PERFORMANCE MEASUREMENT OF CAR-TO-CAR AD-HOC

NETWORKS IN THE REAL WORLD ¹

Taehwan Kim, Woojin Han, Jinhyuk Lee, Dowon Hyun and Juwook Jang

Dept. of Electronic Engineering Sogang University Shinsu-dong Mapo-gu, Seoul, Korea 121-742

Keywords: Ad-hoc networks, Car-to-car networks, Inter-vehicle Communications, Wireless LAN.

Abstract: This paper shows the performance of car-to-car ad-hoc networks in the real world. Actually there are many

irregular driving environments, so we have established some scenarios about inter-vehicle communications

can happen in our actual life. And we have measured the transmission characteristics such as throughput,

delay and packet loss rate of the system for each case. We have installed laptops with IEEE 802.11b

wireless LAN and software for the ad-hoc network and packet transmission at each car. A test-bed has been

implemented to support TCP/UDP packet transmission on wireless ad-hoc environment such as AODV

(Ad-hoc On Demand Distance Vector) and OLSR (Optimized Link State Routing). Scenarios we have used

in this experiment present which algorithm of ad-hoc routing and transmission is suitable in real driving

environment and how performance metrics change according to different circumstances.

1 INTRODUCTION

For next generation vehicles, inter-vehicle

communication is necessary. Car-to-car network is

very dynamic from the point of view of mobility.

Moreover, because the inter-vehicle networks must

be wireless new telematics protocols and

applications should be developed in a different

manner compared with other wired network

protocols. For this reason, many telematics

companies try to develop the wireless network

system that is efficient and stable.

We have researched about wireless network for

next generation vehicles, too. As measuring

performance metrics such as throughput, delay, and

packet loss rate in the environment of dynamic car-

to-car networks, we can get information as follows –

first, what routing and transmission protocol is

suitable for next generation vehicles and second,

how performance metrics change according to

different circumstances.

Many researches about our subject have been

achieved by other researchers in past years, however

there are few studies that measure performance

metrics in the real world. Most of studies provide

results by not actual measurement but virtual

simulation. For the actual measurement, we have

implemented test-bed with OSI 7 layer components

using hardware and software.

First of all, as preparing laptop with 802.11b LAN,

we could meet requirements for physical layer and

link layer. And then we have used software to satisfy

the conditions of network layer, transport layer, and

application layer – session layer and presentation

layer, too. Especially, we have installed software

make nodes communicate by ad-hoc routing using

the protocol such as AODV (PERKINS 1999) and

OLSR (LAOUITI 2001). We have just used these

programs to implement our test-bed (AODV for

Windows and olsr-0.4.9). But we have developed a

unique program, the throughput measurement

program, used to send and receive packets and to

estimate throughput of the network. The throughput

measurement program is composed of two parts –

packet sender and packet receiver, and they have

ability to adjust packet generate frequency, size of a

packet, and transmission protocol like TCP or UDP.

_________________

1 This paper is supported by Seoul R&BD Program (10561)

Kim T., Han W., Lee J., Hyun D. and Jang J. (2006).

PERFORMANCE MEASUREMENT OF CAR-TO-CAR AD-HOC NETWORKS IN THE REAL WORLD.

In Proceedings of the International Conference on Wireless Information Networks and Systems, pages 79-85

 SciTePress

Figure 1: Architecture Overview of "Drive-thru" Internet.

The focus of our test-bed is to research the influence

of natural driving conditions that can happen in the

real world on the performance of car-to-car network.

It is clear that the results of our experiments are very

helpful to many developers who try to make more

practical inter-vehicle application program or

hardware.

The rest of this paper is organized as follows.

Section 2 describe the related work and elaborates

our contribution. In section 3, we discuss car-to-car

communication and some limitations of previous

works. Section 4 gives a brief description of the test-

bed we have implemented, and scenarios we have

used are shown in section 5. Section 6 displays

results achieved by the experiment. Finally, in

section 7, we conclude this paper confirming the

suitability of WLAN-based access technologies for

the vehicle network using Ad-hoc and pointing out

next steps in our research.

2 RELATED WORK

There are many trials to get information about the

performance of network in the car. Recently,

research about the “Drive-thru” Architecture by

University Bremen has been done (OTT 2004). This

study is about usage of internet in the car through

infrastructure like access points on the road. Figure 1

shows the architecture of “Drive-thru” internet. And

the study named “IMPORTANT (Impact of

Mobility on Performance of RouTing protocols for

Adhoc NeTworks)” is also the research about car-

network (BAI 2003).

However, the researches as stated above have

limitations. In the study “Drive-thru” internet, there

is of infrastructure for network between cars.

Besides, cars must be in the range of access point

installed on the road simultaneously to communicate

each other. And because study named

“IMORTANT” is by virtual simulation, the

environment and result can be different from those

of research in the real world.

3 BACKGROUND

Inter-vehicle communication

Figure 2: The current network using access point versus

ad-hoc network without access point.

Inter-vehicle communication can be used to

facilitate applications improving driving safety and

convenience. Potential uses of such applications are

dynamic traffic routing, driver assistance and

navigation, entertainment, co-operative driving, etc.

The existing ad-hoc networking infrastructure can be

leveraged and performance enhancement measures

can be innovated for provisioning seamless inter-

vehicle communication. As opposed to centralized

service, and ad-hoc network is much better suited for

vehicle-related applications that exchange data

having local relevance. The existing 802.11

compliant devices can be used for providing wireless

connectivity between moving vehicles. With the

advent of 802.11a hardware, bandwidths of up to 54

Mbps have become realizable. However, Vehicular

traffic scenarios pose greater challenges than the

indoor WLAN applications, due to associated

driving speeds, varying vehicular traffic patterns and

driving environments. Performance measurements

for 802.11 based wireless LANs have been done in

indoor office and industrial environment. These

results do not provide performance indication for the

more challenging vehicular scenarios. Through the

test we conduct, we investigate the performance

achievable by an 802.11b-based WLAN in vehicular

scenarios.

There are two types of inter-vehicle communication

network using Ad-hoc (Figure 2). One is Inter-

vehicle communication supported by AP (Access

Point), which has been discussed. However, this

approach is not cost-effective. It requires

development of exclusive infrastructure. Access

WINSYS 2006 - INTERNATIONAL CONFERENCE ON WIRELESS INFORMATION NETWORKS AND SYSTEMS

Figure 3: Hardware for test-bed.

points may be provided at each street corner, co-

located with traffic lights, or emergency phones, be

placed in parking lots or in rest areas or may be co-

located with gas stations or other shops in service

areas. The other is inter-vehicle communication

supported by ad-hoc routing algorithm (SINGH

2002). In this approach, it is important that each

mobile node can detect other's position and routing

path continuously.

In this paper, we focus on plain WLAN

connectivity and transport protocol behaviour-and

only briefly address implications on applications in

the end. Our goal is to prove that WLAN technology

is capable of enabling the vehicle network using ad-

hoc in the first place and to document the

communication characteristics we have observed

with different measurement configurations using

UDP and TCP as standard transport protocols. After

simulation and measurement of transport protocol in

ad-hoc mode, we describe the multi-hop mobile

vehicular test-bed, our design decisions and driving

experiences

4 TEST-BED IMPLEMENTATION

We have implemented a test-bed to measure the

performance of wireless ad-hoc network and to get

knowledge about the influences of driving

environment upon the state of network. Because the

test-bed has not only hardware but also software, we

could implement the network system with entire

components of OSI 7 layers.

4.1 Hardware

We have used the cars with laptop as shown in

(a)

(b)

Figure 4: Software for test-bed (a) Packet Sender, (b)

Packet Receiver.

Figure 3. Because range that can be covered by

802.11b WLAN card we have used is only about

10m, we have equipped the PCMCIA card with an

external antenna that has been placed at the right

hand side of the vehicles. By the experiments for

reference we have done before the planned

measurement, we have already know that the range

of PCMCIA card with an external antenna come

close to about 100m. Owing to the external antenna,

we could perform experiment without limitation on

range of electric wave.

4.2 Software

We have installed software for wireless ad-hoc

routing and sending and receiving packets. The

program for ad-hoc is not developed by us, we have

used only. However we have developed unique

program for sending and receiving packets – the

name of this program is Throughput Measurement

Program.

As shown in Figure 4. The Throughput

Measurement Program consists of two parts – Packet

Sender and Packet Receiver. Figure 4(a) describes

the Packet Sender and Figure 4(b) describes the

Packet Receiver and their functions.

5 SCENARIOS

Most important point of this study is that

experiments have been done under the driving

environment in the real world. But there are lots of

situations that can be happen in our real life. So we

have established some scenarios treated as important.

Basically, we have used 3 cars for these experiments,

and cars can participate ad-hoc routing and network.

These 3 cars run in order of source node (S) - relay

PERFORMANCE MEASUREMENT OF CAR-TO-CAR AD-HOC NETWORKS IN THE REAL WORLD

node (R) - destination node (D) in a line, and source

node sends packets to destination packet through

TCP or UDP protocol.

5.1 Variable Distance and Velocity

Our first scenario is changing distance between cars

and velocity of cars. Because the environment of

network is wireless ad-hoc with limitation of range,

distance and velocity are very important variables.

We have changed the distance between each car

from 30m to 70m, and velocity of each car from

30km/h to 70km/h

5.2 Overtaking

Figure 5: Scenario description about the overtaking.

When many cars run on the road, a lot of situation

can happen. Second case we have assumed is seen in

Figure 5. At the beginning of experiment, a

destination node (D) and a source node (S) run in a

line. After a few seconds, node D overtake the car

were running ahead of node D, and these 3 cars run

in order for several seconds.

5.3 Breakaway and Re-entry

Figure 6: Scenario description about the breakaway and

re-entry.

Figure 6 shows third scenario of us. Like the

preceding scenario, 3 cars run in order, S-R-D, as

source sends packets to destination. After some time

from the beginning of communication, an

experimenter disables the LAN of node R. And after

some time again, the experimenter turn the LAN of

node R on. We set the interval between each state

for 15 seconds.

5.4 The Relay Effect

Figure 7: The relay effect experiment 1.

Figure 8: The relay effect experiment 2.

This scenario has two parts of experiments. First, on

one line of the road, we have compared with S-D

communication and S-R-D communication. That is,

performance of network with relay node and without

relay node is compared. Figure 7 shows this scenario.

In this case, we have kept all of the cars being stop,

and set distance between each car 50m.

Secondarily, because we have used wireless LAN,

we have chosen the scenario that can show

influences of any obstacles such as walls, trees, and

buildings. As shown in Figure 8, we set up the

environment that the tall building block the electric

wave between node S and node D. The main

purpose of this example is to know how many

influences relay node give to the wireless ad-hoc

network.

WINSYS 2006 - INTERNATIONAL CONFERENCE ON WIRELESS INFORMATION NETWORKS AND SYSTEMS

6 MEASUREMENT

We have measured performances such as throughput,

delay, packet loss rate in each case abovementioned.

For the transport protocol, TCP and UDP have be

used, and AODV and OLSR have be used for the ad-

hoc routing protocol. But because the case of OLSR

displays more efficient performance than the case of

AODV, we present the results gotten by OLSR only.

Figure 9 shows the map the experiments were

accomplished.

Figure 9: Map of the road the experiments were

accomplished.

At the first, we have measured transmission

characteristics such as throughput, delay and packet

loss rate as change the distance between cars and

velocity of cars. Figure 10 shows the results of the

experiments related to distance and velocity. Figure

10(a) and Figure 10(b) display the result of

throughput. According to the results, basically, the

effect by velocity is more remarkable than by

distance. The reason of these results is by

characteristic of 802.11. 802.11 send RTS (Request

to Send) and CTS (Clear to Send) before sending the

main data to make the most suitable condition of

data transmission. If the vehicles speed up, optimal

conditions change, so, the difference of performance

is notable. The phenomenon like this is seen in the

result about the delay and the packet loss rate as

shown in Figure 10(c) through Figure 10(e). By the

way, as we can see through Figure 10, UDP make

system more efficient than TCP. This is because

UDP does not have the process of dividing the

message into packets by one end point and re-

assembling divided packets into a message by the

other.

0.5

1.5

2.5

3.5

30 50 70

Dis tanc e

(

)

Throug hput(Mbps

0km/ h

30km/ h

50km/ h

70km/ h

(a) Throughput - TCP

30 50 70

Distance

(

)

Thoughput(Mbps

(b) Throughput - UDP

100

150

200

250

300

350

30 50 70

Distance

(

)

Delay ( ms)

100

150

200

250

30 50 70

Distance

(

)

Delay ( ms)

(d) Delay - UDP

30 50 70

Distance

(

)

Number of Lost Packets

(e) Number of lost packets – UDP

Figure 10: Results of experiment for various distance and

velocity.

PERFORMANCE MEASUREMENT OF CAR-TO-CAR AD-HOC NETWORKS IN THE REAL WORLD

0.5

1.5

2.5

3.5

4.5

(Before

overtaking)

(Overtaking

peroid)

(After overtaking)

Thro ughp ut(Mb ps)

TCP

UDP

(a)

100

120

140

160

180

200

(Before

overtaking)

(Overtaking

peroid)

(After overtaking)

Delay(ms)

(b)

Figure 11: Results of experiments about the overtaking (a)

Throughput, (b)Delay.

0.5

1.5

2.5

3.5

(Before

breakaway)

(After breakaway) (After re-entry)

Thro ughp ut(Mbps)

(a)

100

150

200

250

300

350

400

(Before

breakaway)

(After

breakaway)

(After re- entry )

Delay(ms)

(b)

Figure 12: Results of experiments about the breakaway

and re-entry (a)Throughput, (b)Delay.

Section 5.2 describes a scenario about the

overtaking, and Figure 11 shows the results of this.

We can discover that the performance in the

overtaking period goes down compared with before

and after the overtaking. This result is by the sudden

increase in velocity of vehicle pass ahead. The

influence of velocity is seen in a previous

experiment and the result.

Figure 12 displays the results by the

experiment performed according to our third

scenario – breakaway and re-entry. In this case,

efficiency of the network after the breakaway of

relay node is worst. This is because this after-

breakaway-period (before re-entry) is the stage each

node reset the ad-hoc routing table and relay node

does not exist. And after re-entry of the node broken

away from the network, the system could not

recover the performance of before-breakaway-period

because the need of processing time to reset the

routing table.

Lastly, experiments for get information about

the role of the relay node produce results shown in

Table 1 and Table 2. As we can show, when the

relay node exists, the performance is more efficient.

Especially, the case that a building disturbs the

communication between source and destination node

produces remarkable results, because the

communication is impossible under the environment

without the relay node. But if the relay node is

placed at the edge of the building, the destination

node can receive packets sent by the source node.

7 CONCLUSION

In this paper, we present a test-bed for wireless ad-

hoc network for car-to-car communication. Due to

the test-bed we have implemented, we could get

results of the experiment designed to know the

influences of driving environment on the

performance of the inter-vehicle network. The core

of this paper is that all of the experiments and

measurements are accomplished in the real world,

not in the virtual world by simulation.

Because of inter-vehicle ad-hoc network have

to use the wireless LAN with limitation of

bandwidth, more efficient routing and transmission

protocols are necessary. So, our study to measure the

performance of wireless network on the various

environments has meaning. But there are more

limitations yet. We would like to investigate into this

issue further in the future and improve this limitation

of MANET over vehicles.

WINSYS 2006 - INTERNATIONAL CONFERENCE ON WIRELESS INFORMATION NETWORKS AND SYSTEMS

Table 1: Results of experiments about the relay effect 1.

Transmission

Protocol

TCP UDP

Relay O X

Rise

Rate

O X

Rise

Rate

Throughput

(Mbps)

1.2 1.36 13% 1.03 2.1 104%

Average

Delay(ms)

214 187 14% 165 80 106%

Packet Loss

Rate(%)

6 3 100%

Table 2: Results of experiments about the relay effect 2.

Transmission Protocol TCP UDP

Throughput(Mbps) 2.84 3.45

Average Delay(ms) 216 145

Packet Loss Rate(%) ㅡ 1

REFERENCES

Perkins, C. E. and Royer, E. M., 1999, Ad-hoc On-

Demand Distance Vector (AODV) Routing. In wmcsa

2nd IEEE Workshop on Mobile Computer Systems and

Applications, 90.

Laouiti, A., Muhlethaler, P., Clausen, T. and Jacquet, P.,

1999, Optimized Link State Routing Protocol. In IEEE

INMIC, Pakistan.

AODV for Windows, Available from : http://moment.

cs.ucsb.edu/AODV/aodv-windows.html

olsr-0.4.9, Available from: http://seattlewireless.net/

OlsrExperiments

Ott, J. and Kutscher, D., 2004, Drive-thru Internet: IEEE

802.11b for “Automobile” Users. In IEEE

INFOCOMM 2004.

Ott, J. and Kutscher, D., 2004, The “Drive-thru”

Architecture: WLAN-based Internet Access on the

Road. In IEEE VTC 2004.

Bai, F., Sadagopan, N. and Helmy, A., 2003,

IMPORTANT: A frame work to systematically

analyze the Impact of Mobility on Performance

RouTing protocols for Adhoc NeTworks. In IEEE

INFOCOMM 2003.

Ferreira, E., Subramanian, E., Klausner, M., Hathout, J. P.

and MANSTETTEN, D., 2001, Inverts project, Inter-

vehicle Real-Time System. The Book, Final Report

2000-2001, Bosch Research and Technology center.

Atheros Communication’s White paper on 802.11a Range

and System Capacity. Measured Performance of 5-

GHz 802.11a wireless LAN System.

Singh, J. P., Bambos, N., Srinivasan, B. and Clawin, D.,

2002, Wireless LAN Performance Under Varied Stress

Conditions in Vehicular Traffic Scenarios. In IEEE

VTC 2002.

Atheros Communication’s White paper, 802.11 Wireless

LAN Performance. 2003.

Johnson, D. B., Maltz, D. A. and Broch, J., 2001, DSR:

The dynamic source routing protocol for multi-hop

wireless ad-hoc networks. The Book, Perkins, C., ed.

Addison-Wesley, 139-172.

Hass and Pearlman, 1999, The Zone Routing Protocol

(ZRP) for Ad-hoc Networks. In IETF Internet Draft.

Ramasubramanian, V., Haas, Z. J. and Sirer, E. G., 2003,

SHARP: A Hybrid Adaptive Routing Protocol for

Mobile Ad-hoc Networks. In MobiHoc'03, June 1-3,

2003, Annapolis, Maryland, USA.

Breslau, L., Estrin, D., Fall, K., Floyd, S., Heidemann, J.,

Helmy, A., Huang, P., McCanne, S., Varadhan, K., Xu,

Y. and Yu, H., 2000, Advances in network simulation.

In IEEE computer, vol.33, no.5, May 2000, 59-67.

PERFORMANCE MEASUREMENT OF CAR-TO-CAR AD-HOC NETWORKS IN THE REAL WORLD