emergency information dissemination in vehicular ad hoc
TRANSCRIPT
Emergency Information Disseminationin
Vehicular Ad Hoc Networks
A thesis submitted in partial fulfillment of the requirementsfor the degree of
Master of Technology (Honours)
in
Computer Science and Engineering
by
Rayman Preet Singh
06CS3023
advised by
Dr. Arobinda Gupta
Department of Computer Science and EngineeringIndian Institute of Technology, Kharagpur
May 2011
Certificate
This is to certify that the thesis entitled Emergency Information Dis-
semination in Vehicular Ad Hoc Networks submitted by Rayman Preet
Singh (06CS3023) to the Department of Computer Science and Engineering
is a bona fide record of research work carried out by him under my supervi-
sion and guidance. This thesis has fulfilled all the requirements as per the
regulations of the institute and, in my opinion, has reached the standard
needed for submission.
Dr. Arobinda Gupta
Professor
Department of Computer Science and Engineering
IIT Kharagpur
May 2011
Acknowledgment
I would like to express my gratitude towards Prof. Arobinda Gupta for
the supervisory role he played to utmost perfection. Taking time out of his
busy schedule, he ensured that the project work was carried out methodically
and meticulously. I especially thank him for his encouragement and his
accurate comments which were of critical importance, and am indebted to
him for extending out all the necessary support throughout the duration of
the project and for being a constant source of inspiration.
I would also like to thank my parents, S. Joga Singh and Narinderjit Kaur,
and friends, Sayantan Ghosh and Anirudha Patro for their invaluable help,
guidance and motivation. Their continuous support and encouragement has
played a key role in the completion of this work.
Rayman Preet Singh
06CS3023
Department of Computer Science and Engineering
IIT Kharagpur
May 2011
Abstract
The main goal of inter-vehicle communication technology is to provide
the drivers with more information regarding the surroundings than they can
visually perceive. A Vehicular Ad-Hoc Network, or VANET, is one such tech-
nology that uses moving vehicles as nodes in a network to create a wireless
ad hoc network. Many applications in VANETs are based on dissemination
of information to the drivers about hazardous situations and can help avoid
potential dangers. Such applications require that the information is delivered
to all vehicles which are within a certain area and is retained in that area
for a certain period of time. Most of the existing information dissemination
protocols for VANETs do not provide this much needed retention of infor-
mation and little research has been done in this regard. Stored Geocast, is an
information dissemination protocol which provides a retention of information
within a pre-determined area of the road. However, the overhead incurred
by it is large and makes it unsuitable for deployment over VANETs.
In this thesis, we propose an information dissemination protocol which
provides both a spatial and temporal retention of information and incurs
less overhead. Experimental results are used to draw a comparison between
the performance of the existing information dissemination protocols. Per-
formance of the proposed algorithm is evaluated and compared with Stored
Geocast and other existing information dissemination protocols.
Contents
Contents i
1 Introduction 1
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Thesis Overview . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2 Background and Literature Overview 6
2.1 Existing Protocols for Information Dissemination . . . . . . . 7
2.1.1 Contention Based Approach . . . . . . . . . . . . . . . 8
2.1.2 Improvements over Contention Based Approach . . . . 9
2.1.3 Congestion Based Approach . . . . . . . . . . . . . . . 10
2.2 Stored Geocast . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3 Problems with Existing Approaches . . . . . . . . . . . . . . . 13
3 Zone Based Forwarding 15
3.1 Overview of Proposed Strategy . . . . . . . . . . . . . . . . . 15
3.2 Description of the Algorithm . . . . . . . . . . . . . . . . . . . 17
3.3 Pseudocode of the Algorithm . . . . . . . . . . . . . . . . . . 18
i
3.4 An Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.5 Overcoming Network Congestion . . . . . . . . . . . . . . . . 23
4 Experimental Study 26
4.1 Simulation Setup . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.2 Performance Evaluation . . . . . . . . . . . . . . . . . . . . . 28
4.3 Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.3.1 Comparison of Existing Approaches . . . . . . . . . . . 29
4.3.2 Performance Evaluation of ZBF . . . . . . . . . . . . . 32
4.3.3 Performance Evaluation of ZBF Extension . . . . . . . 38
5 Conclusion 41
5.1 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
ii
List of Figures
3.1 Method invoked by vehicle i for initializing or relaying the
dissemination . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.2 Method invoked at vehicle i on receipt of a Query message . . 20
3.3 Method invoked at vehicle i for broadcast of an Info message . 20
3.4 Method invoked at vehicle i on receipt of a Reply message . . 21
3.5 Method invoked at vehicle i on receipt of an Info message . . . 21
3.6 Sample scenario with vehicle i as initiating vehicle. Blocks de-
pict the vehicles and arrowheads denote their direction(North/South)
of movement. vehicle i (solid block) initiates the dissemination. 22
3.7 Method invoked at vehicle i on receipt of an Info message . . . 24
3.8 Method invoked at vehicle i on non receipt of a Query message
due to message drop . . . . . . . . . . . . . . . . . . . . . . . 25
4.1 Comparative performance evaluation of existing strategies :
Number of Vehicles Informed (Coverage) . . . . . . . . . . . . 29
4.2 Comparative performance evaluation of existing strategies :
Number of Messages Broadcasts . . . . . . . . . . . . . . . . . 30
4.3 Performance of the strategy proposed by Fathy and Khakbaz
in [3]. Information spreads to all vehicles as time progresses,
starting at t=170 . . . . . . . . . . . . . . . . . . . . . . . . . 31
iii
4.4 Total number of vehicles informed v/s Frequency (f) of peri-
odic broadcast. . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.5 Total number of messages broadcast v/s Frequency (f) of pe-
riodic broadcast. . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.6 Distribution of different types of messages broadcast v/s Fre-
quency(f) of periodic broadcast. . . . . . . . . . . . . . . . . . 35
4.7 Distance inside a zone which a vehicle traversed before being
informed v/s Frequency(f) of periodic broadcast. . . . . . . . . 36
4.8 Total number of messages broadcast per informed vehicles v/s
Frequency(f) of periodic broadcast. . . . . . . . . . . . . . . . 37
4.9 Result of the Fathy algorithm [3] on a sample highway scenario. 38
4.10 Number of vehicles reached v/s Frequency(f) of periodic broad-
cast. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.11 Total number of messages broadcast v/s Frequency(f) of peri-
odic broadcast. . . . . . . . . . . . . . . . . . . . . . . . . . . 40
iv
Chapter 1
Introduction
A Mobile Ad Hoc Network, or MANET, is a wireless network consisting of
mobile nodes in which all the network-level activity is carried out by the
nodes themselves, without additional infrastructure support. Each node in
such a network plays the role of a router as well as an end-machine, and
hence, all nodes in the network participate actively in message forwarding.
The network topology is subject to change owing to the mobility of the nodes.
A Vehicular Ad-Hoc Network, or VANET, is a special kind of MANET
in which mobile nodes are all vehicles equipped with an On-Board Unit
(OBU) that enables them to send and receive messages to and from each
other. Additionally, a VANET might also interface with communication
points provided by on-road infrastructure, commonly termed as Road-side
Units (RSUs).
Several applications have been proposed for VANETs. Bai et al. [4]
survey these applications and divide them into following categories:
1. Safety applications, in which VANET is used to identify scenarios
that could potentially endanger the vehicle drivers’ safety. For instance,
drivers could benefit from real-time alerts about accidents happening in
their vicinity. Road Hazard Condition Notification (RHCN) is a safety
application in which a vehicle detecting a road hazard (e.g. fluid, ice)
1
Chapter 1: Introduction
notifies vehicles within the potentially affected region. Likewise, Emer-
gency Electronic Brake Light (EEBL) is a safety application in which
a vehicle braking hard notifies vehicles in its neighborhood. Other
examples of safety applications in VANETs include, Road Feature No-
tification (RFN) [4], Cooperative Collision Warning (CCW) [5] and
Cooperative Violation Warning (CVW) [7]. Note that, in each of these
applications some form of warning is emitted on the occurrence of a
designated event associated with the warning.
2. Convenience and commercial applications, in which OBUs com-
municate with each other in order to provide the driver such services
as traffic congestion and flow information, parking availability informa-
tion, automatic toll payment etc. Traffic congestion information can
be propagated amongst vehicles and can be used for producing a more
fluid traffic on roads and avoiding traffic jams. Commercial applica-
tions in VANETs include internet access, streaming audio and video
etc.
In this thesis, we shall be addressing the problem of designing the under-
lying strategy which needs to be in place to spread the warning information
associated with different safety applications. Traditionally, information dis-
semination in VANETs is achieved by means of a flooding mechanism, and
vehicles undertake subsequent re-broadcasts to overcome network fragmenta-
tion. However, the information in which a given vehicle might be interested,
is dependent on its current physical location, and hence an indiscriminate
propagation of information does not suffice for the needs of most VANET
applications designed for this purpose.
1.1 Motivation
The process of information dissemination is initiated on the occurrence of
an emergency event, and has associated with it a protocol for sending and
2
Chapter 1: Introduction
receiving messages used to spread the warning information. An efficient
dissemination process should possess the following characteristics :
1. Maximum coverage across relevant vehicles: Any warning infor-
mation usually has a specified area of interest and a fixed time period
within the limits of which it is considered valid and usable by the drivers
receiving it. These are referred to as the effect area and effect time of
the warning information. For instance, warning information about a
slippery road has an effect area equal to the segment of road on which
the road is slippery and effect time equal to a certain estimated fixed
time duration for which the slipperiness will last. Thus, in order to be
effective any warning information dissemination strategy should guar-
antee the delivery of information to all vehicles lying within the effect
area and effect time.
2. Minimum latency: Since warning information is directly linked to
the safety of the drivers on the road, the time difference between the
actual occurrence of the event and the receipt of the associated warning
information is desired to be as small as possible. Hence, a warning
information dissemination strategy is required to minimize the latency
of dissemination.
3. Low overhead: In addition to latency, the efficiency of an information
dissemination protocol is also governed by the amount of data other
than the warning information, which it sends over the wireless ad-hoc
network. An indiscriminate flooding can cause the channels to get
jammed and cause packet drops.
Imposing a restriction on dissemination of information beyond the effect
area will also stop transmission of warning information to vehicles for which
it is of little significance. This reduces wastage of bandwidth over the wireless
channel and improves channel availability.
Several schemes have been proposed and employed for propagating infor-
mation in VANETs [2] [3] [9] [12] [10]. They aim to spread the information to
3
Chapter 1: Introduction
as many number of vehicles as possible. This leads to the information being
disseminated in a large area which often includes the effect area. However, a
retention of information in the effect area is not maintained. Due to this ab-
sence of retention, vehicles coming into the effect area after the propagation
of information has ceased, do not get the warning information.
Little research has been done on designing information dissemination pro-
tocols which provide retention of information. Stored Geocast [10] is an in-
formation dissemination protocol for VANETs which provides a retention of
information in the effect area. However, the overhead incurred by it is large
and makes it unsuitable for deployment over VANETs. Thus, an information
dissemination protocol for VANETs which provides both retention of infor-
mation in its effect area during the duration of its effect time and incurs less
overhead is needed. If such a information dissemination protocol is available
then it can be used by VANET applications such as RHCN and RFN to effi-
ciently retain the information within the effect area of the event and deliver
it to incoming vehicles.
1.2 Contributions
In this thesis we first present a detailed analysis of the existing information
dissemination protocols in VANETs. We identify that most of the dissemi-
nation existing protocols do not retain the warning information within the
effect area for the duration of the effect time. This is a crucial requirement
for designing safety applications for VANETs. However, Stored Geocast [10]
fulfills this requirement but incurs large overhead. We propose an efficient
information dissemination strategy which incurs small overhead and is capa-
ble of retaining the information within the effect area for the entire duration
of the effect time. The strategy is formalized as an algorithm and analytical
results quantifying its performance are presented.
We use experimental results to draw a comparison between the perfor-
mance of the existing dissemination strategies. Further, we evaluate the
4
Chapter 1: Introduction
performance of the proposed algorithm and compare its performance with
Stored Geocast [10] and other existing information dissemination protocols.
1.3 Thesis Overview
The thesis is further organized as follows :
• Chapter 2 presents a survey of related research work on informa-
tion dissemination in VANETs. Some background and essential no-
tions related to our work are also incorporated. Finally, an analysis
of the shortcomings of existing information dissemination protocols in
VANETs is presented.
• Chapter 3 discusses the proposed strategy for information dissemina-
tion in VANETs. We formally describe our proposed algorithm, along
with a discussion and a sample run on a sample scenario. The chapter
ends with the description of a proposed extension of our algorithm,
which counters network congestion if present.
• Chapter 4 describes the system setup employed for simulation and
performance evaluation of various information dissemination strategies.
We describe, in detail, the measures we have adopted to evaluate the
performance evaluation. Further, an evaluation of the relative per-
formance of the existing approaches for information dissemination is
presented, followed by a performance evaluation of the proposed algo-
rithm. Finally, we analyze the performance of our proposed approach
and comment on its performance in relation to other dissemination
strategies in its class.
• Chapter 5 summarizes the contributions of this thesis, and identifies
possible directions for future research.
5
Chapter 2
Background and Literature
Overview
The mode of communication in VANETs involves broadcast of messages on
the 5.9Ghz frequency band licensed for VANET applications. We assume
that these broadcast messages are delivered in single-hop, using IEEE 802.11p
(or Dedicated Short Range Communications, DSRC) technology pursued by
industry and governments [13].
In VANETs, there are two types of transmission carried out by the ve-
hicles: event-driven messages, whose transmission takes place on the occur-
rence of certain well-defined events, such as the transmission of safety mes-
sages, and periodic transmissions called beacons used for providing mutual
awareness. Hence, a broadcast of a message containing warning informa-
tion is an event-driven message since it is triggered by the occurrence of a
hazardous situation. Note that, warning information messages may also be
re-broadcast periodically for dissemination. Slow Vehicle Advisory (SVA),
Road Hazard Condition Notification (RHCN), and Emergency Electronic
Brake Light (EEBL) described in [4] are some examples of event-driven mes-
sages that are transmitted by vehicles, which are used for ensuring safety
of vehicles and also for better coordination among vehicles. A SVA alert
is broadcast by a vehicle that suddenly slows down or stops on the road,
6
Chapter 2. Background and Literature Overview
enabling the recipients to take evasive action. A RHCN alert is broadcast
by a vehicle that observes a possibly hazardous condition on the road, such
as road slipperiness, to warn other vehicles about this condition. An EEBL
alert is broadcast by a vehicle which is braking hard, to enable the recipients
to take evasive action. However, certain convenience applications such as
Congested Road Notification (CRN) also employ event-driven messages. A
beacon message is a periodic message broadcast by a vehicle and contains its
positional information, speed and direction at the time of broadcast.
In the following section, we first provide a brief insight into the existing
information dissemination protocols in VANETs. We then look at one of the
existing protocols, Stored Geocast in more detail as it is the protocol that is
of most relevance to the work presented in this thesis. We end this chapter by
stating the shortcomings present in the existing approaches for information
dissemination, which forms the motivation for our proposed algorithm in the
next chapter.
2.1 Existing Protocols for Information Dis-
semination
In the classical flooding protocol [1], the initiating vehicle starts the dis-
semination by sending the information messages to all of its neighbors, and
any vehicle receiving the message broadcast re-broadcasts exactly once. In
addition to this, the vehicles periodically rebroadcast the message in order
to overcome network fragmentation. This flooding strategy causes a lot of
redundant broadcasts as well as redundancy in message delivery. Several
strategies have been suggested that improve upon this simple flooding strat-
egy.
7
Chapter 2. Background and Literature Overview
2.1.1 Contention Based Approach
Torrent-Moreno [2] presents a strategy where one vehicle initiates the dis-
semination by broadcasting the information. This strategy employs a con-
tention mechanism to select intermediate vehicles (forwarders) which will
re-broadcast the message so as to carry forward the propagation. The vehi-
cles receiving the message decide by means of a waiting period, as to which
one must forward the message.
Each vehicle receiving the message first computes its distance (in meters)
from the sender, and is referred to as the progress (denoted as P ) of the
vehicle. Each vehicle then assigns itself a waiting time (denoted as t(P ))
according to the following rules :
t(P ) =
∞ if P > rCA
T ′.(1− P
rCA
) if rCA ≥ P ≥ 0m
∞ if P < 0m
where T ′ is the maximum waiting time; and rCA fixes the radius of the area
where potential forwarders must be located, and is typically equal to the
transmission range R of the vehicles (assumed equal for all vehicles). There-
fore, of all the vehicles that received the sender’s broadcast, the vehicle at
the largest distance (P ) from the sender vehicle will select the shortest wait-
ing time, after which it will re-broadcast the information so as to propagate
it. Upon reception of the re-broadcast message, vehicles which are still con-
tending will cancel their contention process. The vehicles which receive the
warning information for the first time via this re-broadcast then repeat the
contention process to carry forward the propagation.
Note that this approach is vulnerable to a partitioned network and fails
to propagate the information in the presence of network partitions across the
VANET. The broadcasting vehicle does not detect whether a new intermedi-
ate vehicle (forwarder) has been selected by the contention process or not. In
the absence of a forwarding vehicle, the propagation stops at the boundary
8
Chapter 2. Background and Literature Overview
of that connected component of the VANET containing the vehicle initiating
the dissemination process.
2.1.2 Improvements over Contention Based Approach
A more reliable method for disseminating safety information is presented by
Fathy and Khakbaz in [3]. They employ the Torrent-Moreno [2] contention
based approach (with rCA = R) for propagation of information, but propose
certain improvements so as to improve upon its shortcomings. Between-
ness conditions proposed by them aim to ensure that during the contention
process, of all the vehicles that receive a re-forwarded message from the re-
broadcasting (farthest) vehicle, only the vehicles present between the actual
sender and the re-broadcasting vehicle cancel their contention process on re-
ceipt of the re-broadcast message. This is important because in case the
re-broadcasting (farthest) vehicle has a progress P less than rCA, vehicles on
the opposite side of the sender can cancel their contention on receiving the
rebroadcast.
Additionally, they also propose a connectivity hole detection method. If
the last broadcasting vehicle does not receive a re-broadcast from one of the
vehicles in the contention process within a fixed time interval, a connection
hole is deemed to be detected. After detection of a connectivity hole, the
vehicle shall start sending small periodic hello messages which contain their
location and ID. Vehicles receiving hello messages reply by a similar message
containing their location and ID. If the vehicle receives a hello message that
shows the entrance of a vehicle in direction of dissemination, it will broadcast
the information message and by assigning the forwarding task to the new
entered vehicle, propagation is continued. The hello messages aim to bridge
the network fragments, so that whenever a vehicle travels to a new fragment
it can repeat the propagation by inducing Torrent-Moreno’s contention based
approach.
9
Chapter 2. Background and Literature Overview
2.1.3 Congestion Based Approach
Many VANET applications assume that each vehicle knows its neighbors
through exchanging periodic beacon messages. By use of these periodic bea-
cons, vehicles can measure the level of vehicular congestion around them.
One such approach to measure vehicular congestion was proposed in [9], in
which the vehicles measure two kinds of vehicular congestion around them
: Instantaneous Congestion (IC), which is the instantaneous picture of the
traffic in the vicinity of a vehicle, and Stabilized Local Congestion (SLC),
comprising of the neighboring vehicles of a vehicle which have been stable
members of its instantaneous congestion over a period of time.
An information dissemination method based on vehicular congestion es-
timates was proposed in [9]. Each vehicle computes its instantaneous con-
gestion and stabilized local congestion using the period beacon messages. In
addition, each vehicle u possesses a measure (bcastu) which, on receipt of
an information message, is used to decide whether to re-broadcast or not as
follows :
bcastu = q |SLC||IC| + (1− q)bcastv, where q is a constant.
vehicle u re-broadcasts if and only if bcastu ≥ bcastTH . Any sender encloses
its bcast value in the information message that it sends, with the initial sender
using a value of 1. The parameter q provides for a weighted sum between
controlled flooding using the local congestion information, and the traditional
flooding strategy. The weighted flooding component drives the propagation
of information whereas the vehicular congestion estimates are used at the
vehicles to decide whether to re-broadcast the information or not. Hence,
a vehicle re-broadcasts the information only when most of the vehicles in
its immediate neighborhood are stable and in all probability will receive the
information. Additionally, a probabilistic broadcast based on the congestion
around a vehicle is also possible, where a vehicle with high congestion around
it broadcasts with a lower probability p, (and vice versa), as only few of the
vehicles need to broadcast in such an environment (p = 1|SLC|).
10
Chapter 2. Background and Literature Overview
Yu and Heijenk in [12] propose a dissemination strategy which utilizes
the concept of ‘effect line’ and ‘safety line’ near an emergency event, but
their work is based on the assumption of a Poisson distribution model of
traffic volume and involves message broadcast by all vehicles with a dynamic
adjustment to inter-broadcast wait time. However, vehicular traffic has been
characterized to exist in three difference categories of Free Flowing, Wide
Moving Jams and Synchronized Flow, and hence is not liable to follow Pois-
son distribution over the number of vehicles as is assumed in this strategy.
2.2 Stored Geocast
The transmission of a message to some or all vehicles within a geo-graphical
area is referred to as a Geocast. A Stored Geocast is a time stable geocast,
which is delivered to all vehicles that are inside a destination region within
a certain period of time.
Manihofer et al. in [10] proposed an infrastructure-less approach based
on Stored Geocast. This would serve as a solution for location based services
in VANETs, like realizing a virtual traffic sign, which in turn is also a method
to achieve warning information dissemination. On detection of an emergency
event, the vehicles in a VANET can coordinate to setup such virtual traffic
signs which will relay the information to vehicular traffic which is inbound
to the area of interest.
In their proposed approach, Manihofer et al. in [10] propose to elect a
vehicle in the destination region of a geocast message to store information
messages. This dynamically elected vehicle within the destination region is
responsible for storing and delivering the message. A periodic broadcast or
notification-triggered broadcast is adopted for message delivery. A handover
of messages is done when this elected vehicle leaves the destination region
and a process to elect a new leader is started. However, to avoid frequent
handovers, Manihofer et al. suggest that it is desirable to choose one that
stays as long as possible in the destination region. Such a vehicle is charac-
11
Chapter 2. Background and Literature Overview
terized by low velocity and closeness to the center of the destination region.
They propose the use of a leader election algorithm suggested in [11] as a
part of the GeoGRID protocol.
The leader election algorithm described by Liao et al. in the GeoGRID
[11] protocol divides the destination region into a set of 2-D grids and uses a
set of guidelines in order to elect a leader for each grid. Firstly, it is suggested
that the vehicle nearest to the physical center of a grid is to be elected as
the leader because such a host is more stable and more likely to remain in
its grid for a long time. Secondly, once a vehicle is elected as leader it will
remain so until it moves out of its grid.
When a vehicle moves closer to the physical center of the grid it will
not be elected as leader until the earlier one leaves the grid. Periodically,
the elected leader will broadcast its existence by sending a GATE(g,loc)
message, where g is its grid coordinate and loc is its current location. All the
other vehicles monitor the current leader in their grid and if a GATE(g,loc)
message is not received for a predefined time period, they will broadcast a
BID(g, loc) message, where g is the grid coordinate and loc is the vehicle’s
current location.
Upon the leader vehicle (if it is still in the grid) hearing the BID message,
it will reply with a GATE to reject the BID sender’s bid. Upon a non-leader
vehicle at a location closer to the physical center of the grid hearing the BID
message, it will reply with a BID(g, loc′) message to reject the former’s bid,
where loc′ is the sending vehicle’s current location. However, if no such
messages are received by the bidding vehicle for a predefined time period,
the bidding host will silently elect itself as the leader.
When a leader leaves its current grid, it broadcasts a RETIRE(g, T )
message where g is the grid coordinate where it served as a gateway. All
other vehicles on receiving a RETIRE message send BID messages in order
to elect a new leader. Lastly, to eliminate the possibility of having multiple
leaders in a grid, when a vehicle which assumes itself as a leader receives
a GATE message from another vehicle at a location closer to the physical
12
Chapter 2. Background and Literature Overview
center of its grid, it silently changes itself to a non-leader vehicle without
sending any message. We shall refer to this algorithm as Stored Geocast
algorithm.
2.3 Problems with Existing Approaches
With regards to the existing information dissemination approaches discussed
in Section 2.1, we observe that they are all message-centric in nature and aim
to attain maximum coverage in terms of vehicles informed/are covered. How-
ever, emergency warning information in VANETs have a definite area (effect
area) associated with it, within the confine of which it is valid. Likewise,
there exists a time limit (effect time, measured from the start of dissemina-
tion), within the duration of which the emergency information holds valid.
Thus, it is imperative to produce persistence of the warning information in
its effect area for the duration of the effect time, rather than focusing ex-
haustively on achieving coverage. The approaches described in Section 2.1
spread the information to a majority of vehicles around the initiating vehi-
cle (or hazard), but this ignores the possibility, that an uninformed vehicle
visiting the area at a later instant of time (but within the effect time), will
still be deprived of the warning information. In order to ensure that warning
information reaches all vehicles, who are in the effect area at some point of
time within the effect time, repeated broadcast by some of the informed ve-
hicles is necessary. In addition, the congestion based approach assumes the
existence of underlying periodic beaconing by the vehicles, which increases
the network congestion.
Hence, a mechanism is needed which causes warning information to per-
sist in the effect area for the duration of the effect time. However, causing
all the informed vehicles to perform a periodic re-broadcast will amount to
large number of broadcasts and redundancy. Note that, a periodic broadcast
by one vehicle with a transmission range R will ensure the persistence of
information within a road segment of length R.
13
Chapter 2. Background and Literature Overview
As described in Section 2.2, Manihofer et al. in [10] suggest that a vehicle
that stays in the destination region for as long as possible is to be elected
as the leader, and propose the use of GeoGRID [11] as the leader election
algorithm. However, GeoGRID is a grid based approach, based on electing
the vehicle closest to the center of a region (grid box) as the leader. But
in a VANET scenario, the amount of time a vehicle spends in a particular
region is dependent upon factors like velocity, acceleration, driver behavior
and traffic congestion. Thus, for its use in VANETs, a leader election algo-
rithm which addresses these needs should be employed. Secondly, GeoGRID
employs periodic GRID and BID messages, which add to the overhead, as
warning message dissemination is to be carried out only when an emergency
event occurs and not at all times. Hence, in order to disseminate warning
information with a nominal overhead, a reactive protocol is needed rather
than a proactive one.
Finally, an exhaustive simulation and testing of the existing dissemination
strategies in the context of VANETs would have provided justification for the
use of Stored Geocast in VANETs, which has not been addressed in [10] and
[11].
14
Chapter 3
Zone Based Forwarding
In this chapter, we propose a strategy for information dissemination which
overcomes the shortcomings of the existing approaches, as identified in Sec-
tion 2.3. In the following sections, we first present a brief overview of the
proposed approach followed by a formal description of the algorithm and an
example run on a sample scenario. We end this chapter by presenting an
extension of the algorithm which overcomes network congestion encountered
during the dissemination process.
3.1 Overview of Proposed Strategy
Apart from assuming that the transmission range for all vehicles is the same
and equal to R, we make the following assumptions:
• Roads have two-way traffic with lower and upper speed limits as 0 and
vMAX respectively.
• Message propagation delay is negligible in comparison to time required
for a considerable kinematic motion.
• At any vehicle i, its current position (Pi), velocity (vi), acceleration
(ai) and direction of motion (d̂i) are available.
15
Chapter 3. Zone Based Forwarding
• Warning message propagation begins with exactly one vehicle broad-
casting the first warning message.
A periodic broadcast of a message by a vehicle while traveling a distance
R, guarantees the delivery of the message to all vehicles which are present
on that road segment of length R. This guarantee is dependent upon the
frequency (f) of the periodic broadcast. Our strategy involves dividing the
entire effect area of the warning information into segments of length R, each
of which is referred to as a zone. We assign one vehicle in each zone the task
of periodically broadcasting the warning information to notify other vehicles
in that zone. This vehicle is referred to as a forwarder for that zone. For
each zone, the vehicle that spends the longest time in the zone is elected
to be the forwarder. Whenever a forwarder exits a zone a new forwarder is
elected from amongst the vehicles present in the zone and the task of periodic
broadcasting is then handed over to the new forwarder. At the end of the
effect time duration all forwarders stop their periodic broadcast.
Table 3.1 shows the three three kinds of messages employed. The Info
message is used by the forwarder for periodic broadcast. It contains all the
details about the particular warning information and its effect area and effect
time. The Info message also contains the forwarder’s position, direction of
motion and ID. The Query and Reply messages are used for electing a new
forwarder when a forwarder leaves its zone. In addition to the contents of
the Info message, these messages contain a type.
Any vehicle is always in one of three modes {Receive, Forward, Relay}and performs functions accordingly. The Forward mode signifies that the
Message Type ContentInfo Warning Information Content, Effect Time, Effect Area,
Sender position, Sender Direction, Sender IDQuery Info, QueryReply Info, Reply
Table 3.1: Message forms.
16
Chapter 3. Zone Based Forwarding
vehicle is a forwarder. Whenever a forwarder leaves a zone and has to elect
a new forwarder it changes its mode to Relay. All other vehicles, which are
also the recipients of the periodic Info messages are in Receive mode.
3.2 Description of the Algorithm
Before start of dissemination all vehicles are in Receive mode. A vehicle i
initiating the dissemination changes its mode to Relay, broadcasts a Query
message and waits for T time for receiving a Reply message. Any vehicle j
at position Pj, on receiving the Query message, computes distance s which it
needs to travel to reach the current position (Pi) of the sender if it is headed
towards it. In case it is headed away from Pi, s is the distance it needs to
travel to reach the boundary of i’s transmission range (R). Using vj, aj,
and d̂j vehicle j computes the time ts it will take to travel the distance s.
Since a Query message is always broadcast by a forwarder exiting a zone, the
distance s can be computed as follows:
s =
|Pj − Pi| if d̂i = d̂j
R− |Pj − Pi| if d̂i = −d̂j.
Vehicle j then computes time ts which it will take to travel the distance s.
Time ts is computed as
ts =s
vj
where vj is the current speed of vehicle j. Time ts can also be computed as√u2 + ai.s− u
aifor more accuracy. Since our aim is just to estimate which
vehicle will be present in a zone for the longest duration, we prefer to use
(s
vj) for simplicity.
Vehicle j then waits for a time twait, computed using ts, such that twait ∝1
tsand twait ∈ [0, T ]. Time twait is calculated as the maximum of
N
tsand
T , where N is a suitable normalizing factor. After expiration of the twait
period vehicle j broadcasts a Reply message. However, if a Reply message,
17
Chapter 3. Zone Based Forwarding
broadcast from another vehicle, is received during the wait period (twait),
vehicle j cancels its transmission of the Reply message contingent upon the
condition that both vehicle j and the vehicle sending the Reply message are
on the same side of the vehicle which sent the Query message. The vehicle
k which sends a Reply message becomes a forwarder (changes to Forward
mode). The road segment of width R between Pi and i’s transmission range
boundary is the zone for this forwarder. On becoming the forwarder, vehicle
k records its current position Pk as RECXY in order to compute the zone
boundaries.
The forwarder k uses RECXY to detect when it reaches the zone border,
at which point it broadcasts a Query message. As described earlier, of all the
vehicles in the zone, the one whose twait expires first sends a Reply message
and gets elected as the new forwarder for the zone. The task of periodic
rebroadcast of Info messages is hence, handed over to the new forwarder.
Note that, a Query message is always broadcast by a forwarder which is
at either zone boundary and is heading out of the zone. Also, if a forwarder
fails to find a new forwarder at the time of exiting the zone, it repeats the
process to find a forwarder at a later instant, until which it remains the
forwarder (and retains its mode).
The frequency of periodic re-transmissions of Info messages f , is upper
bounded byvMAX
R. Wait time T is empirically selected such that T > 2 ×
(message propagation delay). Message propagation delay is a characteristic
of the medium and is typically in the range [10,15] ms.
3.3 Pseudocode of the Algorithm
Figure 3.1 shows the pseudocode for INITIALIZE, the method invoked at the
time of initialization of dissemination. INITIALIZE is also invoked when a
forwarder leaves a zone and a new forwarder is to be elected. In this method,
vehicle i broadcasts a Query message and waits for a Reply message for a
timeout period of T seconds. If a Reply message is not received, vehicle
18
Chapter 3. Zone Based Forwarding
Initialize(i)1 Mode← Relay2 Broadcast Query message3 Receive Reply message4 if Reply message not received within T sec
5 then Sleep(1
f)
6 Initialize(i)
Figure 3.1: Method invoked by vehicle i for initializing or relaying the dis-semination
i repeats the process of broadcasting a Query with frequency f . Hence,
vehicle i continues to be the forwarder until a new forwarder is found. Note
that, the Info message is contained in the Query message. Figure 3.2 shows
the pseudocode for RECEIVE-QUERY, the method invoked on receipt of a
Query message at a vehicle i. In this method, vehicle i computes its wait
time twait and waits for twait seconds before sending a Reply message, only if
no other Reply message is received within that period. Figure 3.3 shows the
pseudocode for the method invoked for periodic broadcast of Info messages
at a vehicle i. In this method, vehicle i first checks whether it has reached the
zone boundary, in which case it invokes INITIALIZE otherwise it re-invokes
BROADCAST-INFO after (1
f) seconds. Figure 3.4 shows the pseudocode
for RECEIVE-REPLY the method invoked on receiving a Reply message
at a vehicle i. In this method, vehicle i first checks if it lies on the same
side of the Reply message’s sender and Query message’s sender, in which
case it cancels its scheduled broadcast of a Reply message. Figure 3.5 shows
the method invoked on the receipt of an Info message. In this method, the
enclosed warning information is extracted and is delivered to the respective
application.
A vehicle i at position Pi generates the warning information and invokes
the INITIALIZE method. The mode of i changes to Relay and a Query mes-
sage is broadcast. Note that road segments between Pi and boundaries of i’s
transmission range R, form two zones. Any vehicle j at position Pj receiv-
19
Chapter 3. Zone Based Forwarding
Receive-Query(i, Query)1 j ←Query.Sender ID2 Pj ←Query.Sender Position
3 d̂j ←Query.Sender Direction
4 if (d̂j = d̂i)5 then6 s← |Pj − Pi|7 else8 s← R− |Pj − Pi|9 RECXY ← Pi
10 ts ←Time remaining(s)
11 twait = Max(N
ts, T )
12 Sleep(twait)13 if (ReceivedReply)14 then ReceivedReply ← False15 else Mode← Forward16 Broadcast Reply message
17 Sleep(1
f)
18 Broadcast-Info(i, Info)
Figure 3.2: Method invoked at vehicle i on receipt of a Query message
Broadcast-Info(i, Info)1 if (Now < Info.Time Limit and Pi < Info.Effect Area)2 then Broadcast Info message3 if (|Pi −RECXY | = s)4 then Initialize(i)
5 Sleep(1
f)
6 Broadcast-Info(i, Info)
Figure 3.3: Method invoked at vehicle i for broadcast of an Info message
20
Chapter 3. Zone Based Forwarding
Receive-Reply(i)1 j ←Reply.Sender ID2 k ←Query.Sender ID3 if (Pi ≤ PkandPj ≤ Pk)‖(Pi ≥ PkandPj ≥ Pk))4 then Mode← Receive5 ReceivedReply ← True
Figure 3.4: Method invoked at vehicle i on receipt of a Reply message
Receive-Info(i, Info)1 Content← Info.Warning information2 Deliver Content to relevant application
Figure 3.5: Method invoked at vehicle i on receipt of an Info message
ing the Query message invokes the RECEIVE-QUERY method. Distance
s is computed as the distance between Pj and the closest zone boundary
in the direction of motion d̂j of j. The position Pj is recorded in RECXY .
Time remaining() is used to compute the time (ts) vehicle j will take to
travel distance s. Wait time twait is calculated as the maximum ofN
tsand T ,
where N is a suitable normalizing factor.
After the expiration of the wait period twait vehicle j broadcasts a Re-
ply message. However, if a Reply message was received during the wait
period, method RECEIVE-REPLY would have been invoked. The flag
ReceivedReply would be set to true in this method, and on expiration of twait
vehicle j can use this flag to cancel its scheduled Reply message broadcast.
Otherwise, vehicle i changes mode to Forward and invokes BROADCAST-
INFO. In BROADCAST-INFO vehicle j broadcasts the periodic Info mes-
sage if the effect area and effect time are valid. After traveling a distance
s from position RECXY vehicle j is at the zone boundary and the method
INITIALIZE is invoked.
21
Chapter 3. Zone Based Forwarding
i
j
k
Transmission Range of i
l
Zone I Zone IIA B C
sk
sj
Figure 3.6: Sample scenario with vehicle i as initiating vehicle. Blocks depictthe vehicles and arrowheads denote their direction(North/South) of move-ment. vehicle i (solid block) initiates the dissemination.
3.4 An Example
We now present an example of the working of the proposed algorithm for the
scenario shown in Figure 3.6. Before start of the dissemination all vehicles i,
j, and k are in the Receive mode. Vehicle i on reaching point B on the road
detects a hazard, changes its mode to Relay and starts the warning message
dissemination. Vehicle i broadcasts a Query message, with sender ID as i,
sender direction as +X, sender location as B and other fields (as shown in
Table 1) accordingly. The region of i’s transmission range is divided into two
zones - I and II, as shown. Vehicles j, which belongs to zone I, receives the
Query and computes its distance sj as the distance to the zone boundary B.
22
Chapter 3. Zone Based Forwarding
Likewise, vehicle k computes its distance sk, as distance to the zone boundary
A. Similarly, respective ts (time to traverse distance s) values are computed,
and using this each vehicle computes its respective twait ∝ 1ts
, normalized
to lie within [0, T ]. Each vehicle waits for twait time before sending a reply
message, if it has not received one during the wait period. Note that a similar
process shall take place in Zone-II.
As sj << sk, we can assume that vehicle k waits less and sends a Reply
message, in which case it becomes the next forwarder and changes its mode
to Forward. Both vehicles j and i, on receiving the Reply message, check if
it was broadcast from the same zone to which they belong, and change back
to Receive mode. Now, as vehicle k had the least wait time, this implies
it is expected to take the most time to reach the zone boundary of Zone-I
towards which it is headed. Any broadcast of Info message by vehicle k is
guaranteed to reach any vehicle in Zone-I, and vehicle l entering the zone at
a later instant of time, shall be informed by such a periodic broadcast of Info
message by vehicle k.
Note that point A is the last point in Zone-I such that a broadcast by
vehicle k from that point is guaranteed to reach all nodes in Zone-I. Hence,
on reaching point A, vehicle k changes its mode to Relay and broadcasts a
Query message, to handover the task of being the zone-forwarder to another
vehicle, which is currently in zone-I, and is most likely to spend the most
amount of time in this zone.
3.5 Overcoming Network Congestion
In a traffic scenario with a large number of vehicles, which is a characteristic
of heavy or bumper-to-bumper traffic, the delivery of a Query message may
not be guaranteed. This can also be attributed to the fact that message
drop can occur if the periodic Info messages sent by the forwarder in the
adjoining zone collide with the Query message. If the Query message sent
by a forwarder vehicle which is leaving a zone fails to be delivered then the
23
Chapter 3. Zone Based Forwarding
warning information is no longer retained in that zone. In this section, we
propose an extension of the algorithm proposed in section 3.3 which counters
this problem caused by congestion in the wireless network, and can lead to
a misalignment in the protocol.
We propose to counter this problem by suggesting a variant of the algo-
rithm in which all the vehicles in the receiver mode anticipate the receipt of
a Query message in advance. Hence, even if the Query message fails to be
delivered to them, they can emulate the receipt of a Query and broadcast
a Reply message as in the previous case. The periodic Info message sent
to the receiver vehicles by the forwarder delivers the warning information,
but it also encloses the sender’s position. Based on this position the receiver
vehicles can estimate as to when the forwarder vehicle will reach the zone
boundary and broadcast the Query message. Hence, if the Query message
is not received within a timeout period from the expected instant, the re-
ceiver vehicles emulate the receipt of the Query message and broadcast Reply
messages so as to elect the new forwarder for the zone.
Receive-Info(i, Info)1 k ←Info.Sender ID2 Pk ←Info.Sender Position3 d̂k =Info.Sender Direction4 s′ =Compute Distance Remaining(Pk, d̂k, Pj, d̂j)5 if s′ ≤ s′TH
6 then Execute REINITIATE-ELECTION(i,k) after Time remaining(s′) + tTH sec7 Content← Info.Warning information8 Deliver Content to relevant application
Figure 3.7: Method invoked at vehicle i on receipt of an Info message
On receipt of the Info message the receiver vehicles obtain the sender’s
position and direction, and by using that along with the query sender’s po-
sition and direction they are able to obtain the distance s′ which the for-
warder will travel before reaching the zone boundary. Now, if s′ is less than
a threshold distance(s′TH), then the receiver vehicles schedule the execution of
RECEIVE-QUERY at an instant after tTH seconds, at which the forwarder
24
Chapter 3. Zone Based Forwarding
Reinitiate-election(i, k)1 if ReceivedReply == False2 then {}3 else RECEIVE-QUERY(k,Query)
Figure 3.8: Method invoked at vehicle i on non receipt of a Query messagedue to message drop
is supposed to reach the boundary. If the Query sent by the forwarder is
actually received then the scheduled execution of RECEIVE-QUERY does
not take place. Figures 3.5 and 3.5 show in detail the steps of this variant of
the algorithm. Apart from these, all other steps of the algorithm remain the
same shown earlier in Figures 3.1 - 3.4.
25
Chapter 4
Experimental Study
4.1 Simulation Setup
The network simulator employed in the experiments is NS-version 2.28 [15],
and VANET Mobisim-version 1.1 [16] is used as the traffic simulator to sim-
ulate the dynamic traffic scenario and mobility patterns of the vehicles. The
movement pattern of the vehicles, generated by the traffic simulator, in terms
of the current position, speed, and direction, is fed into the network simula-
tor to simulate the associated dynamic network topology over the VANET
at real time. This process is automated by means of a data-pipe between
the two simulators and a server-client, two-way communication takes place
between the traffic and network simulators which synchronize the flow of
data amongst themselves and run in parallel at run time. The detailed spec-
ification and implementation of this integrated simulator can be found in
[14].
The network simulator effectively simulates the characteristic mobility
pattern of moving traffic by periodically seeking position and velocity data
of each vehicle from the traffic simulator. The value of this periodic time
step, VMStime−interval, is kept appropriately small so as to effectively report
movement and changes in the movement pattern of a vehicle to the network
simulator without substantial delay, and is commensurate with the granular-
26
Chapter 4. Experimental Study
Channel : WirelessPropagation : Two Ray GroundMac : Mac/802 11Queue : Drop Tail / Priority QueueAntenna : Omni Antenna (Ht. 1.5m)Transmission Power : 0.2818dBm (250m)Frequency : 914 Hz
Table 4.1: Network parameters
ity of the latter. All vehicles in a scenario being studied are associated with
a single “mobile node group” and hence the update process of their position
and velocity for all vehicles is done concurrently with little time-lag. Hence,
the time granularity of the simulation is indeed VMStime−interval.
The network parameters are summarized in Table 4.1. The time interval
at which the updated position and velocities of the vehicles are read by the
network simulator, VMStime−interval, is set to 1 second, and all other algo-
rithms’ parameters are varied, measured and studied in its respect. Hence,
the absolute values of the parameters bear little significant and all analysis
and performance evaluation of the algorithm is conducted by varying param-
eters in multiples of VMStime−interval. This also contributes to the fact that
the results are independent of the absolute values of parameters. A change
in the value of VMStime−interval would just indicate a changed granularity
of the observed mobility pattern of the vehicles, as perceived by the net-
work simulator followed by an associated increase in the number of periodic
messages and other network-related events occurring in the VANET.
In what follows, we shall refer to the algorithm proposed in Chapter 3 as
the Zone Based Forwarding (ZBF) Algorithm.
27
Chapter 4. Experimental Study
4.2 Performance Evaluation
Irrespective of the information dissemination protocol in use, the dissemi-
nation process is initialized from one single vehicle which then coordinates
with other vehicles in its range to propagate the warning information. In
order to measure the performance of an information dissemination protocol,
we primarily employ the following two metrics:
1. Coverage: The total number of vehicles in the scenario which receive
the information after the dissemination is initiated. This is a measure
of the effectiveness of the information dissemination protocol in use.
2. Message broadcasts: The total number of message broadcasts that
took place to propagate the warning information. This is a measure
of the overhead incurred by the information dissemination protocol in
use.
However, some dissemination protocols deliver the information to the vehicles
when they are physically present within the perimeter of a demarcated road
stretch (zone/grid). In this case, the distance a vehicle travels before it is
delivered the warning information is employed as a metric to have a further
insight at the performance such a dissemination protocol provides. To be
employed in any VANET application any such information dissemination
protocol should deliver the warning information to an inbound vehicle as
soon as possible.
4.3 Experiments
In this section, we present results of simulations of different dissemination
protocols. We first present a performance comparison of the existing strate-
gies, followed by a comparison of the proposed ZBF algorithm with the rel-
evant existing strategies. Finally, we present results for performance evalua-
tion of the modified ZBF algorithm proposed in Section 3.5.
28
Chapter 4. Experimental Study
4.3.1 Comparison of Existing Approaches
We conduct the relative performance evaluation of the existing dissemina-
tion strategies, as described in Section 2.1, on a sample highway scenario of
length 2 Km. It consists of one single stretch of road with two lanes per
driving direction and a total number of 200 vehicles. The minimum and the
maximum speed limit for the vehicles are 0 m/sec and 25 m/sec (90 Km/hr)
respectively. The dissemination begins at a randomly chosen instant after
the traffic flow becomes stable. However, this randomly chosen instant is
the same for each dissemination strategy under observation. The periodic
beacons, if employed, are periodic with a period of 1 sec. Figure 4.1and 4.2
10
12
14
16
18
20
22
24
26
28
30
0 0.2 0.4 0.6 0.8 1
Num
ber
of
veh
icle
s in
form
ed (
Cover
age)
q
Torrent-MorenoCongestion Based
Congestion + Delay BasedProbabilistic (p : 1/|SLC|)
Figure 4.1: Comparative performance evaluation of existing strategies :Number of Vehicles Informed (Coverage)
show the coverage induced and message broadcasts incurred by the existing
strategies discussed in Section 2.1, namely, the contention based approach
by Torrent-Moreno and the congestion based approach. In addition, we also
present a variation of the congestion based approach, in which the broad-
29
Chapter 4. Experimental Study
0
2
4
6
8
10
12
14
16
18
20
22
24
26
28
30
0 0.2 0.4 0.6 0.8 1
Num
ber
of
mes
sages
bro
adca
st
q
Torrent-MorenoCongestion Based
Congestion + Delay BasedProbabilistic(p:1/|SLC|)
Figure 4.2: Comparative performance evaluation of existing strategies :Number of Messages Broadcasts
casting vehicles use a randomly chosen delay period from [0,200 ms] before
broadcasting the message. A vehicle which receives a message within this
delay period does not broadcast the message. Lastly, as discussed in Sec-
tion 2.1.3 we also compare the results of the probabilistic approach, where
the probability that a vehicle re-broadcasts a warning information (p) is in-
versely proportional to its Stabilized Local Congestion.
We observe that both the coverage and number of broadcasts change with
changing q, which is a parameter in the congestion based approach, but re-
main the same for other approaches. The congestion based approach informs
the maximum number of vehicles for q = 0, however the number of broad-
casts is also large. However, the variant of the congestion based approach
incurs far less overhead but reaches almost the same number of vehicles (at
q = 1). Also, the probability based approach has the least coverage. Lastly,
the contention based approach by Torrent-Moreno [2], incurs less number of
message broadcasts but at the same time reaches a greater coverage than any
30
Chapter 4. Experimental Study
0
20
40
60
80
100
120
140
160
180
200
170 175 180 185 190 195 200
Num
ber
of
veh
icle
s/m
essa
ge
bro
adca
sts
Timescale(in sec)
Number of Vehicles Informed (Coverage)Number of Message Broadcasts
Figure 4.3: Performance of the strategy proposed by Fathy and Khakbaz in[3]. Information spreads to all vehicles as time progresses, starting at t=170
other strategy. Hence, we observe that both the congestion based approach
(with random delay) and the contention based approach perform equally well,
but the performance of the latter is marginally better in some cases. In this
light, the strategy proposed by Fathy et al. in [3] which improves upon the
Torrent-Moreno approach attains greater importance.
Figure 4.3 shows the performance of the strategy proposed by Fathy et al.
in [3]. Since, this approach carries through the dissemination over a period of
time, we see that the number of vehicles receiving the information increases
substantially as time progresses. The number of vehicles informed at the
start of dissemination are better than the Torrent-Moreno approach owing
to the betweenness conditions introduced in this approach. Also, since this
approach is able to deliver the warning information across network partitions
it delivers the warning information to all vehicles in the scenario within a
period of 30sec. The delivery of the warning information to all the vehicles
31
Chapter 4. Experimental Study
in the scenario is not observed in any of the other existing strategies.
Since, the Torrent-Moreno approach and the approach by Fathy et al. [3]
outperform all existing dissemination strategies, we use them as a reference
for further comparisons in the next section.
4.3.2 Performance Evaluation of ZBF
We now compare the performance of the ZBF algorithm proposed in Chapter
3, and the Stored Geocast algorithm as discussed in Section 2.2. We use
a sample highway scenario of length 10 Km for this evaluation, with two
lanes per driving direction. The total number of vehicles in the scenario
are 500, and the minimum and the maximum speed of the vehicles is 0
and 30 m/sec (108 Km/hr) respectively. We observe the performance of
the two approaches by varying the frequency (f) of the periodic Info and
GATE message respectively. Recall that in the Stored Geocast algorithm
the elected gateway vehicle periodically broadcasts its existence by sending
a GATE(g,loc) message, where g is its grid coordinate and loc is its current
location. The values used for normalizing factor (N) and maximum waiting
time (T ) are 0.1 and 1 sec respectively.
Figure 4.4 shows the coverage achieved by both the algorithms varied with the
frequency (f) of the periodic Info and GATE messages. In case of the Stored
Geocast algorithm, the warning information to be spread is piggybacked upon
the GATE messages broadcast periodically by the gateway vehicle, whereas
in case of the ZBF algorithm, the warning information is contained in all of
the Info, Query, and Reply messages. As discussed in Section 3.1, frequency
(f) of the periodic broadcasts is upper bounded byvmax
Rand is equal to 0.12
(vmax=30 m/sec, R=250 m). The results are sampled at 5 random instances
for a period of 5 minutes for each of the three cases of one, two and three
zones or grids existing in the scenario. Further, the grid length and the zone
length respectively in the two approaches are equal to the transmission range
(R=250 m) of the vehicles. Note that, the results shown in Figure 4.4 are
averaged across all the random instances and all the three cases of one, two
32
Chapter 4. Experimental Study
0
50
100
150
200
0 0.015 0.03 0.045 0.06 0.075 0.09 0.105 0.12
Num
ber
of
veh
icle
s
Frequency of periodic Info/GATE messages
Number of vehicles moving in/out of a zone
ZBF Algorithm
Contention based
Stored Geocast
Figure 4.4: Total number of vehicles informed v/s Frequency (f) of periodicbroadcast.
and three zones/grids.
We observe that since the information is delivered to only those vehicles
which are present in the zone at any moment of time within the duration of
the experiment, the maximum coverage is bounded by the total number of
vehicles moving in/out of a zone within that period. This also includes the
vehicles which are initially present within a zone at the time of the start of
the dissemination.
The number of vehicles to which the Stored Geocast algorithm delivers
the warning information increases as f increases. However it is surpassed by
the ZBF algorithm when f attains its maximum value of 0.12. At that value
of f the ZBF algorithm ensures that the warning information is delivered
to all the vehicles which are present in the zone at any instant after the
start of dissemination. Note that, the Torrent-Moreno approach delivers the
warning information to a far lesser number of vehicles as it does not retain
the information within the area of interest.
33
Chapter 4. Experimental Study
0
50
100
150
200
0 0.015 0.03 0.045 0.06 0.075 0.09 0.105 0.12
Num
ber
of
mes
sages
bro
adca
st
Frequency of periodic Info/GATE messages
ZBF Algorithm
Contention based
Stored Geocast
Figure 4.5: Total number of messages broadcast v/s Frequency (f) of periodicbroadcast.
Likewise, Figure 4.5 shows the total number of message broadcasts by
each of the strategies. As f increases the GATE and Info messages re-
spectively are broadcast more frequently, and hence the number of message
broadcasts increases. We observe that the number of message broadcasts by
the Stored Geocast algorithm far exceeds those by the ZBF algorithm. The
Torrent-Moreno approach incurs less message broadcasts, but it also delivers
the message to less number of vehicles, as was observed in Figure 4.4.
We analyze the different type of message broadcast used by the two strate-
gies in Figure 4.6. We observe that the increased message broadcast in the
Stored Geocast algorithm are due to the large number of BID message broad-
casts employed by it. They are separate messages used by the algorithm for
electing a new leader whenever the previous one exits the grid. Each vehicle
in the grid broadcasts one such message on receiving a RETIRE message
from the outbound leader vehicle. On the other hand, in case of the ZBF
algorithm only one Reply message, in response to a Query message suffices
34
Chapter 4. Experimental Study
0
20
40
60
80
100
120
140
0 0.015 0.03 0.045 0.06 0.075 0.09 0.105 0.12
Nu
mber
of
mes
sages
bro
adca
st
Frequency of period Info/GATE messages
Proposed Algorithm: Number of Query Messages
Proposed Algorithm: Number of Reply Messages
Proposed Algorithm: Number of Info Messages
GeoGRID Algorithm : Number of RETIRE Messages
GeoGRID Algorithm : Number of BID Messages
GeoGRID Algorithm : Number of GATE Messages
Figure 4.6: Distribution of different types of messages broadcast v/s Fre-quency(f) of periodic broadcast.
to elect the new forwarder for the zone.
Figure 4.7 shows the average distance any vehicle which is inbound to
a zone travels before it is delivered the warning information. In case of
the Stored Geocast algorithm, this is the distance traveled before receiving
the first GATE message broadcast from the leader vehicle. For both these
approaches it is seen that as the frequency (f) of broadcast of Info and
GATE messages respectively is increased, the distance decreases. This is
because with an increased frequency of broadcast, the time period between
successive re-broadcasts decreases and an inbound vehicle is more likely to
receive the information sooner. At lower frequency (f) values the Stored
Geocast algorithm delivers the information within 80 m of the vehicle entering
35
Chapter 4. Experimental Study
0
10
20
30
40
50
60
70
80
0 0.015 0.03 0.045 0.06 0.075 0.09 0.105 0.12
Dis
tance
tra
vel
led i
nsi
de
a zo
ne
Frequency (f) of periodic Info/GATE messages
ZBF Algorithm
Stored Geocast Algorithm
Figure 4.7: Distance inside a zone which a vehicle traversed before beinginformed v/s Frequency(f) of periodic broadcast.
the grid which is less than the distance in case of the ZBF algorithm. But
as f is increased to the ideal value of (vmax
R) the distance in case of the
ZBF algorithm is observed to be lesser than in case of Stored Geocast by
10 m. Thus, in terms of the distance traveled by a vehicle inside a zone
before getting the warning information for the first time, Stored Geocast
outperforms the proposed strategy at lower frequency (f) values but at higher
values the ZBF algorithm performs better.
We also present a comparison of the performance of the two strategies
under consideration by using the total number of message broadcasts per
informed vehicle as a metric. Figure 4.8 shows the number of messages
broadcast per informed vehicle in case of the Stored Geocast algorithm, the
Torrent-Moreno algorithm and the ZBF algorithm. We observe that the ZBF
algorithm uses far less number of messages per informed vehicle as compared
to the Stored Geocast algorithm. Although, the Torrent-Moreno algorithm
uses lesser number of messages per informed vehicle, but as is evident from
36
Chapter 4. Experimental Study
0
0.2
0.4
0.6
0.8
1
1.2
0 0.015 0.03 0.045 0.06 0.075 0.09 0.105 0.12
Mes
sage
bro
adca
sts
per
info
rmed
veh
icle
Frequency of periodic Info/GATE messages
ZBF Algorithm
Torrent-Moreno
Stored Geocast
Figure 4.8: Total number of messages broadcast per informed vehicles v/sFrequency(f) of periodic broadcast.
Figure 4.4, it delivers the information to lesser number of vehicles.
Lastly, we draw a comparison between the ZBF algorithm and the algo-
rithm proposed by Fathy et al. [3] which improves upon the contention based
approach proposed by Torrent-Moreno [2]. Figure 4.9 shows the performance
of the algorithm by Fathy et al. [3] on the same sample highway scenario,
as was used for the other approaches. Since this approach essentially uses
the contention based approach for propagation, we observe that the vehicle
coverage and the number of message broadcasts is similar at the start of dis-
semination (denoted as 0 on the Timeline axis). However, as time progresses
this approach bridges across network fragments and delivers the information
to more number of vehicles and finally to approximately all the vehicles in
the scenario.
Note that this approach delivers the warning information to as many
vehicles as possible and does not take into account whether or not the in-
formation content is relevant for the receiving vehicles. However, if we do
37
Chapter 4. Experimental Study
0
100
200
300
400
500
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80
Num
ber
of
veh
icle
s/m
essa
ge
bro
adca
sts
Timeline(in seconds, starting from the time of dissemination)
Number of Message BroadcastsNumber of Vehicles Informed
Figure 4.9: Result of the Fathy algorithm [3] on a sample highway scenario.
assume that the information is relevant all the receipts we observe that this
approach takes 20−25 seconds to deliver the information to the same number
of vehicles as in the case of the ZBF algorithm. However, the ZBF algorithm
delivers the warning information, within a distance of 40−80m from the zone
boundary, to only the vehicles for which it is relevant. At an average speed
ofvMAX
2= 15m/sec (vMAX = 30m/sec) this results in a delay of 2.6 − 5.3
seconds. Thus, we observe that the ZBF algorithm delivers the information
to only the vehicles for which it is relevant and in the worst case incurs a
delay of 5.6 seconds for each receiving vehicle, which is lesser than the worst
case delay of 25 seconds incurred by the Fathy et al. algorithm.
4.3.3 Performance Evaluation of ZBF Extension
In this section we evaluate the performance of the extension of the ZBF
algorithm presented in Section 3.5 for overcoming network congestion, if
present in the scenario. In order to observe the effectiveness of this extension
38
Chapter 4. Experimental Study
0
100
200
300
400
500
600
700
0 0.015 0.03 0.045 0.06 0.075 0.09 0.105 0.12
Num
ber
of
veh
icle
s
Frequency of periodic Info messages
Number of vehicles moving in/out of a zone
ZBF Algorithm
Extended ZBF Algorithm
Figure 4.10: Number of vehicles reached v/s Frequency(f) of periodic broad-cast.
in overcoming congestion, we simulate the protocol over the same highway
scenario but with the number of vehicles increased to 1000. This induces
heavy bumper-to-bumper vehicular traffic and allows us to simulate the drop
of the Query message which is responsible for the handover of the forwarding
task to a new forwarder vehicle in a zone. The values of tTH and s′TH used
for the simulations are 1 sec and 15 m respectively.
Figure 4.10 shows the comparison of the vehicle coverage achieved by the
ZBF algorithm and its proposed extension. We see that in heavy traffic, as
the network becomes more and more congested some Query messages are
dropped however, the proposed extended version overcomes this and still re-
tains the information in a zone by electing a new forwarder. Consequently,
we observe that the proposed extension achieves a greater vehicle coverage
than the standard ZBF algorithm. However, because the receipt of the Query
is emulated by the receiver vehicles in this extension and a subsequent broad-
cast of Reply message occurs, we observe an increased number of message
39
Chapter 4. Experimental Study
0
20
40
60
80
100
120
0 0.015 0.03 0.045 0.06 0.075 0.09 0.105 0.12
Num
ber
of
mes
sage
bro
adca
sts
Frequency of periodic Info messages
ZBF Algorithm
Extended ZBF Algorithm
Figure 4.11: Total number of messages broadcast v/s Frequency(f) of periodicbroadcast.
broadcasts due to these Reply message broadcasts. In case of a Query mes-
sage being dropped in the standard ZBF algorithm, the receiver vehicles
would not have received the Query message and hence broadcast of Reply
message in response to it would not have had occurred. This is evident from
the result shown in Figure 4.11 which compares the total number of message
broadcasts in the two cases.
Hence, we see that in case of heavy or bumper-to-bumper traffic being
present in the scenario, than the proposed ZBF algorithm extension over-
comes Query message drops and produces a better coverage. But the total
number of message broadcasts incurred are also more as compared to the
standard ZBF algorithm of Section 3.3.
40
Chapter 5
Conclusion
This thesis presented an efficient information dissemination strategy which
incurs small overhead and is capable of retaining the information within
the effect area of the information for the entire duration of its effect time.
The preceding chapters have shown that most of the existing dissemination
strategies do not retain the information within its target area. However,
there exists an algorithm which provides retention of information in the effect
area but incurs a large overhead, making it unsuitable for deployment over
VANETs. In this light, the contribution of this thesis has been twofold:
1. A detailed comparison of the existing information dissemination strate-
gies in VANETs is presented and certain essential characteristics that
an information dissemination strategy must possess are identified. It
is shown that all these characteristics are not present in most of the
existing dissemination protocols.
2. An information dissemination strategy which incurs little overhead is
presented which is capable of retaining the information within the effect
area of the warning information for the entire duration of its effect time.
Further, we demonstrate that how the proposed strategy outperforms
the existing information dissemination protocols in VANETs.
41
Chapter 5. Conclusion
5.1 Future Work
Both safety and convenience applications in VANETs might benefit from
further research on warning information dissemination. Some possible future
work on warning information dissemination in VANETs, with regards to the
ZBF algorithm, can be as follows:
1. Although, the frequency (f) of the periodic broadcasts is empirically
chosen, the exact method to compute it dynamically based on the cur-
rent traffic scenario around a vehicle has been left un-specified. Further
work, may investigate its correlation with other network and mobility
characteristics of the vehicle as well as develop an algorithm to compute
it on the fly maybe devised.
2. The ZBF algorithm aims to build up consecutive zones covering the
entire area of interest of the warning information. However, a zone
with a periodic broadcast frequency f equal to (vMAX
R) guarantees
that any vehicle passing the zone will be delivered the information.
Hence, two such zones which bound the area of interest will in fact
ensure that all inbound traffic is delivered the warning information. A
mechanism which allows these two bounding zones to be set up needs
to be investigated.
3. The zone length, which is assumed here to be equal to the transmission
range (R), can also be setup as a value more than R. We suggest
that such an assumption would also lead to an aspect of probabilistic
delivery of the warning information to the vehicles visiting such a zone.
Note that, in this case the number of forwarders and hence the number
of broadcasts can be further reduced.
4. The periodic Info messages which are of greatest significance for spread-
ing the warning information can also be used as a method of electing a
new forwarder. Any vehicle receiving an Info message which senses its
time remaining to be spent in the zone as more than that of the for-
warder can assume the forwarding role by suitable informing the latter.
42
Chapter 5. Conclusion
However, specific semantics of the changing of the forwarder need to be
determined. Also, the fact that excessive handovers of the forwarding
role might lead to an increase the overhead incurred by the protocol
needs to be investigated.
43
Bibliography
[1] H. Sabbineni and K. Chakrabarty, “Location-aided Flooding: An Energy
Effecient Data Dissemination Protocol for Wireless Sensor Networks”,
IEEE Transactions on Computers, vol. 54, no. 1, January 2005 .
[2] Marc Torrent-Moreno, “Inter-Vehicle Communications: Assessing Infor-
mation Dissemination under Safety Constraints”, 4th Annual IEEE/IFIP
Conference on Wireless On Demand Network Systems and Services
(WONS), Obergurgl, Austria, January 2007.
[3] M. Fathy and S. Khakbaz,“A Reliable Method for Disseminating Safety
Information in Vehicular Ad hoc Networks Considering Fragmentation
problem”, 4th International Conference on Wireless and Mobile Commu-
nications, 2008.
[4] F. Bai, H. Krishnan, V. Sadekar, G. Holland, and T. ElBatt, “Towards
Characterizing and Classifying Communication-based Automotive Appli-
cations from a Wireless Networking Perspective”, 1st IEEE Workshop on
Automotive Networking and Applications, AutoNet, 2006.
[5] Tamer ElBatt, Siddhartha Goel, Gavin Holland, Hariharan Krishnan
and Jayendra Parikh, “Cooperative Collision Warning Using Dedicated
Short Range Wireless Communications”, The Third ACM International
Worskshop on Vehicular Ad Hoc Networks (VANET 2006), Los Angeles,
California, USA, September 2006.
44
Bibliography
[6] Y. Toor, P. Muhlethaler and A. Laouiti, “Vehicular Ad Hoc Networks:
Applications and Related Technical Issues”, Communications Surveys
and Tutorials, IEEE, vol. 10, no. 3, pp. 74-88, September 2008.
[7] C. L. Robinson, L. Caminti, D. Caveney and K. Laberteaux, “Effecient
Coordination and Transmission of Data for Cooperative Vehicular Safety
Applications”, The Third ACM International Worskshop on Vehicular Ad
Hoc Networks (VANET 2006), Los Angeles, California, USA, September
2006.
[8] T. Kosch, “Local Danger Warning Based on Vehicular Ad-hoc Networks:
Prototype and Simulation”, 1st International Workshop on Intelligent
Transportation, 2004.
[9] Rayman Preet and Arobinda Gupta, “Traffic Congestion Estimation in
VANETs and Its Application to Infomation Dissemination”, 12th Interna-
tional Conference on Distributed Computing and Networking, Bangalore,
India, January 2011.
[10] C. Maihofer, W. Franz and R Eberhardt, “Stored Geocast”, Kommu-
nikation in Verteilten Systemen (KiVS), Leipzig, Germany, 2003.
[11] W.H. Liao, Y.C. Tseng, K.L. Lo and J.P. Sheu, “GeoGRID: A geocasting
protocol for mobile ad hoc networks based on GRID”, Journal of Internet
Technology, vol. 1, no. 2, pp. 23-32, December 2000.
[12] Yu Qiangyuan and Heijenk Geert, “Abiding Geocast for Warning Mes-
sage Dissemination in Vehicular Ad Hoc Networks”, 2nd ACM Interna-
tional Workshop on Vehicular Ad Hoc Networks, ICMCN, 2005.
[13] Wireless Access in Vehicular Environment (WAVE) in Standard 802.11
Information Technology Telecommunications and Information Exchange
Between Systems, Local and Metropolitan Area Networks, Specific Re-
quirements, Part 11: Wireless LAN Medium Access Control(MAC) and
Physical Layer(PHY) Specifications, IEEE 802.11p/D1.0, Feb. 2006
45
Bibliography
[14] Vinu Rajshekhar, “An Integrated Traffic and Network Simulator for
Realistic VANET Simulation”, Bachelors Thesis, Indian Insitute of Tech-
nology, Kharagpur, India, May 2009.
[15] “Network Simulator ns-2”, http://www.isi.edu/nsnam/ns/.
[16] “VanetMobisim”, http://vanet.eurecom.fr/.
[17] “Car2Car Communication Consortium,” http://www.cat-to-car.org/.
[18] Z.D. Chen, H.T. Kung and D. Vlah, “Ad hoc relay wireless networks
over moving vehicles on highways”, 2nd ACM International Symposium
on Mobile Ad Hoc Networking & Computing, Long Beach, CA, USA
2011.
[19] T. Kosch, Christian Schwingenschlogl and Ai Li, “Information dis-
semination in multihop inter-vehicle networks”, IEEE 5th International
Conference on Intelligent Transportation Systems, Singapore, September
2002.
[20] M. Rudack, M. Meincke, K. Jobmann and M. Lott,“On traffic dynamical
aspects of inter vehicle communications (IVC)”, IEEE 58th Vehicular
Technology Conference, March 2003.
[21] M. Rudack, M. Meincke and M. Lott, “On the Dynamics of Ad Hoc
Networks for Inter Vehicles Communications (IVC)”, International Con-
ference on Wireless Networks, Las Vegas, USA, 2002.
[22] H. Fubler, M. Mauve, H. Hartenstein, M. Kasemann and D. Vollmer,
“A comparison of routing strategies in vehicular ad-hoc networks”, ACM
International Conference on Mobile Computing and Networking (MOBI-
COM), Atlanta, USA, September 2002.
[23] T. Kosch, “Local Danger Warning Based on Vehicular Ad-hoc Networks:
Prototype and Simulation”, 1st International Workshop on Intelligent
Transportation, Hamburg, Germany, 2004.
46