performance analysis of vehicular ad hoc (2)
TRANSCRIPT
8/7/2019 Performance Analysis of Vehicular Ad hoc (2)
http://slidepdf.com/reader/full/performance-analysis-of-vehicular-ad-hoc-2 1/6
1
Abstract — In the past, many researchers depend
highly on various network simulation tools such as Ns2,
OMNET++, Jist, Glomosim, OPNET, and JSim [1] to
analyze the performance of various routing algorithms.
However many of these simulation tools have limited
mobility models such as the Random Waypoint and fixed
direction movement. Recent ever-increasing interest in
vehicular ad hoc networks communication (VAC)
necessitates vehicular movement performance study via
realistic mobility pattern simulation. Unlike many other
works, we present a study on the performance of various
routing algorithms using Ns2 except that the mobility
pattern is generated using a microscopic vehicular
movement simulation tool called SUMO. We then
investigate and attempt to quantify the effects of the
vehicular traffic parameters such as the average speed,
vehicle density and road topology on the overall VANET
performance.
Index Terms —vehicular ad hoc networks (VANET),
intervehicle communication (IVC), routing.
I. I NTRODUCTION
ver increasing recent interest growth in VAC were initially
aimed to provide comforts and safety for passengers on
the road. Indeed, VAC has benefited us in many ways such as
higher reliability telecommunication service at a more
affordable cost. Many people aren’t aware of this special
variation of network service is free compared to the costly 3G
and WiMAX network services. Hence, VAC able to offers a
very attractive low-cost network services. This criteria
benefited many parties especially telecommunication service
provider in extending their coverage area without substantial
investment of wireless infrastructure. Unfortunately, the
performance of VANET highly depends on its routing
protocols. Many existing routing performance deteriorates as it
runs within the versatile vehicular environment. VAC demands
for a highly self-organized enabling extension of coverage area
via adaptive routing mechanism such as path extension. Such
F. A. Author is with the National Institute of Standards and Technology,
Boulder, CO 80305 USA (corresponding author to provide phone: 303-555-
5555; fax: 303-555-5555; e-mail: author@ boulder.nist.gov).
requirement necessitates a study of routing performance and its
behavior across various mobility patterns in vehicular
environment. In this article, we will study of routing protocol
performance under realistic vehicular mobility environment.
We model the vehicle movement using Sub-Urban Mobility
Simulator (SUMO) [2], a microscopic vehicular simulator.
This movement model is then ported into network simulator
ns2 and verified using network animator (NAM) [5]. We then
attempt to quantify the effect of the vehicle movement
parameters (speed, density, and topology). Using these
parameters, we conduct an extensive simulation and examinehow they affect the performance of VANETs.
Arguably, simulation results are not a total representation of
actual vehicular network performance, but it is a key to study
and verify the vehicular routing algorithm performances and
its behavior. Such study is essential to aid the design of
VANET routing algorithm.
II. SIMULATION NETWORK MODEL
We modeled a few typical vehicular environments namely
high-speed highway, variable vehicle speed (VSS) region, and
bus in a city environment. Then we investigate the routing
protocol performances using the realistic mobility patterns.
Fig. 1. Mobility Pattern for Highway
Performance Analysis of Vehicular Ad hoc
Networks with Realistic Mobility Pattern
Wai Foo Chan, Moh Lim Sim, and Sze Wei LeeThird C. Author, M er , I EEE
E
318
Proceedings of the 2007 IEEE International Conference on Telecommunications andMalaysia International Conference on Communications, 14-17 May 2007, Penang, Malaysia
1-4244-1094-0/07/$25.00 ©2007 IEEE.
Wai Foo Chan, Moh Lim Sim, and Sze Wei Lee.Faculty of Engineering, Multimedia University, 63100 Cyberjaya, Selangor, MALAYSIA
8/7/2019 Performance Analysis of Vehicular Ad hoc (2)
http://slidepdf.com/reader/full/performance-analysis-of-vehicular-ad-hoc-2 2/6
2
a. High-speed Highway Environment
A typical high-speed highway movement with dual lane
opposite directions across a stretch of 4 kilometers distance is
considered within the simulation. This highway vehicle model
consists of 4000 nodes. Each 2000 nodes pairs are moving on
opposite direction. Each vehicle is assigned with varying speed
with various maximum speeds ranging from 10km/h to
120km/h. This vehicular environment model is simulated on
SUMO simulation tools. The mobility model is ported into ns2simulator. Figure 1 illustrates the vehicular model portrays
within the microscopic mobility simulator. The ported mobility
pattern is tested and verified with NAM that comes with ns2
simulator. In this simulation we attempts to quantify typical
movement parameters as such vehicle speed and examines the
effects on ad hoc network performance.
b. Variable speed scenario (VSS) Environment
VSS is a typical vehicular environment where all vehicles
slows down at certain point of the road due to presence of
external factor along the road such as road constructions,
accident, and submergence of lanes, pedestrian crossing and
school children zone. Such condition reduces distance between
vehicles and increases vehicle network density. To model VSS
scenario, we consider a 4km stretch road with a slight curve
corner slowing down the vehicle and caused accident. 500
vehicle nodes are considered for 534 seconds in this
simulation using SUMO. The accident car is labeled in red
color as illustrated in Figure 2b. At time t0, the vehicle queue
in lane 1(L1), before the curve corner were relatively short. Atlane2 (L2), The vehicles were separated relatively sparse as
illustrated in Figure 2a. Here the communication route path
can be formed from node within L1 and node within L2.
At t1, the accident occurred at the curve corner. This
accidents block vehicle from going through the road and
lengthen the queue in L1. The increases of the queue length
introduce higher density network at L1. Meanwhile, at lane2
(L2) the vehicles were sparsely as illustrated in Figure 2b.
Hence, the communication between vehicles from L1 to L2
becomes impossible.
At t3, the accident has been cleared. Such scenario causes
the network topology changed. L2 is accommodated withvehicles nodes enabling the communication between vehicles
from L1 to L2 again.
c. Public Bus within City Environment
We use grid to represent a city environment. Figure 3
depicts the modeled city. Figure 3 is just the overview of the
modeled grid city. The zoom in view of the city model Figure
4 illustrates the details of the modeled city. In the city, each
empty cell is 200 meter-square spaces represents building
spaces between the roads. Between each junction, we have
modeled traffic lights within the road. Within the model, bus
stops are placed on 3 spots within the city. Each bus stop is
configured with varying bus line. In this model we imitate
realistic bus stop for different bus line service within the city.
Presence of traffic lights and bus stops enables us emulates
natural vehicular environment as such “stop and go”
condition. Each node is labeled with different color to
indicate with different route path as shown in the Figure 4.
Each node is being assigned vehicle type, ranging from bus
with 5 meters length (long yellow vehicle) and car (triangle
green, pink or yellow node) with an average length of 2
meters as illustrated in Figure 4. For bus node, the bus
services available for every 1 minutes interval. A total of 90
buses are simulated for about 1200 seconds in this modelusing SUMO. Using this mobility model, we manage to
simulate and study the network performance under realistic
mobility traces.
Using this mobility pattern, we examine the network
performance of a heterogeneous wireless cum wired network
in urban city environment. We consider simple wireless
application such as Internet that requires city bus to
communicate with base-station on peak hour and non-peak
hours. On peak hours, the frequency of bus trip is higher Fig. 2a-c. Accident Environment
2b. During Accident, t1
2c. Right After Accident Clearance, t2
L1
L2
L2
L1
L1
L2
2a. Before Accident, t0
319
8/7/2019 Performance Analysis of Vehicular Ad hoc (2)
http://slidepdf.com/reader/full/performance-analysis-of-vehicular-ad-hoc-2 3/6
3
within the urban city. The packet throughput from bus to the
base-station is then calculated to measure the network
performance. Based on this statistic, a study on network
performance and behavior is then conducted.
Fig. 3. City Model as Grid
Fig. 4. Public Bus within City
Simulation Parameters
a. Highway
The realistic mobility model is then imported into ns2
network simulator mobility model. Assuming not all vehicle in
the highway were equipped with WiFi equipment, we
simulated model consists of 1 percent random sampling of the
generated 4000 nodes within the highway. Each of therandomly sampled vehicles is assumed to be equipped with
IEEE 802.11b network adapter with a transmission range/
radius of 250 meters moving within the highway. In the
simulation we study and compare the performance of Dynamic
Source Routing Algorithm (DSR) [3], Ad hoc On-Demand
Distance Vector (AODV) [6] and Optimized Link State
Routing Algorithm (OLSR) [4]. OLSR is classified as a
proactive routing algorithm while DSR can be classified as
reactive routing algorithm. We sampled 50% of the overall
vehicles in the model as communicating pairs for 200 seconds.
For each simulation run, the source node and destination were
randomly picked to imitating the typical random connection
establishment within highway. An assumption of not all
vehicles were involved in the communications within the
highway is made here. Network performance is analyzed in
terms of packet throughput, delay and jitter. For varying
vehicle speed setting, generally the simulation was repeated
for many times.
TransmissionRange (m)
250 m
No of nodes 40
Area (mxm) 12100x100Average Speed(km/h)
10,30,50,70,100,120
Mobility Pattern Highway
Simulation time(s)
920-1100
Routing Algorithm DSR, OLSR, AODV
MAC 802.11 b
Table 1. Simulation Parameter for Highway
b. Variable Speed Scenario
This network simulation model consists of 60 vehicle nodes.
Each vehicle is assumed to be equipped with IEEE 802.11bnetwork adapter with a transmission range/ radius of 250
meters moving across different section of roads with different
speed. We modeled in such way where each vehicle attempts
to communicate with others vehicle imitating the traffic
scenario of information sharing between vehicles from one end
to another end of the road to accessing Internet services/base-
station at the other end of road. Upon accident took place,
vehicles slow down their car and stops at lane L1 separating
the vehicles in lane L2. At this point of time, there would be a
sudden increased of access of network service for the
Vehicle
Bus Stop
Bus
Traffic Lights
ZOOM VIEW
320
8/7/2019 Performance Analysis of Vehicular Ad hoc (2)
http://slidepdf.com/reader/full/performance-analysis-of-vehicular-ad-hoc-2 4/6
4
passenger to seek for infotainment info while waiting for the
accident to be cleared. Many calls or message to be triggered
to inform the others party on the delay of trip during this time.
Having this scenario, vehicles are required to be connected to
the base station at the other end of the road, however the base
station along the road is considerably costly and installed
distance apart. Hence vehicles required behaving in a self-
organized manner to form ad-hoc network to extend the
network between vehicles from right end to left end of lane L1.We assume that the base station to be at the left end of lane
L1. Network performance is analyzed in terms of packet
throughput from one end of the road to another end. The
behavior of network performance is then investigated.
TransmissionRange (m)
250 m
No of nodes 40
Area (mxm) 4000x100
Average Speed(km/h) 30
Mobility Pattern VSS
Simulation time(s)
534
Routing Algorithm DSR
MAC 802.11 b
Table 2. Simulation Parameter for VSS
c. Urban City
This network simulation model consists of 95 busses node
and a base station node. Each bus is assumed to be equipped
with IEEE 802.11b network adapter with a transmission range/radius of 250 meters travels at an average speed of 30 km/h
within an area 900 square-meter of urban city.
TransmissionRange (m)
250 m
No of nodes 95, 1 base station
Area (mxm) 900x900
Average Speed(km/h)
30
Mobility Pattern Urban City
Simulation time(s)
1200
Routing Algorithm AODV, OLSR
MAC 802.11 b
Table 3. Simulation Parameter for Urban City
Within this simulation, we modeled in a way where each bus
attempts to communicate with others bus and an access-point
(modeled as base-station) imitating real-life traffic scenario of
information sharing between busses and user accessing
Internet services. This variant of application is inherently
observable from our daily life such as passenger accessing
Internet from a bus. For varying vehicle environment in urban
city, generally the simulation was repeated for 20 times using
both peak hour and non-peak hour mobility model. Network
performance is analyzed in terms of packet throughput from
the bus to the base station. The behavior of network
performance is then investigated. In the simulation we also
study and compare the performance of both proactive
Optimized Link State Routing Algorithm (OLSR) [4] andreactive routing algorithm, Ad hoc On-Demand Distance
Vector (AODV) [6].
III. SIMULATION RESULT
We executed 50 – 60 runs of the network simulator
for each varying vehicle speed in each time simulation. Each
simulation last about 900-1200 seconds. The ad hoc network
in the simulation consists of 40 mobile nodes with movement
pattern that mimic a vehicle movement pattern within highway.
What we present here is the average packet throughput, delay
and jitter over 50 – 60 runs of the simulator for each of the
different speeds; standard deviation for all cases is within the
range of 5% - 20% of the average value. Figure 5, 6, 7 shows
the average packet throughput, delay and jitter of each node
for both OLSR and DSR routing algorithms.
Throughput vs Speed
0
5
10
15
20
25
0 50 100 150
Speed (km/h)
T o t a l T h r o u g
h p u t
( k b p s ) DSR
AODV
OLSR
Fig. 5. Packet Throughput vs. Vehicle Speed
We observed that averagely DSR routing algorithm
conceives higher throughput than any other compared routing
algorithm. All the routing algorithm’s performance (DSR,ODV and OLSR) deteriorates over node speed as illustrated in
Figure 5. Frequent and high network fragmentation causes this
trend of packet throughput ratio declination. Packet end-to-end
delay and jitter on the other hand increases over node speed as
illustrated in Figure 6 and 7.
321
8/7/2019 Performance Analysis of Vehicular Ad hoc (2)
http://slidepdf.com/reader/full/performance-analysis-of-vehicular-ad-hoc-2 5/6
5
Packet Delay vs Speed
-1000
0
10002000
3000
4000
5000
6000
7000
0 50 100 150
Speed (km/h)
D e l a y ( m s )
Delay DSR
Delay AODV
Delay OLSR
Fig. 6. Packet Delay vs. Vehicle Speed
Packet Jitter
-500
0
500
1000
1500
2000
0 50 100 150
Speed (km/h)
J i t t e r ( m s )
Jitter DSR
Jitter AODV
Jitter OLSR
Fig. 7. Packet Jitter vs. Vehicle Speed
To study the effect of vehicle density, we measure the
performance of packet throughput when number of vehicle
increases in the queue of lane L1 after the accident as
illustrated in Figure 2b. The packet delivery ratio performance
due to the effects of network density is presented at Figure 8
and 9.
Total Throughput
010
20
30
40
0 10 20 30
Number of Vehicle per Lane
Communicating
T o
t a l P a c k e t
T h r o u g h p u t ( k b i t )
TotalThroughput
Fig. 8. Packet Throughput vs. No of vehicle per lane.
From Figure 8, we observed that the total packet throughput
increases when the number of total vehicle communicating per
lane increases after the accident. This appreciation of
throughput value is due to the number of communicating
within the same lane in a near distance. However when
vehicle density increases, the throughput portray some increase
of throughput but the throughput per node decline after that.
This declination is due to the average load of data transmission
within each vehicle.
Throughput per Node
0
0.5
1
1.5
2
0 10 20 30
Number of Vehicle per Lane
Communicating
T h r o u g h p u t ( k b p s )
Throughput
Fig. 9. Packet Throughput vs. No of vehicle per lane.
Graph in Figure 10 depicts the average packet throughput
for peak hours and non-peak hours. We observed that the
total packet throughput for peak hours is higher compared to
the non-peak hours. During peak hours, the bus trip
frequency is higher and vehicle moves slower in urban city
as people were rushing to work accommodating the roads
space. The increase of bus density within the city assists the
ad hoc network formation. Hence the average total
throughput ratio much higher to the throughput value
portray within the sparse network model.
Average Packet Throughput
050
100
150
200
250
300
350
400
450
Environment (Dense/Sparse)
A v
e r a g e P a c k e t T h r o u g h p u t
( k b i t )
AODV-Peak
AODV-OffPeak
OLSR-Peak
OLSR-OffPeak
Fig. 10. Average Packet Throughput.
During off peak-hours, bus trip frequency drops and
subsequently reduces the density of buses within the road.
This scenario contributes to the declination of the average
total packet throughput from bus to the base-station. The
322
8/7/2019 Performance Analysis of Vehicular Ad hoc (2)
http://slidepdf.com/reader/full/performance-analysis-of-vehicular-ad-hoc-2 6/6
6
graph also indicates with the presence of ad hoc network
aids packet transmission, yielding higher packet throughput.
IV. SUMMARY
A realistic vehicular simulation model has been presented.
And yet for future, there are still ample of space for further
improvement remains for the research community to explore.
Realistic traffic model for instance have yet to be studied
thoroughly and modeled. Researcher may also consider on
expanding network and provide extensive simulation results on
various vehicle scenarios. Our future work aims to enhance
performance for existing routing algorithm for VANET.
R EFERENCES
[1] http://www.idsia.ch/~andrea/simtools.html
[2] SUMO - http://sumo.sourceforge.net
[3] Dynamic Source Routing (DSR), David B. Johnson,
David A. Maltz,1998, http://www.ietf.org/internet-
drafts/draft-ietf-manet-dsr-10.txt
[4] Optimized Link State Routing (OLSR), Adjih
Clausen, Jacquet Laouiti, Minet Muhlethaler,
Qayyum Viennot, 1998,
http://hipercom.inria.fr/olsr/draft-ietf-manet-olsr-
09.txt[5] NS2, http://www.isi.edu/nsnam/ns/
[6] AODV, Charles E. Perkins,Elizabeth M. Belding-
Royer,Santa Barbara,Ian D. Chakeres.
http://moment.cs.ucsb.edu/pub/draft-perkins-manet-
aodvbis-00.txt
323