 1975 1980 1985 1990 1995 2000 20051

1.5

2

2.5

3

3.5

4

4.5

5

5.5

Year

Fat

aliti

es p

er 1

00 m

illio

n ve

hicl

e−m

iles

Fig. 1.1: Fatalities per 100 million vehicle-miles

importance of DSRC can be observed in ITS architecture illustrated in Fig. 1.2: DSRC

is the enabling technology for supporting both vehicle to vehicle (V2V) and vehicle to

roadside infrastructure (V2R) communication.

To accommodate the need of ITS for communication infrastructure, in 1999, the

U.S. Federal Communication Commission (FCC) allocated 75MHz bandwidth at 5.9GHz

[8] to public and private vehicular communication applications based on DSRC. The

75MHz bandwidth is divided into seven 10MHz channels. Among the seven designated

channels, one is assigned to V2V public safety communication (ch 172), one is assigned

to intersection public safety (ch 184), four channels are assigned to public safety and/or

private applications (ch 174, ch 176, ch 180, ch 182), and one channel is the control

channel (ch 178) used mainly for broadcast traffic. Our goal in this work is to provide a

Medium Access Control (MAC) protocol in ad hoc mode for broadcast communication.

Such a MAC protocol must be able to reliably deliver safety-critical messages. Due to

stringent delay requirements of safety traffic, transmission delay of a protocol designed

for vehicular communication must be very low. Furthermore, a vehicular MAC must be

capable of supporting mobility and effectively coordinating tens of sources of broadcast

traffic.


Remote Traveler Support

Personal Information

Access

Traffic Management

Information Service Provider

Emergency Management

Emissions Management

Toll Administration

Fleet and Freight Management

Commercial Vehicle

Administration

Maintenance and Construction Management

Transit Management

Archived Data

management

Fixed-Point to Fixed-Point Communication

Wid

e Ar

ea W

irele

ss (M

obile

) Com

mun

icat

ion

Vehicle to Vehicle Communication

Ded

icat

ed S

hort

Ran

ge C

omm

unic

atio

n

Vehicle

Emergency Vehicle

Commercial Vehicle

Transit Vehicle

Maintenance and Construction Vehicle

Roadway

Security Monitoring

Toll Collection

Commercial Vehicle Check

Parking Management

Centers

FieldVehicles

Travelers

Fig. 1.2: US Department of Transportation ITS architecture [7]

The rest of this chapter is organized as follows. In Section 1.2 we consider the choice

between infrastructure and ad hoc communication for safety systems. Characteristics

of the communication traffic of cooperative safety systems are presented in Section 1.3,

noting that knowledge of a relative map of neighbouring vehicles is effective in collision

prevention. Delay requirements of safety messages are described in Section 1.4. An

overview of vehicular communication standards is given in Section 1.5. The scope and

objectives of this work are presented in Section 1.6 and the main contributions of this

thesis are listed in Section 1.7.


1.2 Infrastructure Networks and Ad Hoc Networks

One of the main questions that must be explored for designing a MAC protocol is the need

for and the possibility of using infrastructure in the network. In this section, we identify

situations in which a vehicular network requires infrastructure and situations that can

be handled in ad hoc mode by considering some of the envisioned safety applications.

Line-of-Sight Collision Prevention

In this situation, all the vehicles, bikes, and pedestrians involved in a dangerous situation

are within line-of-sight of others. This has two consequences. First, drivers usually

identify the danger but in cases that result in accidents, recognition of dangerous situation

is so late that preventive actions taken by drivers are not effective in preventing a collision.

Therefore, safety systems must be capable of warning drivers noticeably faster than their

own ability to identify a dangerous situation. Second, the possibility of line-of-sight

communication alleviates the need for infrastructure. We provide two examples of line-

of-sight crash prevention in the following.

Obstacle Information Dissemination Fig. 1.3 shows a sample scenario in which part

of a road is closed due to an obstacle. Although in some cases, such as constructions,

(visual or radio) warning signals, set up in advance, inform the drivers to decelerate,

in many situations, such as a piece of fallen freight being in the road, warning signals

are absent. In the latter case, vehicular communication can be used to disseminate

information about abnormal road conditions. In Fig. 1.3 car 2 transmits a warning

message to all cars traveling behind it and within its communication range. It can be

observed that an ad hoc communication system is capable of delivering the warning

message and roadside units are not necessary.


Obstacle

1

2

3

4

5

6

1

2345

Fig. 1.3: Obstacle causes abnormal road situation

Preventing Chain Collisions A similar but more dangerous situation occurs when a

car is forced to stop due to a sudden event such as abnormal behavior of another vehicle

or an accident as illustrated in Fig. 1.4. Sudden deceleration in a highway may result in

a collision or even a chain of collisions. To avoid rear-end collisions, drivers rely on brake

lights of the car ahead of them to be able to stop in time. Needless to say, close distance,

high speed, slow driver reaction, poor visibility, and/or poor road condition may, and

occasionally do, prevent appropriate stopping and lead to accidents. Warning messages

can be useful in informing drivers faster than they would recognize the danger without

receiving warning messages. Warning messages can be particularly effective in avoiding

chain collisions because drivers’ reaction usually depends on the physical reaction of the

car immediately in front of them, i.e., a driver does not recognize the danger until the

car immediately ahead brakes. Radio warning messages, on the other hand, are free from

this limitation and, hence, can prevent chain collisions.

Non-Line-of-Sight Collision Prevention

An example of non-line-of-sight collision prevention is an intersection collision warning

system. Intersection collisions constitute a major category of traffic collisions that are

largely preventible. As illustrated in Fig. 1.5, however, line-of-sight communication is


Obstacle

1

2

3

4

5

6

1

2345

Fig. 1.4: Chain collision avoidance

not usually possible especially in more dangerous cases in which visual recognition is

difficult. In this situation, the warning message should be relayed from one vehicle to

another one by an intermediate node. In cases with high traffic density, relaying can

be done by another vehicle in the line-of-sight of the vehicles entering the intersection.

However, in most cases, a relaying vehicle is not present and a roadside unit is needed to

relay the warning message.

Fig. 1.5: Intersection collision avoidance

In situations that a roadside unit is not necessary, vehicular ad hoc network can pro-

vide effective and inexpensive networks to support ITS safety applications. Furthermore,

vehicular ad hoc networks can provide communication with lower delay by delivering mes-


sages from vehicle to vehicle, eliminating the delay caused by transmitting messages to

infrastructure and back to vehicles. Because vehicles moving in the same direction have

lower speed relative to each other than to roadside units, problems caused by mobility

are also alleviated in ad hoc networks.

Nevertheless, in cases in which line-of-sight communication is not possible and one

cannot rely on the presence of other vehicles to relay critical messages the presence of

roadside units is necessary. From the above discussion, we conclude that a hybrid solution

in which vehicular ad hoc networks provide the communication platforms in highways

and rural areas, and roadside units are utilized in intersections and urban areas, is best

suited to enable ITS safety message delivery.

1.3 Collision Avoidance Operation: A Map of the

Neighbourhood

In this section, we study the characteristics of the communication traffic in a vehicular

network in a channel dedicated to safety communication. We assume that safety systems

installed on each vehicle require a map of relative position of neighbouring vehicles. If

positions of neighbouring vehicles are known to the safety systems, many collisions can

be avoided. The distances to neighbouring vehicles being known, a safety system can

appropriately warn the driver of potentially dangerous situations. The map can be used

by the safety system to assist the driver in changing lane, entering and exiting highways

and main roads, avoiding unsafe close distance to the vehicle immediately in front, and

many other operations a driver needs to perform. If the velocity of the neighbours is

also known, each vehicle can predict future positions and avoid possible collision-prone

situations.

Building a local map in each vehicle requires that: 1) each vehicle be able to discover

its own absolute or relative position, and 2) vehicles be able to communicate position


information. Discovering the position of a vehicle can be done via GPS [9], radio rang-

ing techniques [10], and/or, radar. Our focus in this thesis is on providing the second

requirement, i.e., the design of a medium access protocol that is capable of delivering

position information messages, as well as other data.

At 100km/hr, a vehicle moves 6m (approximately the accuracy of GPS) in 216ms.

Therefore, update frequency of approximately 5 messages/second guarantees accurate

and up-to-date maps. However, if vehicles can accurately estimate their velocity, position

prediction can be used to alleviate the need for update and lessen the number of update

messages required. A well-designed medium access control layer must be capable of

successfully delivering messages with said frequencies.

Since vehicular update messages need to deliver limited information such as vehi-

cle ID, message ID, position, velocity, road condition, warning, etc, the size of these

messages is under a few hundred bytes. Location information, on the Earth’s sur-

face, in a spherical system with fixed r coordinate and 1 cm resolution can be deliv-

ered with log2(2π6.4× 106 m/10−2 m) + log2(π6.4× 106 m/10−2 m) = 62.81 ≤ 63 bits

where 6.4 × 106m is the Earth’s radius. Relative location information within 100m (in

a 200m×200m square centered at the reference point) in a Cartesian system with 1 cm

resolution can be delivered with 2 log2(200 m/10−2 m) = 28.6 ≤ 29 bits. Assuming each

vehicle transmits its position in absolute form and its velocity and the positions and

the velocities of vehicles immediately in front, behind, left, and right in relative to the

absolute position, 63 + 29 + 4(29 + 29) = 324 bits or 41 bytes need to be transmitted.

Adding 2 bytes for the ID of each vehicle, in total 51 bytes is needed. If about 49 bytes

are allocated for other uses, such as obstacle and its position, emergency car and its

position, emergency braking, etc, the length of the safety message is about 100 bytes.

Therefore, in vehicular communications, safety messages are short compared to data or

multimedia messages.

Assuming 30m separation between adjacent cars traveling in the same lane, in 300m of


one lane of a road, there are approximately 10 cars. If the radius of the neighbourhood

for which vehicles construct a local map is considered to be 300m, depending on the

number of lanes, tens of vehicles are within the communication range of a vehicle. Any

protocol designed for vehicular communication must be able to support tens of nodes in

the network.

Whether or not a map is used for collision avoidance, any medium access control

protocol must be scalable and flexible. A vehicular communication system in a crowded

highway spans hundreds of kilometers and thousands of cars. Although communication

is local and vehicles far from each other need not communicate for safety reasons, vehicles

that are physically close to each other must always be able to communicate regardless of

the logical structure of a protocol.

1.4 Delay Requirement of Vehicular Communication

An automatic safety system is successful if it can recognize a dangerous situation before

the driver of a vehicle does. For example, if the car immediately ahead suddenly stops,

the driver needs to detect the brake lights, decide that the brakes should be applied, and

move the appropriate muscles to apply the brakes. The mental processing time, i.e. , the

time from the moment an event occurs until the moment a decision is made, is between

500ms to 1.2s, depending on how unexpected the event is [11]. Noting that the warning

message alerting a driver, itself needs to be processed, we conclude that communication

delay must not exceed 100-200ms. This value is, henceforth, called the lifetime of a safety

message. Note that the lifetime value assumed here is in agreement with 5 messages per

second update frequency explained in the previous section.


1.5 Overview of Vehicular Communications Standards

IEEE 802.11a is adopted as the base MAC/PHY layer standard for DSRC [12]. The

modification to 802.11a, to make it suitable for vehicular communications and capable

of supporting ITS applications, is called IEEE 802.11p standard for Wireless Access in

Vehicular Environments (WAVE) [13]. WAVE is based on testing and analyses of wireless

communications in a mobile environment published in [5]. Draft 3.05 of 802.11p WAVE

is the most recent version [14].

According to IEEE 802.11p, a compliant vehicular communication network supports

both vehicular on-board units (OBU) and roadside units (RSU). An RSU is similar to

a wireless LAN access point and can provide communications with infrastructure [13].

Also, if required, an RSU must be able to allocate channels to OBUs. There is a third

type of communicating nodes called Public Safety OBU (PSOBU) which is a vehicle

with capabilities of providing services normally offered by RSUs. These units are mainly

utilized in police cars, fire trucks, and ambulances in emergency situations.

DSRC provides several channels (seven 10 MHz channels in North America) for com-

munications which are divided into two categories: a control channel and service channels.

The control channel is reserved for broadcasting and coordinating communications which

generally takes place in other channels. Although DSRC devices are allowed to switch

to a service channel, they must continuously monitor the control channel. There is no

scanning and association as in the conventional 802.11. All such operations are done via

a beacon sent by RSUs in the control channel. While OBUs and RSUs are allowed to

broadcast messages in the control channels, only RSUs can send beacon messages.

The complement of WAVE in higher layers is IEEE 1609 which is a family of standards

dealing with issues such as management and security of the networks [15]:

• 1609.1 Resource Manager: This standard provides a resource manager for WAVE,

allowing communication between remote applications and vehicles.


• 1609.2 Security Services for Applications and Management Messages.

• 1609.3 Networking Services: This standard addresses network layer issues.

• 1609.4 Multi-channel Operation: This standard deals with communications through

multiple channels.

1.6 Scope and Objectives

In this thesis, we propose and analyze a reliable medium access control protocol with

low delay for supporting safety message broadcasting in a vehicular network. We aim

at developing a protocol that is capable of satisfying low delay requirements of safety

messages while being able to deliver messages with high reliability.

A major difference between an ad hoc network in vehicular environment and a con-

ventional ad hoc network is that in vehicular networks, as discussed earlier, traffic is of

broadcast type; routine safety messages are issued from all vehicles several times per

second and are intended for all their neighbours. Transmission of safety messages must

be reliable and with very low delay. Conventional MAC protocols for ad hoc networks are

not designed to handle broadcast traffic from many nodes in the network. For example,

as explained in more detail in Chapter 2, in IEEE 802.11 no mechanism exists to reduce

the probability of collision for broadcast traffic. In IEEE 802.11, Request To Send (RTS)

and Clear To Send (CTS) packets are transmitted before unicast communications to

avoid collisions. It may seem straight forward to add RTS/CTS handshake to broadcast

communications as well. However, in vehicular communication, the length of broadcast

messages is short and comparable to that of RTS. Therefore, the probability of collision

is not significantly lower for RTS packets. The short length of messages also contributes

to inefficiency since the payload (safety message) is not significantly larger than the over-

head (RTS+CTS). Furthermore, RTS/CTS handshake needs to be performed with more

than one receiver to obtain the same reliability as that of unicast communication.


A solution to reliable delivery of safety broadcast messages is repetition-based broad-

cast. In repetition-based broadcast, the safety message is transmitted several times

according to a transmission pattern. While repetition increases the probability of suc-

cessful delivery compared to the ALOHA approach of IEEE 802.11, previously proposed

transmission patterns are random and have no mechanism for avoiding collision among

users.

In this work, after reviewing the literature in Chapter 2, we propose a new protocol

based on Optical Orthogonal Codes (OOC) to decrease collisions in Chapter 3. Opti-

cal orthogonal codes, originating from optical code division multiple access, have good

properties which, as we show with analysis and simulations, decrease the probability of

collision when used as transmission patterns in a repetition-based broadcast protocol. We

present a code assignment mechanism that assigns a transmission pattern to each user

which is unique within its two-hop neighbourhood. The generation of optical orthogonal

codes is also discussed.

A detailed analytical performance study is provided in Chapter 4, resulting in ex-

pressions for the probability of success and the average delay. The analysis is valid for

previously proposed repetition-based broadcast protocols, OOC-based repetition broad-

cast, and other repetition-based broadcast protocols with fixed number of transmissions.

Numerical results, obtained from the analysis, provide performance evaluation for dif-

ferent protocols and indicate that repetition broadcast based on OOC exhibits higher

probability of success and similar delay compared to other repetition-based protocols.

Following the analytical performance study, in Chapter 5, improvements to the pro-

tocol are presented. Adaptive elimination is introduced to further increase the reliability

of the protocol by eliminating potential collisions. Furthermore, the use of coding in

the proposed protocol is explained. Coding helps increase the throughput of the net-

work. Since QoS provisioning is a desirable feature of vehicular networks, a method for

providing different QoS levels is also introduced.


In Chapter 6, simulation results are presented to show the performance of the pro-

posed protocol, in an environment with a traffic load similar to that required by the ITS

safety applications, and compared to the performance of other repetition-based protocols.

Simulation results demonstrate the superior performance of OOC-based broadcast and

show that its performance is improved by adaptive elimination and coding.

Finally, in Chapter 7, we end this work with our concluding remarks and provide

directions for future work.

1.7 Contributions

The main contribution of this work is the design of a novel medium access control pro-

tocol for broadcast communication in vehicular communication networks. The proposed

protocol is suitable for transmission of short routine safety messages issued by vehicles

to inform others of their position and other useful information. The protocol makes use

of good correlation properties of optical orthogonal codes to avoid collisions.

The following lists the contributions of this work, the chapter in which they were

presented, and any publication reporting them.

• MAC Design based on OOC: Introduced and studied OOC-based repetition

protocols for broadcast communications (Chapter 3 and [16], [17])

• Performance Study: Developed a generalized analytical framework not only ca-

pable of analyzing OOC-based broadcast but also previously proposed repetition-

based broadcast protocols and any repetition-based broadcast protocol with a fixed

number of repetitions (Chapter 4 and [16]) and presented simulation results (Chap-

ter 6 and [16], [17])

• Coding, QoS, and Adaptive Elimination: Introduced and studied methods

to improve reliability, throughput, and quality of service provisioning in repetition-


based broadcast (Chapter 5 and [16], [17])

Chapter 2

Background

In this chapter, we look at previous works focusing on the reliability of broadcast commu-

nication in ad hoc networks (vehicular or otherwise). Protocols applicable to vehicular

communications can be divided into five categories as shown in Fig. 2.1. Among these

categories we discuss protocols from the first three categories: Carrier Sense Multiple

Access/Collision Avoidance (CSMA/CA)-based protocols, reservation-based protocols,

and repetition-based protocols.

CSMA/CA-based protocols are modifications of IEEE 802.11 that extend the use of

Collision Avoidance (CA) mechanism, which in IEEE 802.11 is used only for unicast

communications, to broadcast and are discussed, along with IEEE 802.11, in Section 2.1.

These modifications incorporate RTS/CTS/ACK handshake into IEEE 802.11 broadcast

to force hidden terminals into yield state in order to avoid collisions. They may require

the handshake to be performed with all receivers, a subset of receivers, or only one

receiver.

Reliable Reservation ALOHA (RR-ALOHA), a slot reservation protocol suggested

for vehicular networks [18], is discussed in Section 2.2. In RR-ALOHA users reserve a

timeslot in a frame for transmission.

Synchronous p-Persistent Retransmission (SPR) and Synchronous Fixed Retransmis-

16

Chapter 2. Background 17

Broadcast Ad hoc MAC protocols

Reservation-based protocols

BTMA-based protocols

Cluster-based protocols

RR-ALOHA

Repetition-based protocols

CSMA/CA-based protocols

SPR SFR OOC

IEEE 802.11

LAMM

BMWBSMA BMMM

DDTTRADE UMB

Fig. 2.1: Broadcast ad hoc medium access control protocols

sion (SFR) [19], repetition-based protocols proposed to solve broadcast problems, are

discussed in Section 2.3. The proposed protocol in this thesis is also a repetition-based

protocol. Analytical and simulation results are presented for SPR and SFR along with

the proposed protocol in later chapters.

The main challenge in cluster-based protocols is not the medium access control pro-

tocol but the formation of stable clusters. Therefore, these protocols are not considered

here. Also, Busy Tone Multiple Access (BTMA) protocols, such as [20], cannot be used

in the single control channel of DSRC and, therefore, not discussed in this chapter.

For each protocol described in this chapter, we discuss its suitability for broadcast in

vehicular environments.


2.1 CSMA/CA-based Broadcast Protocols

2.1.1 IEEE 802.11 Wireless LAN

In IEEE 802.11, Request To Send (RTS)/ Clear To Send (CTS) handshake is the main

means of combating the hidden terminal problem. However, broadcasting nodes do not

generally use RTS/CTS packets [21]. The lack of RTS/CTS handshake for broadcast

communication results in unreliable and inefficient communication. The situation is worse

in vehicular environments when there are tens of nodes within single-hop communication

range.

For unicast communications, acknowledgement (ACK) packets are transmitted by

receiving nodes to ensure recovery if transmission has not been successful. For broadcast

communication, however, acknowledgement is not required in IEEE 802.11 [22]. This

increases the unreliability of transmission of safety broadcast messages.

Lack of acknowledgements also increases the risk of network instability. Since the

transmitting node cannot detect an unsuccessful transmission, it fails to change the size

of its Contention Window (CW) [23]. Therefore, under heavy load, the size of contention

window is not increased to reduce the probability of collisions and the possibility of

instability is increased.

Although WAVE is built upon 802.11 wireless LAN, the requirements for reliable and

low delay communication, especially broadcast, are not achieved by using the current

standard without appropriate modification and enhancements as commonly observed in

the literature [6], [8], [20], [24]– [32].

Furthermore, unicasting such short messages as in vehicular safety communications

(see Section 1.3), with 802.11 approach is very inefficient. According to a model devised

by Bianchi [33], the maximum bandwidth utilization of 802.11a with RTS/CTS hand-

shake, at 54 Mb/s, with payload size of 100 bytes is less than 7% [8]. Multiple RTS/CTS

handshakes, as proposed by some of the protocols discussed in this chapter, will further


decrease the efficiency.

2.1.2 Broadcast Support Multiple Access (BSMA)

A simple extension to IEEE 802.11 which adds RTS/CTS signaling to broadcast packets

is presented in [34]. The protocol works as follows. After going through carrier sensing

and a contention phase, the source node broadcasts an RTS packet and waits a certain

time for a CTS, which, if received, triggers DATA transmission. With this approach,

however, the collision of CTS packets is very likely. To solve the problem of collision of

CTS packets, the authors suggest:

• Using a radio with capture capability so that one CTS packet is successfully re-

ceived. In this case, it is assumed that one of the CTS packets is sufficiently stronger

than others and can be received with a capture-enabled radio. This assumption,

however, is unrealistic since CTS packets are likely to have similar received power.

• The source node’s radio detects the busy period resulting from a collision as a CTS

and starts transmitting if it detects a busy period in the channel after transmitting

RTS. In this case the main purpose of CTS packets, which is to force hidden

terminals to yield, is overlooked. Hidden terminals cannot distinguish a busy period

resulted from colliding CTS packets from any other collision.

In Broadcast Support Multiple Access (BSMA) [35], NAK signaling is added to the above

protocol. The collision of control packets, however, remains unresolved.

2.1.3 Broadcast Medium Window (BMW)

The fundamental idea behind Broadcast Medium Window (BMW) [27] is to increase

the reliability of broadcast transmission by transmitting the broadcast packet to every

neighbour using RTS/CTS/ACK transmissions in a round robin fashion. If a neighbour


has already received the packet, the data packet is not transmitted. All transmissions in

BMW also follow CSMA/CA rules.

In BMW, each node maintains a list of its neighbours. Consider a node wishing to

broadcast a queue of packets to its neighbours. The broadcasting node chooses one of its

neighbours as the current receiver and transmits an RTS to that node. The receiving node

transmits a CTS in which it includes the sequence number of the packet up to which all

packets are received successfully. Then, the broadcasting node transmits packets starting

from the sequence number indicated in the CTS packet using DATA/ACK transmissions.

Meanwhile, all other nodes, other than the current receiver, are also receiving the packets

and store all successfully received packets. The packet transmission to the current receiver

continues until a new packet, which has not previously been transmitted to any other

node, is transmitted to the receiver. At this point the broadcasting node chooses another

neighbour and continues this process until all packets are transmitted.

BMW is only efficient when the transmission queue contains several packets. There-

fore, it is only suitable for burst traffic, or delay insensitive traffic which can be shaped

into bursts of packets. When traffic arrives in the form of single packets in the presence of

n neighbours, in ideal channel with no collision, as illustrated in Fig. 2.2, there are n con-

tention periods, n RTS transmissions, n CTS transmissions, n ACK transmissions, and

one DATA transmission. High number of control packets and contention periods induce

unacceptably high delays, which is increased, even more, in realistic channel conditions

with collisions, with back-offs, and retransmissions. Since traffic arrival in vehicular en-

vironments is in form of single packets with low delay requirements (in the range of

100ms) and there are tens of nodes in the network, BMW is not suitable for vehicular

communications.


Contention Phase

Contention Phase

Contention Phase

Contention Phase

CTS1RTS1 DATA

RTS1CTS1

ACK1 CTS2RTS2

RTS2CTS2

ACK2 CTSnRTSn ACKn...

RTSnCTS2

CTS1RTS1 DATA

RTS1CTS1

CTS2RTS2

RTS2CTS2

CTSnRTSn ...... ACK1RAK1 ACK2RAK2 ACKnRAKn...

...RTS2

CTS2

Fig. 2.2: BMW protocol under ideal conditions for transmitting one data packet

Contention Phase

Contention Phase

Contention Phase

Contention Phase

CTS1RTS1 DATA

RTS1CTS1

ACK1 CTS2RTS2

RTS2CTS2

ACK2 CTSnRTSn ACKn...

RTSnCTS2

CTS1RTS1 DATA

RTS1CTS1

CTS2RTS2

RTS2CTS2

CTSnRTSn ...... ACK1RAK1 ACK2RAK2 ACKnRAKn...

...RTS2

CTS2

Fig. 2.3: BMMM protocol under ideal conditions for transmitting one data packet

2.1.4 Batch Mode Multicast MAC (BMMM)

Batch Mode Multicast MAC (BMMM) [31] is a protocol that decreases the number of

contention periods of BMW. This is done by sequentially transmitting a RTS to all neigh-

bours and receive a CTS. DATA is transmitted after receiving the last CTS packet. After

transmitting the DATA packet, the broadcasting node sequentially requests acknowledge-

ments from the receivers via Request for Acknowledgement (RAK) packets. The ideal

operation of the protocol, where channel is perfect and none of the receiving nodes are

in yield state, is illustrated in Fig. 2.3.

By decreasing the number of contention periods from n to 1, BMMM decreases the

delay. However, BMMM does not decrease the significant overhead produced by control

packets. Furthermore, note that the yield duration indicated in CTS packets is signif-

icantly increased. For example, receiving node 1 in Fig. 2.3 forces all its neighbours

to yield state for the whole duration of transmissions to all nodes while its neighbours

need to be in yield state only for the duration of DATA and RAK1. This unnecessary

increase in yield state duration results in increased inefficiency. Probability of blocking

and blocking propagation [36], also, increases because of long yield states. Blocking is

illustrated in Fig. 2.4. N7 becomes blocked due to the following events:


• At t1, N1 transmits an RTS to N2. N3, N4, and N5 yield upon reception.

• At t2, N2 transmits a CTS to N1. N6 yields upon reception.

• At t3, N7 transmits an RTS to N6. N8, N9, and N10 yield upon reception.

• At t4, N7 does not receive a CTS and goes into exponential back-off.

While transmission from N7 would not interfere with any ongoing transmissions, N7 is

prohibited from transmission. Other nodes in yield state may also result in blocking of

other nodes. Longer durations indicated in RTS and CTS packets cause longer yield

states and, therefore, the probability that a node in yield state becomes the recipient of

an RTS is increased and so is the probability of blocking of the node transmitting the

RTS.

N1

N9

N7N6

N3

N5

N2N4

N10

N8

t2:CTS

t1:RTS

t2:CTS t3:RTS

Fig. 2.4: Blocking: N7 becomes blocked because N6 is not able to transmit CTS

Another problem with long yield durations indicated in CTS/RTS arises from high

mobility in vehicular environments. A node outside the range of a RTS/CTS packet may

move into the transmission range, not having heard the RTS/CTS. This node may po-

tentially start transmission that collides with ongoing transmissions. When the duration

indicated in RTS/CTS is larger, the probability of topology change during that time is

higher and, therefore, the probability of collision due to topology change is also higher.


2.1.5 Location Aware Multicast MAC (LAMM)

Location Aware Multicast MAC (LAMM) [31] improves the performance of BMMM by

introducing cover sets. A cover set of the receivers of a broadcast message, transmitted

by a certain transmitter, is a subset of receivers whose radio coverage, is the same as

the radio coverage of all receivers. Therefore, if only nodes in the cover set send CTS

packets, the effect is the same as when all receiving nodes send CTS. Furthermore, only

nodes in the cover set receive RAK and transmit ACK. Determining the cover set requires

position estimation which can be obtained via GPS. While LAMM decreases the number

of control transmissions, the amount of decrease is not discussed in [31]. The duration

indicated in CTS/RTS control packets is also still larger than that of BMW. It might be

possible to modify LAMM to adapt it to vehicular environment and decrease its overhead

and CTS durations significantly. Exploring this possibility requires further research.

The next three protocols introduce methods of finding one receiver and perform hand-

shake with only that node instead of all the recipients to decrease the overhead caused

by control packets.

2.1.6 TRAcking DEtection (TRADE)

TRAcking DEtection (TRADE) protocol introduced in [37], is a protocol which uses

location information to find the farthest node in each direction to eliminate the broadcast

storm problem [22] caused by flooding in multi-hop broadcast. However, this protocol

can be modified to increase the reliability of single-hop broadcast compared to IEEE

802.11. By modifying a simplified version of TRADE we obtain a protocol that works

as follows. The broadcasting node chooses the farthest node within its communication

range, and unicasts the broadcast packet to that node in each direction on a straight road

using RTS/CTS/ACK handshakes. Approximating the positions of vehicles as being on a

straight line, if the farthest vehicle receives the message without collision, so do all vehicles


between the source and the farthest vehicle. This approach, however, has drawbacks.

• TRADE assumes the positions of vehicles are known by transmitting GPS messages.

This can be problematic since GPS messages also need to be broadcast routinely.

• Since collisions are not the sole reason of packet loss, an intermediate node may

still lose the packet due to noise, fading, or a large vehicle blocking line-of-sight

while the farthest node receives it successfully.

• A broadcast message is of much more importance to the nearest vehicle than it is

to the farthest. Therefore, if the farthest node is in yield state, the message is not

transmitted and its lifetime may expire, while the closet node, which is its more

important target, would be able to receive it without interference.

2.1.7 Distance Defer Transfer (DDT)

Distance Defer Transfer (DDT) [37] is proposed to eliminate the need for a priori position

information in TRADE, which is one of its drawbacks. DDT, similar to TRADE, is

designed to combat the broadcast storm problem. However, it can be modified to increase

the reliability of broadcast communication. In this modification of a simplified DDT

protocol, the source node broadcasts the RTS packet in which the position of the sender

is included. Each of the receiving nodes start a timer whose value is a decreasing function

of their distance from the source. When a node’s timer expires, if no other node has

transmitted a CTS, that node transmits a CTS. This approach solves the first problem

mentioned above, but the second and third problems remain unresolved. Furthermore,

the likely collision of CTS’s (or re-broadcasts in the original protocol) when the distance

of two nodes is roughly the same from the source is ignored.


2.1.8 Urban Multihop Broadcast (UMB)

The design of Urban Multihop Broadcast (UMB) protocol [38] is aimed at finding the far-

thest node without a priori neighbourhood information. This protocol, basically, works

as follows. The broadcasting node sends a Request-To-Broadcast (RTB) packet in which

a desired direction of transmission is indicated. Each receiving node in the correct direc-

tion, transmits a jamming signal (black-burst), duration of which is proportional to its

distance from the source and is an integral number of Short InterFrame Spaces (SIFS).

After transmitting black-burst, each node listens to the channel. If it is the last node

transmitting black-burst, it is chosen as the farthest node and continues to transmit a

Clear-To-Broadcast (CTB) packet. The broadcasting node, then, transmits the DATA

packet and receives an ACK packet from the farthest node. It is possible for two or

more nodes to transmit black-bursts with the same length. In this case, they all trans-

mit CTB packets leading to collision. If the broadcasting node senses the channel to be

busy but is unable to decode the transmission, it assumes that two or more CTB packets

have collided. In this case, it issues another RTB packet. In responding to the second

RTB, only nodes that have transmitted after receiving the previous RTB compete for the

channel by transmitting black-bursts. This procedure is iterated until a node successfully

transmits CTB or a certain number of iterations is reached in which case the nodes that

have transmitted the last CTB packets enter a random resolution phase which, if nec-

essary, is repeated a certain number of times. If random resolution is also unsuccessful,

the broadcasting node restarts the transmission from the beginning. Starting the node

selection from the beginning may happen RETmax times, after which the broadcasting

node enters the 802.11 back-off mode.

While UMB manages to decrease the overhead, it puts too much emphasize on finding

the farthest node while, as mentioned before, in vehicular communications, the closest

node is the most important receiver. By the time the contention between far nodes is

resolved, the life time of the message may have expired.


When far nodes are in yield state and unable to transmit black-bursts, close nodes

respond to RTB packets with CTB and DATA/ACK transmissions follow. In this case,

far nodes, due to ongoing transmissions, are unable to receive the broadcast packet. This

is illustrated in Fig. 2.5. Assume that communication between nodes H and I takes place

first. Nodes C, D, E, and F are unable to reply to the RTB issued by node A. Node B

responds to node A with a CTB packet and DATA/ACK transmission follows. Nodes

C, D, E, and F do not receive the broadcast message due to the ongoing transmission

between H and I. Immediate transmission to node B and not delaying the transmission

for nodes C, D, E, and F is beneficial. However, the problem is that node A assumes it

has successfully transmitted the broadcast message and will not try to reach other nodes

by retransmissions. In spite of the problems mentioned above, UMB is logically a good

option among CSMA/CA-based protocols for transmitting broadcast packets because of

its relatively low overhead and the increase in reliability.

A B C D E F G H I J K L M N O

RTS

CTS

RTB

CTB

H I A

Fig. 2.5: Far nodes are in yield state, only node B receives the broadcast message

2.2 Reliable Reservation ALOHA (RR-ALOHA)

Reliable Reservation Aloha (RR-ALOHA) [30] is a medium access control protocol based

on Reservation ALOHA (R-ALOHA) which works only in an environment where all

nodes are able to receive transmitted signals from all other nodes [29]. To eliminate this

limitation and make R-ALOHA suitable for vehicular environments, in which hidden

terminals exist, RR-ALOHA is devised.


In R-ALOHA time is divided into timeslots, which are grouped into frames with length

N timeslots. If a node’s transmission in a timeslot is successful, that timeslot becomes

reserved for that node in the subsequent frames and is no longer available for other nodes

until it is released. Some mechanism is required to synchronize the information among

all nodes about the status of the transmissions in the last N timeslots, i.e. , being suc-

cessful and hence reserved, collision, or no transmission. Note that while other nodes can

distinguish between successful and collided transmissions, the transmitting nodes cannot

detect collision and hence need feedback, for example piggybacked acknowledgment.

In RR-ALOHA, each node with a reserved timeslot in the frame, transmits a Frame

Information (FI) in its timeslot which indicates the status of each of the N previous

timeslots as observed by that node. A timeslot is indicated in FI as either BUSY or

FREE. It is indicated as BUSY if an FI was successfully received in that timeslot. When

it is indicated as BUSY, the ID of the node transmitting FI in that timeslot is also

included. FI transmitted by a node, say node i, has two purposes. First, consider nodes

j and k, which are within the communication range of node i but are not within each

other’s communication range. If node j transmits in timeslot tj, node k will learn about

that timeslot through the FI transmitted by node i and will not transmit in timeslot tj.

Therefore, a collision at node i in timeslot tj is avoided. Second, since FI includes the

ID of transmitters in BUSY timeslots, it can also act as acknowledgments.

RR-ALOHA only works as intended in good channel conditions. In harsh channel

conditions, such as in vehicular environments, a timeslot may be announced as free while

the transmission may have failed due to noise or fading. In the next frame, the probability

of collision is increased in the timeslot because it is announced as free when it actually

belongs to a node. Furthermore, FIs add significant overhead. In a simulation performed

in [30], with frame length equal to 200 and maximum number of nodes in a two-hop area

equal to 100, the overhead introduced by FIs in each timeslot is 2400 bits. Note that

safety messages in vehicular communication are small (about 100 bytes). Therefore, the


overhead in percentage is very large. Additionally, further study is required to determine

the suitability of RR-ALOHA for mobile environments.

2.3 Repetition-based Broadcast Protocols

The fundamental idea behind repetition-based broadcast is repeating a message several

times in an interval shorter than or equal to its lifetime to ensure high probability of

reception. In repetition-based broadcast protocols, time is divided into frames, the max-

imum length of which must not be greater than the lifetime of a safety message. Each

frame, in turn, is divided into L timeslots with length equal to the transmission time of

a single packet. The division of time into frames and timeslots is shown in Fig. 2.6. Each

packet is transmitted a number of times inside the frame according to a transmission

pattern. In each timeslot, if a node is not transmitting, it switches to the receive mode.

Each pattern can be represented by a binary vector of length L in which a ‘1’ denotes

a transmission and a ‘0’ represents an idle timeslot, as illustrated in Fig. 2.7. Each of

these vectors is called a codeword and the set of all codewords is called a code.

FrameLifetime Timeslot

t

... ...

1 0 010100010 0100

Transmission Pattern

Binary representation

Transmission Idle/Receive Mode

1 2 3 L

Fig. 2.6: Division of time into frames and timeslots for repetition-based broadcast with

L = 15, each time unit is equal to the length of one timeslot.

In the following, we only consider synchronous protocols. In a synchronous protocol,

timeslots are synchronized. Synchronization can be achieved using a variety of methods.

One that is particularly appealing to vehicular communication applications is Global

Positioning System (GPS) [9] because many vehicles are already equipped with GPS

devices and more will be equipped in the future.


FrameLifetime Timeslot

t

... ...

1 0 010100010 0100

Transmission Pattern

Binary representation

Transmission Idle/Receive Mode

1 2 3 L

Fig. 2.7: A transmission pattern and its binary representation.

In this section, we study two random protocols proposed in [39], [40], and [19]. In

random repetition protocols, the transmission patterns are chosen randomly.

2.3.1 Synchronous p-Persistent Repetition (SPR)

In SPR, the source node transmits the packet in each timeslot in a frame with probability

p and remains idle with probability 1−p. Note that a packet may be transmitted L times

or not transmitted at all. In this approach, the total number of possible codewords is 2L.

2.3.2 Synchronous Fixed Repetition (SFR)

In SFR, each packet is transmitted w times in each frame, i.e., w timeslots are randomly

chosen out of the L available timeslots for repeated transmissions of the packet.

It is shown in [19], via simulation, that SFR decreases the probability of failure by

one order of magnitude, compared to IEEE 802.11a. The performance of SPR is worse

than SFR but better than IEEE 802.11a.

The above protocols are able to perform well in harsh channel conditions due to

multiple transmissions and are robust against mobility because they do not rely on the

knowledge of position of nodes in the network, and instead take advantage of the fact

that safety messages are intended for all nodes in the neighbourhood area. Furthermore,


although each message is repeated several times, the overhead is still less than many

of the protocols discussed in the previous sections because of the short length of safety

messages. However, these protocols are not able to combat collision. In SPR and SFR, the

timeslots in which a transmission takes place, i.e., the transmission patterns, are chosen

randomly. Randomly choosing transmission patterns results in relatively high probability

of collision. On the contrary, if transmission patterns are chosen deterministically with

the goal of decreasing collision, we are able to mitigate collision among users. In the next

chapter, we propose a repetition-based broadcast protocol in which transmission patterns

are deterministically chosen using optical orthogonal codes to reduce the possibility of

collision.

2.4 Other Protocols and Chapter Summary

In this chapter, we have discussed previously proposed broadcast protocols. We have not

included protocols that rely on clustering to provide broadcast support since the main

challenge in such protocols is the formation of stable clusters and electing best cluster

heads. Examples of cluster-based protocols can be found in [24] and [41]. Also, protocols

that rely on out of band signaling such as [20], [42], and [43], are not discussed here

because out of band signaling is not compatible with DSRC specification.

We discussed the suitability of the previously proposed protocols considered here for

vehicular communications. Unreliable broadcast service, large overhead and inefficiency

in delivering short safety messages, insufficient support for mobility, and neglecting other

causes of unreliability such as harsh channel conditions by focusing only on collision are

among the problems from which some of the discussed protocols suffer. In the next

chapter we aim at eliminating these issues by introducing OOC-based broadcast.

Chapter 3

OOC-based Broadcast Protocol

In the previous chapter, we reviewed the relevant previously proposed broadcast proto-

cols. In particular, we described two repetition-based broadcast protocols, namely SFR

and SPR, which use random codes as transmission patterns. In this chapter, to decrease

the possibility of collision among users, we introduce the repetition broadcast based

on optical orthogonal codes. We also describe the code assignment procedure, explain

frame-synchronous and frame-asynchronous modes of the protocol, and present the code

generation algorithm.

3.1 Broadcast using Optical Orthogonal Codes

In this section, we give a brief introduction to Optical Orthogonal Codes (OOC) and

describe their use as transmission patterns.

One way to choose the transmission patterns is to make sure that any two patterns

do not collide in more than one timeslot. This is the main idea behind choosing optical

orthogonal codes as transmission patterns.

Assuming vectors x and y are two codewords used as transmission patterns, their

cross-correlation, represented by the inner product 〈x,y〉, is the number of collisions in

an ideal channel which occur if two users transmit with patterns indicated by x and

31

Chapter 3. OOC-based Broadcast Protocol 32

y. Therefore, limiting the correlation of two codewords is equivalent to limiting the

possibility of collisions. A synchronous optical orthogonal code, C, with length L and

weight w is a code whose codewords are binary vectors with length L and the number of

‘1’s in each codeword is w. Furthermore, the codewords satisfy the following condition

〈x,y〉 =L∑i=1

xiyi ≤ λ ∀x,y ∈ C (3.1)

where λ is a fixed integer usually taken to be 1. Choosing OOC with λ = 1 as transmission

patterns guarantees that any two users have at most one common timeslot, while in

random methods, the number of collisions can be up to w for SFR and up to L for SPR.

A synchronous optical orthogonal code, C, with length L, weight w, and maximum

correlation λ is equivalent to a constant weight code with minimum Hamming distance

2δ = 2(w− λ) and same length and weight. The size of the largest constant-weight code

with given values for L, w, and 2δ is unknown in the general case [44]. Johnson [45]

provides an upper bound for the number of codewords in such code

‖C‖ ≤⌊L

w

⌊L− 1

w − 1· · ·

⌊L− w + δ

δ

⌋···⌋⌋

(3.2)

where bxc is the largest integer less than or equal to x. In this work, we only consider

λ = 1. For example, for L = 64 and w = 6, code cardinality is bounded by 128 and

for L = 128 and w = 9, code cardinality is bounded by 233. Lower bounds are usually

obtained by constructing a code with given parameters. Note that strict orthogonality,

i.e., λ = 0, leads to a very low code cardinality, namely, at most L/w.

An example code with L = 13, w = 4, and λ = 1 is shown below


C =

0 0 0 0 0 1 0 0 1 0 1 0 10 1 1 0 0 0 0 0 0 0 0 1 10 0 0 1 1 0 0 0 0 1 0 0 11 0 0 0 0 0 1 1 0 0 0 0 11 0 0 1 0 0 0 0 0 0 1 1 01 0 1 0 0 0 0 0 1 1 0 0 01 1 0 0 1 1 0 0 0 0 0 0 00 1 0 1 0 0 0 1 1 0 0 0 00 0 1 1 0 1 1 0 0 0 0 0 00 0 0 0 0 1 0 1 0 1 0 1 00 0 1 0 1 0 0 1 0 0 1 0 00 0 0 0 1 0 1 0 1 0 0 1 00 1 0 0 0 0 1 0 0 1 1 0 0

(3.3)

Typical values for frame length, L, packet length, Lp, and data rate, R, can be 100,

100B or 800bits, and 10Mbps, respectively. Therefore, a typical value for the duration of

a frame is Tf = LLp/R = 100× 800/10M ≈ 10ms.

3.1.1 Code Division Multiple Access, Positive Optical Systems,

and Optical Orthogonal Codes

Consider a synchronous Code Division Multiple Access (CDMA) system with N users,

in which a user’s message, xi(t), is multiplied by a unique code waveform, ci(t), to obtain

xi(t) = xi(t)ci(t) (3.4)

where xi(t) is an on-off (with values in {0,1}) binary waveform and the duration of ci(t)

is equal to the duration of a bit, denoted by Tb. Bit duration, Tb, is divided into L chips

with length Tc, in each of which ci(t) has a constant value. A waveform ci(t) can be

represented by a vector ci = (ci1, . . . , ciL),

ci(t) = cij for (j − 1)Tc ≤ t < jTc. (3.5)


The waveforms ci(t) and, equivalently, vectors ci are orthogonal, as shown in Fig. 3.1. (a).

Formally

〈ci(t), cj(t)〉 ,∫Tb

ci(t)cj(t)dt =L∑k=1

cikcjk = 0 for i 6= j (3.6)

and

〈ci(t), ci(t)〉 ,∫Tb

‖ci(t)‖2dt =L∑k=1

‖cik‖2 = Pc (3.7)

In this system, the signal received by each receiver is the sum of all transmitted

signals,∑Ni=1 xi(t). Because of the orthogonality of waveforms ci(t), the receiver can

obtain xi(t) from this summation by using correlation detectors.

The principles of the above system can be extended to a positive optical systems. In

a positive optical system [46] signals are non-negative, therefore, waveforms with non-

trivial strict orthogonality, such as c1 = (0, 0, 1, 1) and c2 = (1, 1, 0, 0), are limited in

number. However, we can relax the orthogonality condition of (3.6) to obtain

〈ci, cj〉 =L∑k=1

cikcjk ≤ λ for i 6= j (3.8)

where the elements of the vectors ci are ‘0’ or ‘1’ as illustrated Fig. 3.1. (b).

The above equation, in fact, is the defining equation of optical orthogonal codes and

leads to Optical Code Division Multiple Access (OCDMA) based on optical orthogonal

codes [47]. The received signal can be detected using correlation detectors or chip-level

detectors [48].

In OOC-based OCDMA, ‘1’s at the same position must be avoided as much as possible

because the correlation they produce cannot be canceled out since the waveforms are

positive. Similarly in packet networks, packets transmitted at the same time lead to

collision and must be avoided. Therefore, we use the same criteria as in OOC-based

OCDMA to determine transmission patterns with the goal of minimizing the possibility

of collision. This leads to optical orthogonal transmission patterns defined by (3.1). A

sample of OOC-based transmission patterns is shown in Fig. 3.1. (c) for λ = 1.


user 1

user 2

user 3

(a) (b) (c)

One bit One bit One frameChip Time

slot

Fig. 3.1: (a) A CDMA system in which signal are strictly orthogonal, (b) OOC-based

optical CDMA in which each bit is expanded using optical orthogonal codes, and (c)

OOC-based transmission patterns for a repetition-based broadcast networks

From a different point of view, instead of expanding each bit in time as in spread

spectrum communications, we have expanded each packet in time. This approach has

the benefit that the fine synchronization required for CDMA at bit level is not required.

Instead, we need a much coarser synchronization at timeslot level which is, assuming

each packet is 800 bits, about three orders of magnitude longer than the duration of a

bit.

3.2 Distributed Code Assignment

We assume the set of codewords is decomposed into two subsets. A subset of codewords

in the code is reserved only for network association, denoted by set Ca. Once a vehicle

enters a road, it randomly selects a tentative codeword from the subset reserved for

network association. In the network association phase, the vehicle that wants to join

the network, can start transmitting its data packet as usual. However, it must also

acquire a permanent codeword that is unique within its two-hop communication range.

To obtain information about codewords used in the two-hop neighbourhood the joining


node issues Code Information Requests (CIQ). Every node i receiving a CIQ transmits

Code Information Response (CIR) which contains the index of its codeword and its

ID, the codewords of the node’s one-hop neighbours, and the ID of those neighbours.

The codewords indicated in the CIR received from node i, denoted by Ci, are used by

other nodes and hence unusable by the joining node. After receiving several of these

packets, the node with the tentative codeword chooses a permanent codeword from the

set Cp = C\Ca\∪iCi. While network association is performed only once when the vehicle

enters the road, each node with a permanent codeword also periodically transmits a CIR

with frequency of once every few seconds. This enables the network to adapt to topology

changes. If a node with a permanent codeword discovers that its codeword is being used

by one of its two-hop neighbours, it releases that codeword and chooses another one from

Cp. The detailed flowchart of the network association phase is shown in Fig. 3.2 and is

explained in the next section.

3.2.1 Code Information Response Window

When a joining node issues a CIQ, if all neighbours transmit CIR packets in the next

frame, the additional load caused by several immediate CIR packets results in perfor-

mance degradation. Note that because of the use of optical orthogonal codes several

nodes can transmit in the same frame successfully. However, if all one-hop neighbours

of an entering vehicle transmit CIR packets in the subsequent frame, a large increase in

network load results that must be avoided. To resolve this issue, we introduce the Code

Information Response Window (CIRW). Each node that receives a CIQ, sets a counter

to a random number uniformly chosen between 1 and CIRW. At the end of each frame,

the counter is decreased by one. When the counter reaches zero, CIR is transmitted in

the next frame. The joining node determines its permanent codeword after CIRW frames

have passed, as illustrated in Fig. 3.2. CIRW can be either predetermined or determined

by the node that transmits the CIQ packet. If it is predetermined, it is chosen in a way


Enter road

Transmit CIQ

Cp = C \ Cal = CIRW

l > 0

l = l - 1

Wait until end of frame

CIR received?

For all CIRs received

Cp = Cp \ CIR

Randomly choose codeword from Cp

yes

yes

Fig. 3.2: Network association phase


not to cause a significant load increase assuming a typical (or worst-case) number of one-

hop neighbours. Alternatively, the joining node can monitor the channel for the duration

of several frames and obtain an estimation of the number of its one-hop neighbours and

include the desired CIRW value in the CIQ packet.

1

7

3 54

8

2

9

CIQ

CIQ

CIQ

6 10

CIR4CIR2

CIR8

CIR2

CIR2

CIR4

CIR4

CIR8 CIR8

Fig. 3.3: CIQ and CIR transmissions; node 3 joins the network and transmits CIQ,

nodes within one-hop transmission reply by CIR

Fig. 3.3 shows a sample code assignment process. Node 3 is the newcomer to the net-

work and transmits a CIQ packet. Nodes 2, 4, and 8, receiving the CIQ, select a random

number between 1 and CIRW and transmit their CIR after that number of frames has

passed. A possible scenario is shown in Fig. 3.4. Note that, although CIR2 and CIR8 are

transmitted in the same frame, because of the use of optical orthogonal codes, both pack-

ets will be successfully received with high probability. Through CIR2, CIR4, and CIR8,

node 3 discovers the codewords used by its two-hop neighbours, {c1, c2, c4, c5, c7, c8, c9},

and does not choose those codes.

1 2 3 4 5 6 7

CIQ CIR4CIR2CIR8

CIRW=4

Fig. 3.4: Timeline of CIQ and CIR transmissions of Fig. 3.3


It takes CIRW×Tf seconds for an entering car to switch to a permanent codeword.

Collisions occur with high probability if a second car arrives during this time and chooses

the same tentative codeword. Assume that CIRW=10, Tf = 10ms, and cars entering a

road form a Poisson process with rate λr = 1 cars per second. When a car enters the road,

the probability that another car arrives before CIRW×Tf seconds has passed since the

arrival of the first car is 1−eCIRW×Tf = 1−e−10×0.010 = 0.0952. The probability that both

cars choose the same codeword is 1/‖Ca‖. To make the total probability, 0.0952/‖Ca‖

sufficiently small, ‖Ca‖ must be large. If this is not possible due to the limited cardinality

of the code, more codewords can be added to the code with higher cross-correlation. Since

these codes are in use only for a short time, performance degradation caused by their

higher cross-correlation is minimal.

The number of permanent codewords required depends on the desired communication

range, Rc. The cardinality of Cp must be large enough to support all vehicles in length

4Rc of a road. For example, assuming Rc = 100m and adjacent cars in the same lane are

30m apart, in a four-lane road at least 4 × 4 × 100/30 ≈ 53 permanent codewords are

required.

3.3 Frame-Synchronous and Frame-Asynchronous

In OOC-based broadcast, each node has a pattern for transmission, which is a codeword

denoting the available timeslots in which the node is allowed to transmit. This pattern

has length L and is repeated in time. Therefore, it can be properly represented with a

disk, slices of which are timeslots, as well as linearly, as shown in Fig. 3.5. The starting

point of these patterns are synchronized among all nodes.

When a message arrives at the network interface of a node, there are two options.

The node can wait for the beginning of the next frame or can begin transmitting at

the next available timeslot. The former is called frame-synchronous, where transmis-


10%

10%10%

10%10%

10%

10%

0

t

000

Fig. 3.5: Circular and linear representation of a transmission pattern with L = 10 and

w = 3.

sion is synchronized with frames, and the latter is called frame-asynchronous, where

transmission start independently of frames. In either case the message is transmitted in

available timeslots as indicated by the codeword. Fig. 3.6 show both frame-synchronous

and frame-asynchronous transmission by a user.

000

Message Arrival

00

Message Arrival

0

(a)

(b)

Fig. 3.6: Frame-synchronous and frame-asynchronous transmissions: (a) frame-

synchronous transmission, transmission starts in the next frame after a message arrives,

and (b) frame-asynchronous, transmission starts at the next available timeslot after a

message arrives

The analysis and simulation presented here, both assume frame-synchronous mode of


transmission for simplicity. The delay of frame-asynchronous transmission is less than

that of frame-synchronous because the waiting until the beginning of the next frame is

eliminated.

3.4 OOC Code Generation

One can construct an OOC code by building a graph with(Lw

)vertices, each corre-

sponding to a codeword, connected if their Hamming distance is larger than or equal to

2δ = 2(w − λ), and finding the maximum clique [49]. However, in our case, finding a

large maximal clique results in a sufficiently large code and finding the maximum clique

is not necessary. Therefore, we use a greedy depth first search (DFS) algorithm [50] to

find a maximal clique (code). In this algorithm, in each step a ‘1’ is added to the current

codeword in such a way that it does not cause the Hamming distance between the current

codeword and previous codewords to decrease below the given number. When a codeword

with weight w is obtained, it is added to the code. The algorithm ends when it cannot

find a new codeword. The pseudocode of the algorithms used to generate optical orthog-

onal codes is presented in Algorithm 1 and Algorithm 2. Global variables are initialized

outside the procedures. Procedure FindCode(L,w) starts by randomly choosing the first

timeslot for the current codeword from a set of available positions and calls procedure

DFS until there is no possibility for DFS to find another codeword. Procedure DFS

searches for available timeslots to add to the current codeword and calls itself recursively

to complete the current codeword or become unable to find another available timeslot.

In Fig. 3.7 a histogram of the position of ‘1’s in a sample code is shown. Fig. 3.7

shows the number of codewords in which a given position (timeslot) is occupied by a ‘1’

(indicated in the legend as ‘Total’). It can be observed that ‘1’s are roughly distributed

in a uniform manner. Therefore, all codewords must, more or less, have equal probability

of success. Fig. 3.7 also shows the number codewords whose first occurrence of a ‘1’ is in


a given position (indicated in the legend as ‘First’).

Algorithm 1 Procedure FindCode(L,w) finds an OOC code

1: Code← {}

2: depth← 0

3: found← false

4: Codeword← {}

5: procedure FindCode(L,w)

6: bits = {1, . . . , L} . All bits are available

7: while bits 6= {} do

8: b← random choice from bits

9: Codeword← {b}

10: depth← 1

11: found← true

12: while found = true do

13: found← false

14: DFS

15: end while

16: remove b from bits

17: end while

18: end procedure


Algorithm 2 Procedure DFS recursively searches for more available timeslots

1: procedure DFS

2: tmp = Codeword

3: bits← {1, . . . , L} − Codeword

4: Exclude bits that cause low distance with previous codes

5: while bits 6= {} do

6: b← random choice from bits

7: Codeword← {b}

8: if depth = w − 1 then

9: found← true

10: Add Codeword to Code

11: else

12: depth← depth+ 1

13: dfs

14: if found = true then

15: return

16: end if

17: end if

18: end while

19: depth← depth− 1

20: Codeword← tmp

21: end procedure


0 10 20 30 40 50 60 70 80 900

2

4

6

8

10

12

14

Position in codeword

Num

ber

of o

nes

Total

First

Fig. 3.7: Distribution of ‘1’s in a sample code and the first occurrence of a ‘1’ in codewords

with (L,w) = (90, 7) and λ = 1

Chapter 4

Analytical Performance Study of

Broadcast Protocols in Ideal

Channel

To gain a better understanding of repetition-based broadcast protocols, in this chapter,

we analytically study the performance of such protocols. Although we only consider

SPR, SFR, and OOC, our analysis is general and can also be applied to other repetition-

based protocols. In this chapter, we assume a network that is interference limited and an

ideal wireless channel, which carries signals with no attenuation and no noise. Hence, we

neglect the effect of noise on the performance. As the traffic model, we use the binomial

distribution. However, other traffic models can also be used. Furthermore, we assume

frames and timeslots are synchronized.

4.1 Probability of Success

Transmission in a timeslot is successful if only one node transmits in that timeslot. If two

nodes transmit in the same timeslot, a collision occurs and both transmissions fail. Since

45

Chapter 4. Analytical Performance Study 46

the channel is assumed ideal, all nodes are able to receive a transmission when there is

no collision. In each frame, each node transmits in several timeslots. The message is

successfully transmitted if at least one of the transmissions in the frame is successful.

Probability of success is defined as the number of messages successfully transmitted by

a node divided by the number of messages that the node has attempted to transmit.

Probability of success depends on the number of interfering users, i.e., the number of

users transmitting in the same frame as the desired user.

In order to obtain the probability of success, we introduce the following notations.

Events are denoted by calligraphic letters such as R, and the probability of the event R

is denoted by P (R). The probability of an event R, given that there are M interfering

users, is shown by PM(R). The event R is the complement of R. Boldface capital letters,

such as X, denote random variables, and the expectation of the random variable X is

shown by E(X).

Furthermore, we have the following definitions. S is the event that at least one trans-

mission is successful in a frame. When discussing protocols with exactly w transmissions

in a frame, such as OOC and SFR, Si denotes the event that the ith transmission among

w transmissions is successful and Si denotes the event that the ith transmission is the

first successful transmission. When discussing SPR, Si is the event that the transmission

in the ith timeslot is successful and Si is the event that the first successful transmission

occurs in the ith timeslot.

Assume that the desired user is transmitting a message. The probability of success,

Ps, can be written as

Ps =N−1∑M=0

PM(S)P (M = M) (4.1)

where, M is the random variable denoting the number of interfering users and N is the

total number of users in the network.

The probability mass function of M depends on the traffic model. The traffic model

is discussed in Section 4.3. In this section, we focus on obtaining PM(S).


4.1.1 Probability of Success for SPR

For SPR, the desired transmitter is successful in the ith timeslot if it transmits in that

timeslot and all other users are silent. Assuming each user transmits with probability p

in each timeslot, the probability of success in a timeslot, s, is

s , PM(Si) = p (1− p)M 1 ≤ i ≤ L. (4.2)

The desired user fails to transmit its packet successfully if it fails in all L timeslots.

The probability of failure in all timeslots is (1−s)L. Therefore, the probability of success,

PM(S), is

P(SPR)M (S) = 1− (1− s)L = 1−

(1− p (1− p)M

)L. (4.3)

4.1.2 Probability of Success for SFR and OOC

In SFR and OOC, since there are exactly w repetitions, different timeslots are not in-

dependent. Therefore, the probability of success cannot be obtained as easily as that of

SPR. SFR and OOC are two special cases of repetition-based broadcast schemes with

exactly w transmissions.

In a repetition-based protocol with exactly w transmissions, the probability that at

least one transmission is successful among i transmissions, PM(S1 ∪ · · · ∪ Si), can be

written as

PM(S1 ∪ · · · ∪ Si) =i∑

k=1

PM(Sk)−∑

{a1,a2}∈(i)2

PM(Sa1 ∩ Sa2)

+∑

{a1,a2,a3}∈(i)3

PM(Sa1 ∩ Sa2 ∩ Si+1) + · · ·

=i∑

k=1

(−1)k+1∑

{a1,...,ak}∈(i)k

PM(Sa1 ∩ · · · ∩ Sak)

(4.4)

where (i)k is the set of all k-subsets of {1, · · · , i}. As we will see, PM(Sa1 ∩ · · · ∩ Sak)

does not depend on a1, . . . , ak but rather only on k. Therefore, by defining

γk = PM(Sak∩ Sak−1

∩ · · · ∩ Sa1), {a1, . . . , ak} ∈ (i)k (4.5)


we have

PM(S1 ∪ · · · ∪ Si) =i∑

k=1

(−1)k+1

(i

k

)γk (4.6)

The probability of success with M interfering users in a repetition-based protocol with

exactly w transmissions, PM(S) = PM(S1 ∪ · · · ∪ Sw), can be obtained, by substituting i

by w in (4.6), as

PM(S) =w∑k=1

(−1)k+1

(w

k

)γk . (4.7)

Next, we find γk = PM(Sak∩Sak−1

∩ · · · ∩ Sa1) for SFR and OOC and substitute it in

(4.7) to obtain success probability for SFR and OOC.

SFR

The probability that a certain interfering user, i.e., a user that transmits in the same

frame, does not transmit in timeslots ak, ak−1, . . . , a1 with the desired user is equal to(L−kw

)(Lw

) (4.8)

where the transmission pattern of the interfering user can be any of the(Lw

)patterns

with equal probability. Among the possible patterns,(L−kw

)pattern does not include

transmission in the prohibited timeslots ak, ak−1, . . . , a1.

Since the M interfering users are independent,


∩ · · · ∩ Sa1) =

(L−kw

)(Lw

)M (4.9)

Therefore, as claimed earlier, PM(Sak∩ Sak−1

∩ · · · ∩ Sa1) does not depend on a1, . . . , ak

but only on k. Hence,

P(SFR)M (S) =

w∑k=1

(−1)k+1

(w

k

)(L−kw

)(Lw

)M . (4.10)


OOC

In OOC with λ = 1, by definition, an interfering user may transmit in only one timeslot

in which the desired user also transmits. Consider the ajth transmission of the desired

user. Let p1 be the probability that a certain interfering user transmits in the same

timeslot as the ajth transmission of the desired user. The probability that the interfering

user does not transmit at the same time as any of the transmissions ak, ak−1, . . . , a1 of

the desired user is 1− kp1. Considering M independent interfering users, we have


∩ · · · ∩ Sa1) = (1− kp1)M . (4.11)

Note that PM(Sak∩ Sak−1

∩ · · · ∩ Sa1) only depends on k. Substitution in (4.7) yields

P(OOC)M (S) =

w∑k=1

(−1)k+1

(w

k

)(1− kp1)

M . (4.12)

4.1.3 Probability of Success and Interference Probabilities

In the previous section, we derived the probability of success for SFR and OOC in terms

of the probability that k certain transmissions are successful, i.e., γk. In this section, we

present a method for finding γk in terms of the probabilities that another user interferes

with the desired user in a given j-subset of w timeslots in which the desired user transmits,

called interference probabilities. The interference probabilities can be easily determined

for any generated code. We also find p1, introduced in the previous section, in terms of

L and w.

The desired user is susceptible to interference from other users in w timeslots in which

it transmits. For example, in the following code, shown in a matrix form where each row


corresponds to a user’s codeword,

1 0 1 0 1 0

0 0 1 1 1 0

1 1 1 0 0 0

0 1 1 0 0 1

1 0 1 0 0 1

(4.13)

assuming u1 (the first row) is the desired user and the other rows are the interfering users

in the current frame, the desired user is only susceptible in its three transmissions, i.e.,

the first, third, and fifth timeslots.

Let Sr be a subset of {1, 2, · · · , w}, where 1 ≤ r ≤ 2w. Sr is called an interference

pattern and its elements determine the transmissions, among w transmissions of the

desired user, with which an interfering user collides. For example, in (4.13), only u4

interferes with the desired user with interference pattern {2}. ΓSr is defined as the

number of users that interfere with the desired user with pattern Sr. Therefore, for

example, Γ{2} = 1. For other interference patterns we have Γ∅ = Γ{1} = Γ{3} = Γ{1,3} =

Γ{1,2,3} = 0, Γ{2,3} = 1, and, Γ{1,2} = 2. Note that∑2w

r=1 ΓSr is equal to the number of

interfering users.

The probability of interference with the specific pattern Sr is shown by pSr and called

an interference probability. SFR and SPR randomly choose the timeslots in which trans-

missions take place and OOC, while not strictly random, shows some random properties.

For example from Fig. 3.7, it can be observed that the total number of transmissions,

from all users, in a timeslot are roughly the same for all timeslots in a frame. Assuming

randomness in the code, pSr only depends on the cardinality of Sr, shown by ‖Sr‖. Hence

pSr = pSt , as long as ‖Sr‖ = ‖St‖. With small abuse of notation, we can write pj = pSr

where j = ‖Sr‖. Therefore, a set of 2w interference probabilities, pSr , 1 ≤ r ≤ 2w, is

reduced to a set of w + 1 interference probabilities, pj, 0 ≤ j ≤ w. It is desirable to find

γk for 1 ≤ k ≤ w in terms of pj.


The number of interfering users in the timeslot in which the ith (out of w) transmission

takes place is denoted by Zi. We have

Zi =2w∑r=1

ΓSrI ({i} ∩ Sr 6= ∅) (4.14)

where I(.) is the identifier function which is 1 if the condition inside the brackets is true,

and 0 otherwise. For example,

Z1 = Γ{1} + Γ{1,2} + Γ{1,3} + · · ·

+ Γ{1,w} + Γ{1,2,3} + · · ·+ Γ{1,··· ,w}

(4.15)

In code (4.13),

Z1 = Γ{1} + Γ{1,2} + Γ{1,3} + Γ{1,2,3} = 2,

Z2 = Γ{2} + Γ{1,2} + Γ{2,3} + Γ{1,2,3} = 4,

Z3 = Γ{3} + Γ{1,3} + Γ{2,3} + Γ{1,2,3} = 1.

(4.16)

Note that γk can be written in terms of Zi as

γk = PM(Z1 = 0, · · · ,Zk = 0). (4.17)

To find the right-hand-side of (4.17), we first find PM(Z1 = 0). Using (4.14), we can

write

PM(Z1 = 0) = PM

(2w∑r=1

ΓSrI ({1} ⊆ Sr) = 0

)

= PM(Γ{1} = 0,Γ{1,2} = 0,Γ{1,3} = 0, · · · ,Γ{1,w} = 0,Γ{1,2,3} = 0, · · · ,Γ{1,··· ,w} = 0

)=

(1−

[p1

(w − 1

0

)+ p2

(w − 1

1

)+ p3

(w − 1

2

)+ · · ·+ pw

(w − 1

w − 1

)])M(4.18)

To explain (4.18), we reorder the timeslots in a way that all the ‘1’s in the codeword of

the desired user are in timeslots 1 to w and the codeword becomesw←−−−−−→

1111 · · · 11L−w←−−−−→

000 · · · 0.

The codeword of the desired user is shown in the topmost row of Table 4.1. Momentarily,

only consider one interfering user. Only the ‘1’s in the codeword of the interfering user


that are (after reordering) in timeslots 1 to w cause collision with the desired user. The

interfering user has one of the codewords shown in the center column of Table 4.1. The

possible codewords for the interfering user can be divided into two classes, separated in

Table 4.1, the codewords that collide with the desired user in the first timeslot and the

codewords that do not. These two classes are, in turn, divided into groups based on the

number of timeslots in which they collide with the desired user. The codewords in each

group, which collide in the same number of timeslots with the desired user, are divided

into subgroups based on their specific interference patterns with the desired user. For

example, the codewords in the subgroup 1100 · · · 00xxx . . . x only collide with the desired

user in the first and second timeslots, i.e., Sr = {1, 2}. The probability that the codeword

of the interfering user is in a specific subgroup is shown in front of that subgroup, in the

right column. The number of subgroups in a group, which have the same number of

colliding timeslots with the desired user, is indicated in the left column. The probability

the codeword of the interfering user is in a certain group is the number of subgroups in

that group times the probability of each subgroup. For example, the probability that the

codeword of the interfering user collides with the interfering user in the first timeslot and

one other timeslot is p2

(w−1

1

). We have

P1(Z1 = 0) = 1− P1(Z1 6= 0).

P1(Z1 6= 0), when there is only one interfering user, can be obtained by summing the

probabilities of the groups in the first class (groups that collide with the desired user in

the first timeslot) and is equal to

p1

(w − 1

0

)+ p2

(w − 1

1

)+ p3

(w − 1

2

)+ · · ·+ pw

(w − 1

w − 1

).

Since different interfering users have independent interference patterns we can write (4.18)

by noting that PM(Z1 = 0) = (P1(Z1 = 0))M = (1− P1(Z1 6= 0))M .


w←−−−−−→1111 · · · 11

L−w←−−−−→000 · · · 0(

w−1

0

)1000 · · · 00 xxx . . . x p1

(w−1

1

)1100 · · · 00 xxx . . . x p2

1010 · · · 00 xxx . . . x p2

· · ·

1000 · · · 01 xxx . . . x p2

(w−1

2

)1110 · · · 00 xxx . . . x p3

1101 · · · 00 xxx . . . x p3

· · ·

1000 · · · 11 xxx . . . x p3

· · · · · ·(w−1

w−1

)1111 · · · 11 000 . . . 0 pw(

w−10

)0000 · · · 00 xxx . . . x p0

(w−1

1

)0100 · · · 00 xxx . . . x p1

0010 · · · 00 xxx . . . x p1

· · ·

0000 · · · 01 xxx . . . x p1

(w−1

2

)0110 · · · 00 xxx . . . x p2

0101 · · · 00 xxx . . . x p2

· · ·

0000 · · · 11 xxx . . . x p2

· · ·(w−1w−1

)0111 · · · 11 xxx . . . x pw−1

Table 4.1: Possible codewords and their probabilities for one interfering user


We generalize (4.18) to write

γk = PM(Z1 = 0, · · · ,Zk = 0) = PM

(2w∑r=1

ΓSrI ({1, · · · , k} ∩ Sr 6= ∅) = 0

)

= PM

(∧{1,··· ,k}∩Sr 6=∅ ΓSr = 0

)

=

1−w∑j=1

pj

min(j,k)∑i=1

(k

i

)(w − kj − i

)M

=

1−w∑j=1

pj

[(w

j

)−(w − kj

)]M(4.19)

where ∧ is the logical AND operator. Finally, by substituting (4.19) in (4.7), we have

PM(S) =w∑k=1

(−1)k+1

(w

k

)1−w∑j=1

pj

[(w

j

)−(w − kj

)]M (4.20)

The above formula is general and is valid for broadcast communications using packet

repetitions provided 1) in each frame, there are w transmissions, and 2) interference

patterns with same cardinality are equally likely to happen. In the following sections, we

find the interference probabilities pj for SFR and OOC.

Interference Probabilities for SFR

First, we examine SFR to verify that it satisfies the conditions of Eq. (4.20). In SFR,

the desired transmitter, and all other transmitters, transmit in w timeslots. Therefore,

condition 1 is satisfied.

Assume Sr is an interference pattern. Since all codewords are equiprobable, the

probability that an interfering user collides with the desired user with pattern Sr is equal

to the number of codewords with interference pattern Sr divided by the total number of

codewords,

pSr =

(L−w

w−‖Sr‖

)(Lw

) . (4.21)

Since pSr only depends on the cardinality of Sr condition 2 of Eq. (4.20) is also satisfied.

Therefore, we can write

pj =

(L−ww−j

)(Lw

) (4.22)


and

P(SFR)M (S) =

w∑k=1

(−1)k+1

(w

k

)1−w∑j=1

(L−jw−j

)(Lw

) [(w

j

)−(w − kj

)]M (4.23)

Interference probabilities for OOC

Having w transmissions in each frame, OOC satisfies condition 1 of Eq. (4.20). The only

possible interference patterns for OOC are

Sr = {r} 1 ≤ r ≤ w (4.24)

Since OOC is not a random code, pSr may depends on r. However, random properties

can still be seen in optical orthogonal codes. For example Fig. 3.7 shows that distribution

of ones in a sample code is roughly uniform. Therefore, we assume that condition 2 of

Eq. (4.20) also holds and

p1 = pSr = p{r} 1 ≤ r ≤ w (4.25)

By substituting (4.25) in (4.20) we have

P(OOC)M (S) =

w∑k=1

(−1)k+1

(w

k

)(1− kp1)

M(4.26)

p1 is equal to the number of codewords in the subgroup in the second row in Table 4.1,w←−−−−−→

1000 · · · 00L−w←−−−→xxx · · · x, divided by the total number of possible codewords. Codewords in

this subgroup are common in the first timeslot. Therefore, they cannot have any other

common timeslot among the L − w timeslots denoted by x. Since w − 1 ones must be

placed in the L − w timeslots denoted by x with no overlap, there are at most L−ww−1

codewords in this subgroup. From (3.2), the total number of codewords is at most⌊Lw

⌊L−1w−1

⌋⌋. Therefore, p1 may be approximated by

p1 ≈w(L− w)

L(L− 1). (4.27)


p1 L=64 L=128

w=2 3.075377e-02 1.580402e-02

w=3 4.504188e-02 2.247906e-02

w=4 5.977387e-02 3.188442e-02

w=5 7.471811e-02 3.780905e-02

w=6 8.622891e-02 4.499162e-02

w=7 9.913148e-02 5.328069e-02

w=8 1.096639e-01 5.966638e-02

w=9 1.104613e-01 6.668362e-02

Table 4.2: Sample values of p1 obtained by generating OOC codes for various L and w

p1 can also be empirically obtained from a sample generated code using

p1 ≈

‖C‖∑i=1

‖C‖∑j=i+1

〈ci, cj〉

w(‖C‖2

) (4.28)

where C is the generated code and vectors ci are codewords. The division by w is because

p1 corresponds only to one transmission among w transmissions. Sample values of p1 are

presented in Table 4.2. Comparison of approximate values and sample values of p1 is

presented in Fig. 4.1. It is observed that approximation given in (4.27) provide values

close to empirical results, especially when w is not too large.

4.2 Average Delay

When a packet is transmitted several times in a frame, the delay, D, is defined as the first

timeslot in which the packet is successfully received. The average delay, Ds, is defined as

Ds , E(D|S) =N−1∑M=0

Ds(M)P (M = M) (4.29)


2 3 4 5 6 7 8 9 10 11 120

0.02

0.04

0.06

0.08

0.1

0.12

0.14

w

p 1

L=64, sample codeL=64, approximationL=128, sample codeL=128, approximation

Fig. 4.1: Approximate and sample values of p1 for OOC

where Ds(M) is the average delay of a successful transmission when there are M interfer-

ing users. Note that D is not defined when all transmissions in a frame are unsuccessful.

4.2.1 SPR

The average delay for SPR, conditioned on successful transmission when there are M

interfering users, can be obtained as

Ds(M) , EM [D|S] =L∑i=1

iPM(Si|S)

=L∑i=1

iPM(Si ∩ S)

PM(S)=

∑Li=1 iPM(Si)PM(S)

·(4.30)

Timeslot i is the first successful timeslot with probability

PM(Si) = PM(Si ∩ Si−1 ∩ · · · ∩ S1)

= PM(Si)PM(Si−1) · · ·PM(S1)

= s(1− s)i−1

(4.31)


where we have used (4.2). By substituting (4.31) in (4.30) we obtain

D(SPR)s (M) =

1

s− L(1− s)L

1− (1− s)L. (4.32)

Note that the right-hand-side of (4.32) implicitly depends on M through s.

4.2.2 SFR and OOC

In a repetition-based protocol with exactly w transmissions we have

Ds(M) , EM [D|S]

=w∑i=1

PM(Si|S)EM [D|Si]

=w∑i=1

PM(Si)PM(S)

L−(w−i)∑j=i

jPM(D = j|Si)

(4.33)

where PM(D = j|Si) is the probability that the delay is equal to j when the ith transmis-

sion is the first successful transmission. In other words, this is the probability that the

ith transmission takes place in the jth timeslot given that the ith transmission is the first

successful transmission. To calculate (4.33) we need to find PM(Si) and PM(D = j|Si).

To obtain PM(Si), we first find the probability that “transmission i is not successful


but at least one previous transmission is successful”, PM(Si ∩ (Si−1 ∪ · · · ∪S1)). We have

PM(Si ∩ (Si−1 ∪ · · · ∪ S1)) = PM(i−1⋃l=1

(Si ∩ Sl))

=∑{a1}∈(i−1)1

PM(Si ∩ Sa1)

−∑

{a1,a2}∈(i−1)2

PM(Si ∩ Sa1 ∩ Sa2)

+ · · ·+ (−1)k+1∑

{a1,··· ,ak}∈(i−1)k

PM(Si ∩ Sa1 ∩ · · · ∩ Sak) + · · ·

+ (−1)i∑

{a1,··· ,ai−1}∈(i−1)i−1

PM(Si ∩ Sa1 ∩ · · · ∩ Sai−1)

=i−1∑k=1

(−1)k+1∑

{a1,...,ak}∈(i−1)k

PM(Si ∩ Sa1 ∩ · · · ∩ Sak)

(4.34)

where (i− 1)k is the set of all k-subsets of {1, · · · , i− 1}. Let

γk , PM(Sak∩ Sak−1

∩ · · · ∩ Sa1)

ηk , PM(Sak∩ Sak−1

∩ · · · ∩ Sa1) .

(4.35)

Therefore,

PM(Si ∩ (Si−1 ∪ · · · ∪ S1)) =i−1∑k=1

(−1)k+1

(i− 1

k

)ηk+1

=i−1∑k=1

(−1)k(i− 1

k

)(γk+1 − γk)

(4.36)

As discussed earlier, because the probability of an interference pattern only depends on

the number of timeslots in which two users transmit simultaneously and not the position

of those timeslots, γk and ηk only depend on k.


Using (4.36) we can write

PM(Si) = PM(Si ∩ Si−1 ∩ · · · ∩ S1)

= 1− PM(Si ∪ Si−1 ∪ · · · ∪ S1)

= 1− PM(Si)− PM(Si−1 ∪ · · · ∪ S1) + PM(Si ∩ (Si−1 ∪ · · · ∪ S1))

= γ1 −i−1∑k=1

(−1)k+1

(i− 1

k

)γk +

i−1∑k=1

(−1)k(i− 1

k

)(γk+1 − γk)

=i−1∑k=0

(−1)k(i− 1

k

)γk+1 =

i∑k=1

(−1)k+1

(i− 1

k − 1

)γk

(4.37)

where in the fourth step we have used 4.6.

Note that (4.37) can be used to obtain PM(S) since PM(S) =∑wi=1 PM(Si). As seen

below, the result is the same as in (4.7).

PM(S) =w∑i=1

PM(Si)

=w∑i=1

i∑k=1

(−1)k+1

(i− 1

k − 1

)γk

=w∑k=1

(−1)k+1γkw∑i=k

(i− 1

k − 1

)

=w∑k=1

(−1)k+1

(w

k

)γk

(4.38)

Next, we obtain PM(D = j|Si) for SFR and OOC. Let T i be the timeslot in which

the ith transmission takes place.

PM(D = j|Si) = PM(T i = j|Si) = PM(T i = j) (4.39)

The last equality holds because the position of the ith transmission is independent of it

being the first successful transmission.

For SFR, since the position of transmissions in the frame is strictly random, we have

PM(D = j|Si) = PM(T i = j)

=

(j−1i−1

)(L−jw−i

)(Lw

) .(4.40)


For OOC, PM(T i = j) may depend on the code. Assuming ‘1’s are distributed evenly

in each codeword, we can use an expression identical to that of SFR as an approximation.

Finally, substituting (4.37) and (4.40) in (4.33), for OOC and SFR, Ds(M) is obtained

as

D(SFR)s (M) =

1

P(SFR)M (S)

w∑i=1

i∑k=1

(−1)k+1

(i− 1

k − 1

)(L−kw

)(Lw

)M L−(w−i)∑

j=i

j

(j−1i−1

)(L−jw−i

)(Lw

) (4.41)

and

D(OOC)s (M) =

1

P(OOC)M (S)

w∑i=1

i∑k=1

(−1)k+1

(i− 1

k − 1

)(1− kp1)

ML−(w−i)∑j=i

j

(j−1i−1

)(L−jw−i

)(Lw

) (4.42)

which are substituted in (4.29) to obtain Ds. On average, messages wait L/2 timeslots in

a buffer from their arrival at the network interface until the beginning of the next frame.

If delay is defined from the moment that a packet arrives at the network interface, L/2

must be added to the values obtained above.

4.3 Numerical Results

In this section, we present numerical results for Sections 4.1 and 4.2. Let L and N be

128 and 31 respectively. The value of p1 for OOC is calculated from (4.27). We assume

each vehicle independently makes a local decision, whether or not to transmit its location

to neighbour vehicles. Furthermore, we assume these periodical updates are generated

according to a Bernoulli model in each frame with probability µp. Since the decisions

for data transmission are independent, the number of nodes with an active packet in

each frame is a Binomial random variable with parameters N and µp, where N is the

total number of cars in a (loosely defined) cluster. (4.43) shows the pdf of the number

of interfering users.

P (M = M) =

(N − 1

M

)µp

M (1− µp)N−1−M (4.43)

where µp is the average load in packets per user per frame.


Fig. 4.2 shows the probability of failure, Pf , defined as 1 − Ps versus the average

number of transmissions, w, for SPR while load ranges from 0.1 to 1 (packets/user/frame)

with step 0.1 calculated with MATLAB using (4.1). Note that for SPR, w is the average

number of transmissions; the probability of transmission in a timeslot is p = w/L.

From Fig. 4.2, for each value for load, there is an optimum probability of transmission.

This probability can also be approximated by minimizing the probability of success for

the average number of interfering users. In (4.3), if M is substituted with the average

number of interfering users, (N − 1)µp, we have

P(N−1)µp(S) = 1−(1− p (1− p)(N−1)µp

)L. (4.44)

By taking the derivative with respect to p and equating it to zero

d

dpP(N−1)µp(S)

= −L(1− p (1− p)(N−1)µp

)L−1 (− (1− p)(N−1)µp + p(N − 1)µp (1− p)(N−1)µp−1

)= 0

(4.45)

We have

(1− p)(N−1)µp = p(N − 1)µp (1− p)(N−1)µp−1

⇒ (1− p) = p(N − 1)µp

⇒ popt =1

(N − 1)µp + 1

(4.46)

Fig. 4.3 shows the approximate value for the optimum number of average transmis-

sions and the result from numerically finding the minimum Pf . From Fig. 4.3, it is ob-

served that the approximation has very good accuracy for µp ≥ 0.5. The approximation

is not accurate for small µp. Note that in the derivation of (4.46), the main approxi-

mation is the substitution of the random variable M with a binomial distribution with

its average value. Therefore, the accuracy of this approximation can be explained by


5 10 15 20 25 30 35 4010

−4

10−3

10−2

10−1

100

w

Pf

μp increasing

Fig. 4.2: Probability of failure of SPR for N = 31, L = 128, w ranging from 1 to 40 and

µp ranging from 0.1 to 1 (packets/user/frame).

examining the ratio of the standard deviation to the mean,√

Var[M ]/E[M ]. When this

ratio is low, a distribution can be better approximated by its mean. We have

√Var[M ]

E[M ]=

√(N − 1)µp(1− µp)

(N − 1)µp

=

√1− µp

(N − 1)µp

(4.47)

which is a decreasing function of µp. Therefore, the approximation provides better results

for higher values of µp.

Fig. 4.4, and 4.5 show the probability of failure, Pf , for SFR, and OOC, respectively.

For SFR, w ranges from 1 to 40. For OOC, w ranges from 2 to 12. OOC for w = 1 is

a trivial case and the code cardinality for w > 12 is not big enough to accommodate 31

users.

From Fig. 4.4, it can be seen that for SFR, there exists a value for w which results

in the best performance. For OOC, as shown in Fig. 4.5, for high load, there exists an

optimum value for w, but in low load the highest possible w gives the best result. This


0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

5

10

15

20

25

30

35

40

45

μp (packets/user/frame)

w

SPR

SPR approx.

Fig. 4.3: Optimum average number of transmissions, w, and its approximation, for SPR,

versus load, for N = 31, L = 128.

5 10 15 20 25 30 35 4010

−5

10−4

10−3

10−2

10−1

100

w

Pf

μp increasing

Fig. 4.4: Probability of failure of SFR for N = 31, L = 128, w ranging from 1 to 40 and

µp ranging from 0.1 to 1 (packets/user/frame).

is illustrated in Fig. 4.6.

From Fig. 4.2, 4.4, and 4.5, it is observed that the optimum value of Pf of SFR is less


2 3 4 5 6 7 8 9 10 11 1210

−9

10−8

10−7

10−6

10−5

10−4

10−3

10−2

10−1

100

w

Pf

µp increasing

Fig. 4.5: Probability of failure of OOC for N = 31, L = 128, w ranging from 2 to 12

and µp ranging from 0.1 to 1 (packets/user/frame).

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 12

3

4

5

6

7

8

9

10

11

µp (packets/user/frame)

w

SFROOC

Fig. 4.6: Optimum number of transmissions, w, for SPR and OOC, versus load, for

N = 31, L = 128.

than that of SPR, and the optimum value of Pf of OOC is less than that of SFR. This

is illustrated in Fig. 4.7. Furthermore, a more in-depth analysis of the numerical results


shown in Fig. 4.2, 4.4, and 4.5, indicates that the optimum value of SFR can be obtained

by using OOC with fewer transmissions. Similarly, the optimum value of SPR can be

obtained by using SFR with fewer transmissions. A small number of transmissions has

the benefit of producing less interference for non-safety data if it is transmitted in the

same channel. From the numerical results we conclude that the performance of OOC is

better than those of SFR and SPR.

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 110

−9

10−8

10−7

10−6

10−5

10−4

10−3

10−2

10−1

100

µp (packets/user/frame)

Pf

SPRSFROOC

Fig. 4.7: Optimum probability of failure for OOC, SFR, and, SPR, versus load, for

N = 31, L = 128.

Fig. 4.8 shows the average delay of successful transmissions calculated using (4.29).

It is observed that the delay is more or less the same for different protocols. This fact

will also be observed in simulation results.


2 3 4 5 6 7 8 9 10 11 1210

15

20

25

30

35

40

45

50

55

60

w

Del

ay (

times

lots

)

OOC SFR SPR

µp increasing

Fig. 4.8: Delay of successful transmissions for OOC, SFR, and, SPR, for N = 31,

L = 128, w ranging from 2 to 12 and µp = 0.1, 0.4, 0.7, and 1 (packets/user/frame).

Chapter 5

Adaptive Elimination, Coding, and

Quality of Service Provisioning

In this chapter, we describe methods for improving the performance of the broadcast

protocol based on optical orthogonal codes. Adaptive elimination is proposed to further

decrease the probability of collision in Section 5.1. Coding, as a method of increasing

throughput and reliability, is presented in Section 5.2 and Quality of service is discussed

in Section 5.3.

5.1 Adaptive Elimination

Fig. 5.1 shows the transmission pattern of two users with one timeslot in common. It is

seen that both user 1, u1, and user 2, u2, are scheduled to transmit in the third timeslot.

For receiving nodes that are within the range of both users, a collision is likely to happen,

which results in both packets being lost. However, since user 2 receives a packet from

user 1 in the first timeslot, if user 1 includes some information in the transmitted packets

with which user 2 can identify the codeword used by user 1, user 2 can prevent the

collision simply by not transmitting in the third timeslot. We call this method Adaptive

Elimination.

68

Chapter 5. Adaptive Elimination, Coding, and QoS Provisioning 69

2 3 4 5 6 7

2 3 4 5 6 7

u1

u2

Fig. 5.1: Both u1 and u2 are scheduled to transmit in the third timeslot

Adaptive elimination increases reliability by eliminating transmissions in timeslots

that may potentially result in collision. To enable this, the codebook must be stored

in all nodes and nodes must transmit the index of their codeword and indicate which

timeslots are disabled. Each node adds a codeword indicator field with format (index of

codeword, enabled/disabled timeslots) to the header of its data packets to inform other

nodes of its codeword. Each part of this field has a predetermined length.

In Fig. 5.1 in the same frame, u1 and u2 transmit a packet with codewords c1 =

(1011000) and c2 = (0110010) respectively. In the first timeslot, u1 transmits its data

along with the sequence ‘001111’ where ‘001’ indicates that this user uses the fist code-

word and ‘111’ denotes that all transmissions in this codeword are used. Note that an

indexed codebook is stored on the nodes, which they can use to find the first codeword.

After receiving this packet, u2 disables its second transmission to eliminate the collision

in the third timeslot. Hence, its new codeword becomes c2 = (0100010). In the second

timeslot, u2 transmits ‘010101’ along with its data; ‘010’ denotes that u2 is using the sec-

ond codeword and ‘101’ denotes that the first and third transmissions of c2 are enabled

and the second transmission is disabled. When u1 receives the packet from u2, it realizes

that u2 has already disabled the transmission in the third timeslot. Therefore, u1 does

not need to disable its transmission in the third timeslot. The resulting transmission

patterns are shown in Fig. 5.2.

Since the number of OOC codes is limited (see (3.2)), the length of the index part

of the codeword indicator field is 2-3 bytes. Also, to announce which timeslots have


2 3 4 5 6 7

2 3 4 5 6 7

u1

u2

Fig. 5.2: u2 disables its transmission in the third timeslot

been eliminated from its codeword, so that other users do not eliminate transmissions in

the same timeslots, each user only needs to transmits w bits. Thus, this method does

not add significant overhead. For example, for L = 64 and w = 6, there are at most

b64/6b63/5cc = 128 codewords. Therefore, log 128+w = 13 bits are required for enabling

adaptive elimination.

Since a local map between the node ID and the codeword, acquired via CIR packets,

is available, an explicit codeword ID in the code indicator may not be required. However,

including codeword IDs in the code indicator field is more reliable.

5.2 Coding

If a packet with length Lp is transmitted w times in a frame, one successful reception

among w transmissions is sufficient. If the probability of successful reception is high, one

can use coding techniques to transmit more than one packet in a frame. Messages, the

units of data arriving at the network interface of a node to be transmitted, with length

Lm = rLp can be divided into r packets, each with length Lp, which can be coded in

the w transmitted packets. Reception of any r, out of w, packets guarantees successful

decoding of the message.

One can generate a matrixW composed of w packets for transmission, W = [w1 · · ·ww],

from data packets R = [r1 · · · rr] using W = R × A, where A is a matrix of dimension

r × w whose elements are chosen from a finite field with a proper size, Fq, and the mul-


tiplication operation is defined in Fq. Any r columns of A must be linearly independent.

Matrices with this property can be found in the coding literature [51]. If matrix A is

deterministic and made available to nodes in advance, coding can be performed with no

overhead.

Assume that, r packets, Q = [wa1 · · ·war ], are received by the receiver. Matrix B

is formed by choosing only columns a1, · · · , ar from A. Therefore, B is a square and

invertible matrix. The data, R, can be found at the receiver using WB−1. Note that

the index ai of a packet can be found by considering the transmission pattern of the

transmitter and the timeslot in which the packet is received or, alternatively, be explicitly

transmitted in the packet (with negligible overhead).

On the one hand, coding can increase the throughput of the network as presented in

Chapter 6. On the other hand large r can result in low probability of success. When

r = w, for example, loss of only one packet results in the loss of the message. The

throughput versus reliability trade-off provides flexibility in dealing with different types

of traffic. For traffic requiring the highest levels of reliability, repeating a packet w times

is the most suitable. For non-critical applications, whereby it is sufficient to deliver the

message to a subset of nodes in the network, r > 1 can be chosen to increase throughput.

5.2.1 Probability of Success with Coding

In this section we obtain the probability of success when coding is used to transmit a

message larger than one packet. When a message is divided into r packets, at least r

successful transmissions are needed for successful reception of the message. This event

is denoted by Sr.

SPR

For SPR, from (4.2), the probability of success in a timeslot, s, is p (1− p)M . Therefore,

since different timeslots are independent, the number of successful timeslots is a binomial


random variable and the probability of at least r successful transmissions is given by

PM(Sr) =L∑k=r

(L

k

)sk(1− s)L−k

= 1−r−1∑k=0

(L

k

)sk(1− s)L−k.

(5.1)

For example, if r = 2 we have

PM(S2) = PM(S)− Ls(1− s)L−1 (5.2)

Fig. 5.3 shows the probability of success for different values of r versus average load

for L = 64, N = 31, and w = 6. The traffic load arrives at nodes in messages with certain

length Lm. The messages are divided into r packets with length Lp = Lm/r. Since each

timeslot is long enough to transmit one packet, the duration of timeslots of schemes with

different r are different and, therefore, the duration of frames are different. Time unit

for measuring the average load is the duration of the frame when r = 1, denoted by

T1. Note that, on the horizontal axis, the average load for all schemes is the same, but

the arrival pattern for different r’s is different. The reason is that the traffic model is

binomial and a message arrives at each node with a given probability in each frame. For

example average load equal to 0.5 for r = 1 means a message arrives with probability 0.5

every T1 seconds while for r = 2 it means that a message arrives with probability 0.25

every T1/2 seconds.

It is observed from Fig. 5.3 that, for a wide range of load, r = 2 provides better

performance compared to r = 1. For low load, however, r = 1 performs better. When a

higher r is chosen, per frame load is decreased and, therefore, the number of collisions de-

creases. On the other hand, higher values of r need more successful packet transmissions

to successfully transmit a message. The optimum value of r depends both on average

load and the average number of transmissions.


0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Average load (messages/user/T1)

Ps

r=1r=2r=3r=4r=5r=6

Fig. 5.3: Probability of success of SPR for different r’s, L = 64, N = 31, w = 6

SFR and OOC

Since SFR and OOC have exactly w transmissions, the interference in different timeslots

are dependent. Thus, the derivation of the probability of success is more complicated

than that of SPR. We present the final result here and leave the proof to Appendix 5.4.

Proposition 5.2.1. Probability of a successful message delivery when the message is

divided into r packets with M interfering users in a repetition-based protocol with exactly

w transmissions, such as SFR- or OOC-based protocols, is

PM(Sr) = PM(S)−r−1∑l=1

l−1∑k=0

(w

l

)(l − 1

k

)(−1)kPM(Sw−l+k+1) (5.3)

where Si is the event that the ith transmission in a frame is the first successful transmis-

sion in that frame.

Fig. 5.4 and Fig. 5.5 show the probability of success for different values of r versus

average load for L = 64, N = 31, and w = 6. For both of these protocols, r = 2 provides

the best performance.


0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1


Ps

r=1r=2r=3r=4r=5r=6

Fig. 5.4: Probability of success of SFR for different r’s, L = 64, N = 31, w = 6

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1


Ps

r=1r=2r=3r=4r=5r=6

Fig. 5.5: Probability of success of OOC for different r’s, L = 64, N = 31, w = 6

5.3 Quality of Service

One of the desired characteristics of a vehicular network is the ability to provide better

service to high priority messages. In repetition-based broadcast using optical orthogonal


codes, number of transmissions can be varied to provide different levels of Quality of

Service (QoS). Adding more transmission to a frame, or equivalently, increasing the

weight of the corresponding codeword must be done in such a way that does not violate

the cross-correlation bound of (3.1).

Fig. 5.6 shows two users (or applications); one high priority and the other one low

priority. The frame length, L, is 16. The high priority user transmits whigh = 5 times and

the low priority user transmits wlow = 3 times. High priority users can take advantage

of more transmission opportunities in two ways: 1) transmitting with the same r as low

priority users to have increased reliability, and 2) transmitting with higher r to increase

their data rate. This gives the high priority users (or applications) the flexibility to

choose the best mode of operation.

It should be noted that the QoS of the high priority users deteriorates with high total

load. Different classes of service assigned here do not guarantee successful transmission or

high data rate for high priority messages. However, high priority messages are provided

with better service compared to low priority messages.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

High Priority

Low Priority

Fig. 5.6: High and low priority messages with different number of transmissions

5.4 Appendix: Proof of Proposition 5.2.1

Assume A and B are non-overlapping subsets of the w transmissions that take place

in a frame. Let SA,B be the event that, the transmissions in A are successful and the

transmissions in B are not successful. Since the probability of a specific interference


pattern only depends on the number of timeslots in which interference takes place, similar

to the analyses in Chapter 4, the probability of SA,B only depends on the cardinality of

A and B and not their specific elements. The probability of SA,B given that there are M

interfering users is denoted by ζM(‖A‖, ‖B‖).

Lemma 5.4.1. The probability ζM(a, b) can be found from the summation

ζM(a, b) =a−1∑k=0

(a− 1

k

)(−1)kPM(Sb+k+1) (5.4)

where Si is the event that the ith transmission in a frame is the first successful transmis-

sion in that frame.

Proof. Assume A, B, and {k} are non-overlapping subsets of the w transmission in a

frame. We have

SA∪{k},B ∪ SA,B∪{k} = SA,B (5.5)

Therefore,

PM(SA∪{k},B) + PM(SA,B∪{k}) = PM(SA,B) (5.6)

By setting a = ‖A ∪ {k}‖ and b = ‖B‖, ζM(a, b) satisfies the recursion

ζM(a, b) + ζM(a− 1, b+ 1) = ζM(a− 1, b) (5.7)

with the initial condition

ζM(1, b) = PM(Sb+1) . (5.8)

The initial condition follows from the fact that ζM(a, b) does not depend on the specific

transmissions that are successful or unsuccessful but rather only on the number of suc-

cessful and unsuccessful transmissions, i.e. , a and b. Therefore, ζM(1, b) is equal to the

probability that the first b transmissions are unsuccessful and the b+ 1st transmission is

successful, i.e. , PM(Sb+1).

Since we can find ζM from (5.7), a solution to (5.7) that also satisfies the initial

condition is the unique solution to ζM . We prove that (5.4) satisfies (5.7) and the initial

condition and, therefore, is the solution to ζM .


The equation, obtained by substituting (5.4) in (5.7),

a−1∑k=0

(a− 1

k

)(−1)kPM(Sb+k+1) +

a−2∑k=0

(a− 2

k

)(−1)kPM(Sb+k+2) =

a−2∑k=0

(a− 2

k

)(−1)kPM(Sb+k+1)

(5.9)

holds if and only if

a−1∑k=0

(a− 1

k

)(−1)kPM(Sb+k+1) +

a−1∑k=1

(a− 2

k − 1

)(−1)k+1PM(Sb+k+1) =

a−2∑k=0

(a− 2

k

)(−1)kPM(Sb+k+1)

(5.10)

which is true since (a− 1

k

)=

(a− 2

k

), k = 0,(

a− 1

k

)−(a− 2

k − 1

)=

(a− 2

k

), 1 ≤ k ≤ a− 2,(

a− 1

k

)−(a− 2

k − 1

)= 0, k = a− 1 .

(5.11)

It can be easily verified that the initial condition is also satisfied.

To prove Proposition 5.2.1 we note


(w

l

)ζM(l, w − l) (5.12)

where we have subtracted the probability of the cases in which less than r transmis-

sions are successful from the probability that at least one transmission is successful. By

Lemma 5.4.1 we have


(w

l

)l−1∑k=0

(l − 1

k

)(−1)kPM(Sw−l+k+1)

= PM(S)−r−1∑l=1

l−1∑k=0

(w

l

)(l − 1

k

)(−1)kPM(Sw−l+k+1) .

(5.13)

Chapter 6

Simulation Results

In Chapter 4 and Section 5.2, we discussed the theoretical performance of our proposed

repetition-based broadcast protocol as well as similar protocols and provided motivation

for coding to increase throughput and enhance reliability. Analytical study, although

insightful and readily available for obtaining numerical results, has shortcomings in pro-

viding a true picture of the behavior of the studied protocols. These shortcoming are

mainly caused by simplifying assumptions made to make an analytical study tractable.

As mentioned earlier, for obtaining analytical results, we have assumed nodes communi-

cate in an ideal channel in which every node receives a signal from every other transmit-

ting node. Furthermore, the capture effect and adaptive elimination are neglected in the

analytical study.

In an ideal channel, all simultaneous transmissions result in collision. In a non-ideal

channel with capture, however, one of the many simultaneous transmissions may be

successful, resulting in better performance. Note that since we are studying a multiple

access system, we neglect the effect of noise in the system. Therefore, collision is the

only contributor to packet loss.

In this chapter, with the goal of providing a more accurate representation, assuming

a Rician channel with capture effect explained in Section 6.1, we present simulation

78

Chapter 6. Simulation Results 79

results for different protocols. We also consider the effect of adaptive elimination in the

performance of the protocol in the simulation.

6.1 Channel Model

In a Rician fading channel, the received signal, V (t), is the sum of a line-of-sight (LOS)

component and a non-line-of-sight (NLOS) component

V (t) = Aejωt + Bej(ωt+φ) =(A+ Bejφ

)ejωt = Rejθejωt (6.1)

in which Aejωt is the LOS component and Bej(ωt+φ) is the NLOS component. Assuming

the transmitted signal is Atejωt we have

A2 =A2t

dn(6.2)

where d is the distance between the transmitter and the receiver and n is the path

loss exponent, a constant between 2 and 4. B, is a random variable with a Rayleigh

distribution

fB(B) =B

σ2e−B

2/2σ2· (6.3)

It can be seen that E[B2] = 2σ2 and, therefore, the power of the NLOS component is

σ2. Since the power of the NLOS component, Pnlos, is proportional to the power of the

LOS component, Plos, we have

Pnlos = σ2 =PlosK

=A2

2K(6.4)

where K is called the Rice factor. Substituting (6.4) in (6.3) gives

fB(B) =2KB

A2e−KB

2/A2

(6.5)

The pdf of R = |V | is (see e.g. [52])

fR(R) =2KR

A2exp

(−K(1 +

R2

A2)

)I0

(2KR

A

)(6.6)


where I0(.) is the modified Bessel function of the first kind and zeroth order. Therefore,

the pdf of the received power, P = R2

2, is

fP (P ) =2K

A2exp

(−K(1 +

2P

A2))I0

√8K2P

A2

(6.7)

This pdf can be transformed by U = 4KP /A2 to a noncentral chi-square distribution

with 2 degrees of freedom and noncentrality parameter λx = 2K.

In timeslot m, the desired receiver, denoted by u0, receives the power P(m)i from user

ui. In an interference limited network, the desired transmitter, uj, is successful in sending

its packet to u0 in the mth timeslot ifuj ∈ T (m)

u0 /∈ T (m)

P(m)j > 1

β

∑i:ui∈T (m)\{uj}P

(m)i

(6.8)

where T (m) is the set of transmitting users in the mth timeslot and β is the capture ratio.

A message transmitted by uj is successfully delivered if (6.8) is satisfied at least for

one timeslot (or r timeslots, if r packets are coded into w packets) in that frame.

6.2 Protocol Performance

6.2.1 Simulation Setup

In the simulation setup, cars are placed on a three-lane road with 4m lane separation

and the distance between two adjacent cars in the same lane is 30m, as illustrated in

Fig. 6.1. The simulation is performed in MATLAB. The received power by a vehicle

from any other vehicle is randomly driven according to the Rician distribution of (6.7)

with K = 3 and n = 2. The capture ratio, β is 0.2 unless otherwise stated. The number

of cars, N , is 31 which occupy 300m of road. Frame length, L, is 64. Data rate is 5Mbps

and safety message size, after adding the overheads of different layers, is 200 bytes. When


200B is transmitted in each timeslot, the length of each timeslot is 320µs and each frame

is 20.48ms. In an actual implementation, timeslots must be longer to compensate for

non-ideal synchronization. The traffic model is binomial as shown in (4.43); in each

frame, a message arrives at each node with probability µp.

−150 −100 −50 0 50 100 150−20

−10

0

10

20

Fig. 6.1: Map of roadway and cars.

6.2.2 Probability of Success

An important metric in the simulation results in this work is the probability that more

than 90% of the nodes successfully receive the transmitted message, denoted by P (0.9)s .

Fig. 6.2 shows P(0.1)f = 1 − P (0.9)

s , i.e., the probability that more than 10% of the nodes

in the network fail to receive a transmitted message successfully, for w’s from 2 to 8.

Quadratic curves are fitted to the simulation results using MATLAB’s curve fitting tool-

box. Average load, µp, is 0.2 (messages/user/frame); on average each car produces a 200B

message every 20.48ms/0.2=102.4ms. This figure indicates that by choosing a good value

for w, all protocols are capable of delivering messages reliably while OOC performs better

than the other protocols.

When coding is enabled, nodes can code r packets of data into the w transmitted

packets and, therefore, handle larger messages, namely, 400B and 600B for r = 2 and

r = 3, respectively. In these cases, messages arrive with the same probability in each

frame as in r = 1, however, the messages are larger. Fig. 6.3 and 6.4 show P(0.1)f when r


2 3 4 5 6 7 8-3

-2.5

-2

-1.5

-1

w

log[Pf(0.1) ]

SPR

SFR

OOC

Fig. 6.2: Probability of failure versus w, for µp = 0.2.

is 2 and 3, respectively. For r greater than 1, throughput is higher but it can be observed

from Fig. 6.3 and 6.4 that probability of failure is also higher. The trade-off between

reliability and throughput can be seen by comparing Fig. 6.2, Fig. 6.3, and Fig. 6.4. The

two latter figures also indicate that the probability that OOC delivers more than one

packet successfully is higher compared to the other protocols.

Fig. 6.5 shows the average delay versus w. As mentioned earlier, the delay is defined

as the first timeslot, in which a packet is received successfully by a certain user. The

average delay, in time and among users, of all protocols is more or less the same, as

previously seen in Section 4.3. For all protocols the average delay is less than 24 timeslots

or approximately 8ms. One may also consider the time that a message is buffered until

the beginning of the frame by adding 20.48/2=10.24ms to the above values.

Next we plot the average delay and the probability of failure together to be able

to further analyze the effect of w on overall performance. This is done in Fig. 6.6 for

µp = 0.3. The probability of failure corresponds to r = 1. The closer the point to the

origin, the lower the delay and the probability of failure. From Fig. 6.6, it is observed that


2 3 4 5 6 7 8-1.8

-1.6

-1.4

-1.2

-1

-0.8

-0.6

-0.4

w

log[Pf(0.1) ]

SPR

SFR

OOC

Fig. 6.3: Probability of failure for r = 2 versus w, for µp = 0.2.

2 3 4 5 6 7 8

-1

-0.8

-0.6

-0.4

-0.2

0

w

log[Pf(0.1) ]

OOC

SFR

SPR

Fig. 6.4: Probability of failure for r = 3 versus w, for µp = 0.2.

OOC is the most desirable protocol in terms of probability of failure and average delay.

Also, note that the average delay is always a decreasing function of w while probability

of failure, in some cases, is the lowest for a certain value of w. This is because when


2 3 4 5 6 7 810

12

14

16

18

20

22

24

w

Del

ay (

times

lots

)

OOCSFRSPR

Fig. 6.5: Average delay versus w, for µp = 0.2.

w is larger the average time before the first transmission decreases and since being a

successful transmission is independent of the position of transmission, the average time

before the first successful transmission decreases.

14 16 18 20 22 24 260

0.05

0.1

0.15

0.2

0.25

Delay

Pf(0

.1)

OOCSFRSPR

w=8

w=2

Fig. 6.6: Probability failure versus average delay with w ranging from 2 to 8, for µp = 0.3.


If the number of timeslots in a frame, L, and the number of transmissions, w, are

fixed, the length of timeslots, Ts can be varied by changing r to obtain the highest

probability of success for a given load. This is illustrated in Fig. 6.7 for OOC. In this

figure, the packet size is 200B, w is 7, and L is 64. When r = 1 and Ts = 320µs,

each packet can be transmitted once in each timeslot and the length of frame, Tf , is

LTs = 20.48ms. When r = 2 and Ts = 160µs, each packet is divided into 2 packets with

length 100B which are coded into w packets and transmitted and so on. The effect of a

higher value for r is twofold. On the one hand, when r is higher, Ts is smaller, and so

is Tf . Smaller Tf translates into lower per frame load and hence a higher probability of

success. On the other hand, for a higher r, more transmissions need to be successful for

a successful packet transmission and, hence, a lower probability of success. Therefore,

there exists an optimum value for r for a given load. If the network load is known in

advance, the optimum value for r and Ts can be obtained for a given L and w. In Fig. 6.7,

the probability that more than 90% of the nodes receive the packet successfully, P (0.9)s ,

for different values for r is shown. For very low data rates (0-12kbps), r = 1 gives the

highest probability of success. r = 2 and r = 3 can support a wide range of network load

and maintain relatively high probability of success. r = 4 has an inferior performance

compared to r = 2 or 3; since 7 timeslots are used for transmission, the probability of 4

of them being successfully received is too low.

Fig. 6.8 presents the throughput of the network when load varies. In this figure, L, w,

Ts, and Tf are fixed. The number of packets coded into w packets, r, is shown after the

name of the protocols. It is observed that OOC 3 can support a wide range of network

load and maintain relatively high throughput. OOC 4 supports the widest range but has

an inferior performance to OOC 3.

Fig. 6.9 shows the performance for different QoS levels in a network using OOC. In

this figure, codes are generated for w equal to 7. High priority packets use all 7 available

timeslots for transmission while low priority packets transmit in only 5 of the 7 available


0 25

50

50 75.50.65

0.7

0.75

0.8

0.85

0.9

0.95

1

Average load per user (kbps)

Ps(0.9)

r=1, Ts=320µs

r=2, Ts=160µs

r=3, Ts=106.67µs

r=4, Ts=80µs

Fig. 6.7: Probability of success versus average load, for w = 7.

0 50 100 150 200 250 300 3500

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

Average load per user (kbps)

Thr

ough

put (

Mbp

s)

OOC 1OOC 2OOC 3OOC 4SFR 1SFR 2SFR 3

Fig. 6.8: Throughput versus average load, for w = 7.

timeslots. 30% of the incoming traffic in this simulation is high priority and the rest is

low priority. The performance of a system that does not offer different levels of QoS is

also plotted. It can be observed that the system can provide different QoS levels while


keeping the average performance similar to that of a network without QoS provisioning.

The effect of QoS provisioning is more apparent when the load is high.

0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10.4

0.5

0.6

0.7

0.8

0.9

1

Average load (packets/user/frame)

Ps

Low QoSHigh QoSAverage QoSSystem with no QoS

Fig. 6.9: Providing different QoS levels with OOC.

Fig. 6.10 shows the effect of adaptive elimination for OOC. In this figure, the prob-

ability that more than 90% of the nodes successfully receive a transmitted message is

shown for w = 7. It is observed that adaptive elimination increases the probability of

success particularly in high load.

The analytical results presented in Chapter 4 correspond to a case in which the wire-

less channel is perfect and capture and adaptive elimination are disabled. In the simula-

tions, however, we have considered a more general case: Rician channel with capture and

adaptive elimination. If capture and adaptive elimination are disabled, as expected, the

analytical results and the simulation results agree. This is shown in Fig. 6.11 and 6.12,

for the probability of success while w changes and in Fig. 6.13, for the delay while the

average load changes. Irregularities in the curves for OOC can be observed more often

because OOC has less intrinsic randomness compared to the other two methods.


0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Average load (messages/user/frame)

Ps(0

.9)

r=1 with AEr=1 without AEr=2 with AEr=2 without AEr=3 with AEr=3 without AE

Fig. 6.10: Effect of Adaptive Elimination.

2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 70.75

0.8

0.85

0.9

0.95

1

w

Ps

OOC Ana.SFR Ana.SPR Ana.OOC Sim.SFR Sim.SPR Sim.

Fig. 6.11: Comparison of analytical and simulation results: Ps vs w, for µp = 0.3.

6.3 Performance versus Distance

In the last section, performance evaluation was performed without considering the dis-

tance between nodes. In other words, performance metrics were averaged among all


2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 70.4

0.45

0.5

0.55

0.6

0.65

0.7

0.75

0.8

0.85

0.9

w

Ps

OOC Ana.SFR Ana.SPR Ana.OOC Sim.SFR Sim.SPR Sim.

Fig. 6.12: Comparison of analytical and simulation results: Ps vs w, for r = 2 and

µp = 0.3.

nodes with different distances. In this section, we present simulation results that show

performance versus the distance between transmitter and receiver.

6.3.1 Setup

The placement of cars in this simulation is similar to Fig. 6.1 except that there are 61 cars

instead of 31 and the cluster covers 300m on each side of the middle car. The middle

car is the desired receiver and the probability of success is obtained for transmissions

originating from all other cars and plotted versus their distance from the middle car.

With data rate 10Mbps, packet size is 250B, and each timeslot is 200µs. Path loss

exponent, n, is equal to 2, Rice K factor is 3, and β = 0.5 unless otherwise stated. Also,

adaptive elimination is disabled.


0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.910

20

30

40

Del

ay (

times

lots

)

OOC

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.910

20

30

40

Del

ay (

times

lots

)

SFR

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.910

20

30

40

Del

ay (

times

lots

)

μp

SPR

Fig. 6.13: Comparison of analytical and simulation results: Delay vs Average load, for

w = 6 and β = 0.

6.3.2 Results

The simulation is performed for N = 61, µp = 0.3 and two different pairs of values for

(L,w), namely (64, 6) and (94, 8). Probability of success versus distance is shown in

Fig. 6.14 and Fig. 6.15. As observed in these figures, performance improves for higher

L and w, for all schemes. This is because larger frames result in fewer collisions and a

larger w, while L is sufficiently large, increases the probability of success. Note, however,

that since the per frame load in both cases is µp = 0.3, the data rate for (L,w) = (94, 8)

must be higher to decrease the duration of a timeslot to maintain the same duration for

frames in both cases. From Fig. 6.14 and Fig. 6.15, it is observed that OOC performs


better than SFR and SPR both in terms of probability of success and delay. SFR, in

turn, is better than SPR.

50 100 150 200 250 300

0.7

0.75

0.8

0.85

0.9

0.95

1

Distance(m)

Ps

OOC

SFR

SPR

Fig. 6.14: Probability of success vs. distance from the receiver for N = 61, L = 64,

w = 6, µp = 0.3

To study the performance of different QoS levels with varying distance, the simulation

is modified as follows. In each pair of the cars, which have equal distance from the desired

receiver, one has a high number of repetitions and the other one has a low number of

repetitions. A code with weight 6 is used, but for low-priority users, two of the repetitions

are randomly eliminated. As observed in Fig. 6.16 nodes with higher repetitions have

higher probability of success compared to users in no-priority setup in which all users have

6 retransmissions, while nodes with lower repetitions have lower probability of success.

The result of the simulation with β = 0 is plotted alongside the results obtained in

Chapter 4 for each of the three methods for N = 31, L = 64, w = 6, µp = 0.3 in Fig. 6.17.

It can be observed that analytical results with no capture (β = 0) provides a lower bound

to system performance.


50 100 150 200 250 300

0.84

0.86

0.88

0.9

0.92

0.94

0.96

0.98

1

Distance(m)

Ps

OOC

SFR

SPR


w = 8, µp = 0.3

50 100 150 200 250 300

0.7

0.75

0.8

0.85

0.9

0.95

1

Distance(m)

P s

High QoS

No QoS

Low QoS


whigh = 6, wlow = 4, µp = 0.3


20 40 60 80 100 120 140

0.91

0.92

0.93

0.94

0.95

0.96

0.97

0.98

0.99

1

Distance(m)

Ps

OOC

OOC

SPR

SPR

SPR

SFR

OOC

SFR

SFR

Fig. 6.17: Simulation results for β = 0.5 (solid lines), β = 0 (dashed lines), and analytical

results for β = 0 (dotted lines); N = 31, L = 64, w = 6, µp = 0.3

6.4 Discussion

At the beginning of this chapter, we described the channel model used in the simulations.

The design of a medium access control protocol must not be tied to a specific channel

model. However, a non-ideal channel model is useful for obtaining realistic results. Al-

though the channel model assumed here may not be the best channel model for vehicular

environments, it is chosen merely as a vehicle for obtaining simulation results.

In most parts of the simulations, messages with length 200B are issued from each

vehicle approximately 5 times per second. As explained in Section 1.3, message frequency

of approximately 5 messages with length 100B (per second per user) is sufficient for

communicating position and other useful information. After adding different overheads,

the message length will not exceed 200B assumed in the simulations.

With the described load characteristics, we have shown that OOC-based broadcast


can reliably deliver safety messages with low delay. Furthermore, OOC-based broadcast

performs noticeably better than random repetition broadcast protocols.

By using coding, we have shown that it is possible to increase the throughput of

the network. An advantage of coding is that it enables users to locally determine the

level of coding based on network load and required bandwidth and reliability. We have

also shown that adaptive elimination can significantly increase the probability of success

particularly in high load.

Providing different levels of quality of service is a desirable property of vehicular

networks. As observed in simulation results, OOC-based broadcast is capable of providing

different service levels with a simple modification. The advantage of QoS provisioning is

more apparent when it is most needed– in far distances and high loads.

Simulation results presenting performance versus the distance between transmitter

and receiver reveal that even in higher load and with N = 61 the probability of success

of OOC does not fall below 75% at distance of 300m.

We conclude that OOC-based repetition broadcast provides good performance in

vehicular environments and offers features that are desired in a vehicular communication

system.

Chapter 7

Conclusions and Future Work

The potential of Dedicated Short Range Communication (DSRC) for providing a commu-

nication platform has transformed it to the main enabling technology behind the emer-

gence of a new type of communication systems– vehicular communication networks. It is

envisioned that the appeal of Intelligent Transportation System (ITS) technologies for in-

creasing safety, enhancing productivity, and providing connectivity will result in increas-

ing deployment of DSRC devices. Along with Global Positioning Systems (GPS) and/or

other localization technologies, DSRC makes the development of cooperative/active col-

lision avoidance systems possible. However, realization of a cooperative/active collision

avoidance system, as well as other ITS applications, requires a well designed network,

part of which is the medium access control protocol.

In this work, we have presented a novel MAC broadcast protocol which, unlike other

proposed MAC protocols, is capable of delivering short safety messages in a reliable,

fast, and efficient manner. This protocol relies on Optical Orthogonal Codes (OOC)

which are a well-known family of codes used in Optical Code Division Multiple Access

(OCDMA). We have shown that good cross-correlation properties of these codes can

reduce the possibility of collision in a wireless network.

A detailed mathematical analysis has been provided that enables us to find the prob-

95

Chapter 7. Conclusions and Future Work 96

ability of success and the average delay and is extended to cases that coding is used to

achieve higher throughput or reliability. The analysis presented here takes on a general

approach and not only is applicable to broadcast protocols discussed here, but also can be

used for any repetition-based broadcast protocol with a fixed number of transmissions.

The results of the analysis reveal that OOC-based broadcast can provide significant

performance improvements over previously proposed random repetition-based broadcast

protocols.

After providing mathematical analysis for the basic protocol, we introduced enhance-

ments to the protocol that increase reliability and decrease the probability of collision.

The use of coding is proposed as a means of increasing efficiency, flexibility, and network

throughput. We showed that for a given load and number of transmissions, there exist an

optimum level of coding which provides the highest success probability. Adaptive elim-

ination is introduced to further increase the good cross-correlation properties of optical

orthogonal codes on the fly. Significant performance improvements obtained using adap-

tive elimination is presented by simulation. Furthermore, quality of service provisioning

is also presented using different number of transmissions for different types of traffic.

Simulation results presented in this work showed that the proposed protocol is capable

of supporting safety messaging under load conditions expected in vehicular environments.

Furthermore, the agreement of simulation results performed with ideal assumptions of

the analysis and the analysis strengthens the validity of the results provided in this work.

Overall, the protocol proposed in this thesis shows promising performance and pro-

vides features required in a vehicular medium access control protocol.

7.1 Future Work

In this work, we did not consider the use of acknowledgements. While RTS/CTS trans-

mission is inefficient because of the short length of safety messages, piggybacked acknowl-

Chapter 7. Conclusions and Future Work 97

edgements can be implemented efficiently. The existence of multiple sources of broadcast

traffic adds to the appeal of piggybacked acknowledgements. This work can be extended

by using acknowledgements.

Designing better transmission patterns may be possible for repetition-based broadcast

protocols. For example, an interesting approach is designing codes that do not require

synchronization. In these codes, any overlap between the transmission times of two

packets transmitted in an asynchronous system must guarantee that other transmission

in the same frame between those two users do not have any overlap. Since this condition

is more restrictive than the condition we assumed in this thesis, code cardinality will

be smaller. The suitability of asynchronous codes as transmission patterns and their

generation requires more exploration.

Several transmissions in a frame offers a unique opportunity for implementing network

coding in a repetition-based broadcast. In this case, messages from different users are

combined in each transmission to help disseminate messages. Network coding can also

provide path diversity for nodes whose line-of-sight is blocked. Use of network coding

in a vehicular network may offer increased reliability and efficiency but requires further

study.

Bibliography

[1] The World Health Organization, “The World Health Report 2002 - Reducing Risks,

Promoting Healthy Life.” http://www.who.int/whr/2002/chapter4/en/index7.

html, 2002.

[2] E. Krug, “World Health Day 2004: Road Safety, Overview Fact Sheet.”

http://www.who.int/entity/world-health-day/previous/2004/en/traffic_

facts_en.pdf, 2004.

[3] U.S. Department of Transportation, Research and Innovative Technology Admin-

istration (RITA), “Vehicular Infrastructure Integration (VII).” http://www.its.

dot.gov/vii/.

[4] U.S. Department of Transportation, Bureau of Transportation Statistics (BTS),

“National Transportation Statistics: Motor Vehicle Safety Data.” http://www.bts.

gov/publications/national_transportation_statistics.

[5] ASTM E2213-03, “Standard Specification for Telecommunication and Information

Exchange Between Roadside and Vehicle Systems – 5 GHz Band Dedicated Short

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

(PHY) Specificifications,” ASTM Int’l, vol. July, 2003.

[6] S. Biswas, R. Tatchikou, and F. Dion, “Vehicle-to-vehicle Wireless Communication

Protocols for Enhancing Highway Traffic Safety,” Communications Magazine, IEEE,

vol. 44, no. 1, pp. 74–82, 2006.

98

http://www.who.int/whr/2002/chapter4/en/index7.html

http://www.who.int/whr/2002/chapter4/en/index7.html

http://www.who.int/entity/world-health-day/previous/2004/en/traffic_facts_en.pdf

http://www.who.int/entity/world-health-day/previous/2004/en/traffic_facts_en.pdf

http://www.its.dot.gov/vii/

http://www.its.dot.gov/vii/

http://www.bts.gov/publications/national_transportation_statistics

http://www.bts.gov/publications/national_transportation_statistics

Bibliography 99

[7] US Department of Transportation, “National ITS Architecture Version 6.” http:

//www.iteris.com/itsarch/html/entity/paents.htm, 2007.

[8] J. Zhu and S. Roy, “MAC for Dedicated Short Range Communications in Intelligent

Transport System,” Communications Magazine, IEEE, vol. 41, no. 12, pp. 60–67,

2003.

[9] B. Hofmann-Wellenhof, H. Lichtenegger, and J. Collins, GPS Theory and Practice.

New York, NY, 2001.

[10] R. Parker, “Cooperative vehicle position estimation,” Master’s thesis, University of

Toronto, 2007.

[11] M. Green, ““ How Long Does It Take to Stop?” Methodological Analysis of Driver

Perception-Brake Times,” Transportation Human Factors, vol. 2, no. 3, pp. 195–216,

2000.

[12] S. Sibecas, C. Corral, S. Emami, G. Stratis, M. Inc, and B. Beach, “On the Suit-

ability of 802.11 a/RA for High-mobility DSRC,” Vehicular Technology Conference,

2002. VTC Spring 2002. IEEE 55th, vol. 1, 2002.

[13] IEEE 802 Committee, “Draft Amendment to STANDARD FOR Information

technology – Telecommunications and information exchange between systems –

LAN/MAN Specific Requirements – Part 11: Wireless LAN Medium Access Con-

trol (MAC) and physical layer (PHY) specifications: Wireless Access in Vehicular

Environments (WAVE),” January 2005.

[14] IEEE P802.11 - Task Group p, “Status of Project IEEE 802.11 Task Group p,

Wireless Access in Vehicular Environments (WAVE).” http://grouper.ieee.org/

groups/802/11/Reports/tgp_update.htm.

http://www.iteris.com/itsarch/html/entity/paents.htm

http://www.iteris.com/itsarch/html/entity/paents.htm

http://grouper.ieee.org/groups/802/11/Reports/tgp_update.htm

http://grouper.ieee.org/groups/802/11/Reports/tgp_update.htm

Bibliography 100

[15] F. Farnoud, “Vehicular Communication Networks,” Term paper, Advanced Network

Architecture, University of Toronto, 2007.

[16] F. Farnoud, B. Hassanabadi, and S. Valaee, “Message Broadcast Using Optical

Orthogonal Codes in Vehicular Communication Systems,” Wireless Networking for

Intelligent Transportation Systems, ICST QSHINE Workshop, Aug. 2007.

[17] F. Farnoud and S. Valaee, “Repetition-based Broadcast in Vehicular Ad Hoc Net-

works in Rician Channel with Capture,” IEEE INFOCOM MOVE Wrkshp, Apr.

2008.

[18] F. Borgonovo, A. Capone, M. Cesana, and L. Fratta, “RR-ALOHA, a Reliable

R-ALOHA broadcast channel for ad-hoc inter-vehicle communication networks,”

Proceedings of Med-Hoc-Net 2002.

[19] Q. Xu, T. Mak, J. Ko, and R. Sengupta, “Vehicle-to-vehicle Safety Messaging in

DSRC,” Proceedings of the 1st ACM international workshop on Vehicular ad hoc

networks, pp. 19–28, 2004.

[20] S. Gupta, V. Shankar, and S. Lalwani, “Reliable Multicast MAC Protocol for Wire-

less LANs,” Communications, 2003. ICC’03. IEEE International Conference on,

vol. 1, 2003.

[21] K. Xu, M. Gerla, and S. Bae, “How Effective is the IEEE 802.11 RTS/CTS Hand-

shake in Ad hoc Networks,” Global Telecommunications Conference, 2002. GLOBE-

COM’02. IEEE, vol. 1, 2002.

[22] S. Y. Ni, Y. C. Tseng, Y. S. Chen, and J. P. Sheu, “The Broadcast Storm Problem in

a Mobile Ad hoc Network,” Proceedings of the 5th annual ACM/IEEE international

conference on Mobile computing and networking, pp. 151–162, 1999.

Bibliography 101

[23] N. Balon, “Increasing Broadcast Reliability In Vehicular Ad Hoc Networks,” Mas-

ter’s thesis, The University of Michigan, 2006.

[24] X. Zhang, H. Su, and H. Chen, “Cluster-Based Multi-Channel Communications

Protocols in Vehicle Ad Hoc Networks,” IEEE Wireless Communications, p. 45,

2006.

[25] F. Yu and S. Biswas, “Self-Configuring TDMA Protocols for Enhancing Vehicle

Safety With DSRC Based Vehicle-to-Vehicle Communications,” Selected Areas in

Communications, IEEE Journal on, vol. 25, no. 8, pp. 1526–1537, 2007.

[26] S. Katragadda, G. Murthy, R. Rao, M. Kumar, and R. Sachin, “A Decentralized

Location-based Channel Access Protocol for Inter-vehicle Communication,” Proc. of

the 57th IEEE Semiannual Vehicular Technology Conference (VTC03 Spring), 2003.

[27] K. Tang and M. Gerla, “MAC Reliable Broadcast in Ad hoc Networks,” Military

Communications Conference, 2001. MILCOM 2001. Communications for Network-

Centric Operations: Creating the Information Force. IEEE, vol. 2, 2001.

[28] M. Taha and Y. Hasan, “VANET-DSRC Protocol for Reliable Broadcasting of Life

Safety Messages,” Signal Processing and Information Technology, 2007 IEEE Inter-

national Symposium on, pp. 104–109, 2007.

[29] J. Luo and J. Hubaux, “A Survey of Inter-Vehicle Communication,” EPFL, Lau-

sanne, Switzerland, Tech. Rep, 2004.

[30] F. Borgonovo, A. Capone, M. Cesana, and L. Fratta, “ADHOC MAC: New MAC

Architecture for Ad Hoc Networks Providing Efficient and Reliable Point-to-Point

and Broadcast Services,” Wireless Networks, vol. 10, no. 4, pp. 359–366, 2004.

Bibliography 102

[31] L. Min-Te Sun, A. Arora, and T. Lai, “Reliable MAC Layer Multicast in IEEE

802.11 Wireless Networks,” Proceedings of the International Conference on Parallel

Processing, pp. 527–536, 2002.

[32] J. Blum, A. Eskandarian, and L. Hoffman, “Challenges of Intervehicle Ad Hoc Net-

works,” Intelligent Transportation Systems, IEEE Transactions on, vol. 5, no. 4,

pp. 347–351, 2004.

[33] G. Bianchi, “Performance Analysis Of The Ieee 802. 11 Distributed Coordination

Function,” IEEE Journal on Selected Areas in Communications, vol. 18, no. 3,

pp. 535–547, 2000.

[34] K. Tang and M. Gerla, “MAC Layer Broadcast Support in 802.11 Wireless Net-

works,” MILCOM 2000. 21st Century Military Communications Conference Pro-

ceedings, vol. 1, 2000.

[35] K. Tang and M. Gerla, “Random Access MAC for Efficient Broadcast Support in

Ad hoc Networks,” Wireless Communications and Networking Conference, 2000.

WCNC. 2000 IEEE, vol. 1, 2000.

[36] S. Ray, J. Carruthers, and D. Starobinski, “RTS/CTS-induced Congestion in Ad hoc

Wireless LANs,” Wireless Communications and Networking, 2003. WCNC 2003.

2003 IEEE, vol. 3, 2003.

[37] M. Sun, W. Feng, T. Lai, K. Yamada, H. Okada, and K. Fujimura, “GPS-based

Message Broadcast for Adaptive Inter-vehicle Communications,” Proceedings of Ve-

hicular Technology Conference, vol. 6, pp. 2685–2692, 2000.

[38] G. Korkmaz, E. Ekici, and F. Ozguner, “Black-Burst-Based Multihop Broadcast

Protocols for Vehicular Networks,” Vehicular Technology, IEEE Transactions on,

vol. 56, no. 5 Part 2, pp. 3159–3167, 2007.

Bibliography 103

[39] Q. Xu, T. Mak, J. Ko, and R. Sengupta, “Layer-2 Protocol Design for Vehicle Safety

Communications in Dedicated Short Range Communications Spectrum,” Intelligent

Transportation Systems, 2004. Proceedings. The 7th International IEEE Conference

on, pp. 1092–1097, 2004.

[40] Q. Xu, T. Mak, and R. Sengupta, “Medium Access Control Protocol Design for

Vehicle-vehicle Safety Messages,” Vehicular Technology, IEEE Transactions, vol. 56,

no. 2, pp. 499–518, 2007.

[41] J. Kuri and S. Kasera, “Reliable Multicast in Multi-Access Wireless LANs,” Wireless

Networks, vol. 7, no. 4, pp. 359–369, 2001.

[42] A. Toyserkani, E. Strom, and A. Svensson, “An Efficient Broadcast MAC Scheme

for Traffic Safety Applications in Automotive Networks,” Proc. of the WCNC. Las

Vegas: IEEE Press, vol. 2105, 2006.

[43] C. Chiu, E. Wu, and G. Chen, “A Reliable and Efficient MAC Layer Broadcast

(Multicast) Protocol for Mobile Ad hoc Networks,” Global Telecommunications Con-

ference, 2004. GLOBECOM’04. IEEE, vol. 5.

[44] E. Agrell, A. Vardy, and K. Zeger, “Upper Bounds For Constant-weight Codes,”

Information Theory, IEEE Transactions, vol. 46, no. 7, pp. 2373–2395, 2000.

[45] S. Johnson, “A New Upper Bound for Error-correcting Codes,” Information Theory,

IEEE Transactions, vol. 8, no. 3, pp. 203–207, 1962.

[46] K. Jackson, S. Newton, B. Moslehi, M. Tur, C. Cutler, J. Goodman, and H. Shaw,

“Optical Fiber Delay-line Signal Processing,” IEEE Transactions on Microwave

Theory and Techniques, vol. 33, no. 3, pp. 193–210, 1985.

Bibliography 104

[47] J. A. Salehi, “Code Division Multiple-access Techniques in Optical Fiber Networks.

I. Fundamental Principles,” Communications, IEEE Transactions on, vol. 37, no. 8,

pp. 824–833, 1989.

[48] H. Shalaby, “Chip-level Detection in Optical Code Division Multiple Access,” Light-

wave Technology, Journal of, vol. 16, no. 6, pp. 1077–1087, 1998.

[49] A. E. Brouwer, J. B. Shearer, N. J. A. Sloane, and W. D. Smith, “A New Table of

Constant Weight Codes,” Information Theory, IEEE Transactions, vol. 36, no. 6,

pp. 1334–1380, 1990.

[50] T. Cormen, C. Leiserson, R. Rivest, et al., Introduction to algorithms. Cambridge,

Mass.: MIT Press, 2001.

[51] F. MacWilliams and N. Sloane, The Theory of Error-correcting Codes. North-

Holland Amsterdam, 1988.

[52] R. Prasad and C. Liu, “Throughput Analysis of Some Mobile Packet Radio Protocols

in Rician Fading Channels,” Communications, Speech and Vision, IEE Proceedings

I, vol. 139, no. 3, pp. 297–302, 1992.

Reliable Broadcast of Safety Messages in Vehicular Ad hoc ......tions, that the proposed protocol...

Documents

Transcript of Reliable Broadcast of Safety Messages in Vehicular Ad hoc ......tions, that the proposed protocol...