MAC Performance Analysis for Vehicle to Infrastructure Communication Tom H. Luan*, Xinhua Ling, Xuemin (Sherman) Shen* *BroadBand Communication Research.

MAC Performance Analysis for Vehicle to Infrastructure Communication

Tom H. Luan*, Xinhua Ling , Xuemin (Sherman) Shen*

*BroadBand Communication Research GroupUniversity of Waterloo

§

§Research In Motion

2

Outline

1. Introduction to Vehicular Network2. Model of MAC in V2I communication3. Simulation4. Conclusion

2

3

Why Vehicular Networks ?

Internet becomes an essential part of our daily life Watch video on Youtube; order literature on

Amzone; catch the final moments of an eBay auction …

Americans spend up to 540 hours on average a year in their vehicles (10% of the waking time)

Internet access from vehicles is still luxury Vehicular Network

To provide cheap yet high throughput data service for vehicles on the road

3

4

V2V and V2I Communications

4

RSU (roadside unit)

Vehicle to RSU (V2R or V2I)

Vehicle to Vehicle (V2V)

Infotainment: Internet access, video streaming, music download, etc.

MAC throughput performance evaluation of V2I communication

5

Standard and Research Efforts

5

IEEE drafts 802.11p standard to permit vehicular communication 802.11a radio technology + 802.11e EDCA MAC Multi-channel: 6 service channels + 1 control

channel Drive-thru Internet

Using off-the-shelf 802.11b hardware, a vehicle could maintain a connection to a roadside AP for 500m and transfer 9MB of data at 80km/h using either TCP or UDP

[1] J. Ott and D. Kutscher, "Drive-thru Internet: IEEE 802.11 b for 'automobile' users," in IEEE INFOCOM, 2004

Image from http://www.drive-thru-internet.org/

6

CarTel in MIT [2] City-wide experiment showing the intermittent and

short-lived connectivity, yet high throughput while available

6

[2] V. Bychkovsky, B. Hull, A. Miu, H. Balakrishnan and S. Madden, "A measurement study of vehicular internet access using in situ Wi-Fi networks," in ACM MobiCom, 2006

Small scale network without considering MAC Link layer and transport

layer performance What if a great number

of vehicles moving fast?

Standard and Research Efforts (cont’d)

7

Problem Statement

MAC performance evaluation for fast-moving large scale vehicular networks

We consider 802.11b DCF Used by most trail networks, e.g., Drive-thru Compatible to WiFi device (e.g., iPod Touch) The basis of 802.11p MAC

7

8

Network Model

Perfect channel without packet loss and errors Saturated case: nodes always have a packet to transmit Multi-rate transmission according to the distance to RSU Spatial zones: the radio coverage of one RSU is divide into

Z = {0, 1, …, N} zones according to node transmission rate

p-persistent MAC: nodes transmit with a constant probability pz for different zone n in Z

8

Re

ceiv

ed

SN

R (

dB

)

21 NN - 1

1 2 N-1 N Zone

Markov chain

RSU RSU

n nmapMirror zones along RSU

Mobility Model Sojourn time of vehicles in

each zone n is geometrically distributed with mean tn

Within a period , vehicle moves from zone n to n+1 with the probability /tn, and no change with the left probability

9

Markov Model of Vehicle Nodes

9

Each node can be represented by {z(t), b(t)} z(t): zone the vehicle is current

in at time t b(t): the value of backoff

counter of the node at time t

1,0 1,1 1,2 1,W-1

2,0 2,1 2,2 2,W-1

N,0 N,1 N,2 N,W-1

Geometric distribution (p1)

N-1,0

Movement of Vehicles

Back off Interval Countdown

Geometric distribution (pN)

2D Markov chain embedded at the commencement of the backoff counter countdown

Upon the decrement of backoff counter, vehicle may either move to the next zone or stay in the original zone

When coming into a new zone, different transmission probability is applied

10

Simulation Setup

When arriving at the end of the road session (zone N), vehicles reenter zone 0 and start a new iteration of communication

Two schemes Equal contention window (transmission probability p) in all

zones Differential contention window in zones

10

RSU

Zone 0Zone NZone 1 Zone 2Zone 0 ... ...

Radio coverage of RSU is 250m, which is divided into 8 zones

By default, 50 vehicles move at constant speed with v = 80 km/h

11

Nodal Throughput in Each Zone n

11

sn = Average pkt length in each

trans.Mean interval between consecutive trans.

Nodal Throughput in Each Zone

S = ∑

Integrated Throughput

nXn sn

Using equal CW in all zones would suffer from performance anomaly

Where Xn is the node population in zone n

12

Increasing Velocity

With enhanced node velocity, nodes in front zones have higher throughput than the back zones The small CW in zone 4 benefits the following zones

System throughput reduces when velocity increases

12

13

Conclusion

Throughput performance evaluation of DCF in the vehicle to infrastructure communication

Increase the velocity would reduce the system throughput

Future work Optimal design of DCF (contention window) QoS provision with call admission control etc.

13

Question and Answers ?

Thank you !

14

bbcr.uwaterloo.ca/~hluan