1 Vehicular Sensor Networks for Traffic Monitoring In proceedings of 17th International Conference on Computer Communications and Networks (ICCCN 2008)

1

Vehicular Sensor Networks for Traffic Monitoring

In proceedings of 17th International Conference on

Computer Communications and Networks (ICCCN 2008)

2

Outline Introduction

Motivation and Problem

Metric Definition

Traffic Status Estimation

Performance Evaluation

Future Work and Conclusion

3

Introduction

Traffic monitoring in city urban area

Traditional approach: loop detector, camera,etc

infrastructure cost

maintenance cost

communication cost

not scalable

4

Another way?

The existing vehicular sensor networks of taxi companies vehicle dispatching security purposes not special for traffic monitoring

Whether it can be used for traffic monitoring?

If “yes”,Advantage: Low infrastructure cost Low maintenance cost Cover the entire road network, scalable

5

What we have…

Data basis and features: Long sampling interval due to communication cost

Sparse and incomplete information

Error, etc.

6

Outline Introduction

Motivation and Problem

Metric Definition

Traffic Status Estimation

Performance Evaluation

Future Work and Conclusion

7

Motivation

What sort of performance for traffic monitoring we might expect from such vehicular sensor networks providing sparse and incomplete information

Now in Shanghai, we utilize a test

bed with mobile sensors

installed in about 4000

taxis

8

Problem

Whether we can demonstrate the feasibility of taxi-based sensor networks for traffic monitoring?

Whether the tradeoff between the accuracy of traffic status estimation and low communication cost can be well handled?

9

Outline Introduction

Motivation and Problem

Metric Definition

Traffic Status Estimation

Performance Evaluation

Future Work and Conclusion

10

Metric definition Three key characteristics in macroscopic

traffic-flow model:

flow rate

mean traffic speed

density

Public tends to consider more in terms of mean speed rather than flow rate or density in evaluating the quality of their trips

11

Definitions of mean traffic speed freeway VS roads in urban area

Length: Time cost:

iL

t

link >> i

i

12

Whole time cost ∆t to pass a link

=traveling time ∆t1+ intersection delay ∆t2

For a given link Li with length li, the mean traffic speed at time tk is defined as:

)(|)(|1

)(

ki tCcc

ki

iki

ttC

ltV

13

Outline Introduction

Motivation and Problem

Metric Definition

Traffic Status Estimation

Performance Evaluation

Future Work and Conclusion

14

A sample data from a sensor is defined by a 4-tuple D(SID, T, , ), and two consecutive data samples can construct a data pair.

A data pair from sensor s can be defined as:

p(s, t1, t2) = {s, t1, 1, t2, 2}

1 and 2 are the geographic coordinates from the consecutive data samples at t1 and t2, respectively

15

The link-based algorithm (LBA)

LBA only aggregates data pairs of sensing data from link Li as well as links adjacent to either of intersection nodes of Li.

16

The vehicle-based algorithm (VBA) VBA utilizes every available data pairs and

disseminates them back to all links traveled to estimate mean traffic speed.

17

A vehicular mobile sensor system: Intelligent Traffic Information Service (ITIS)

18

Outline Introduction

Motivation and Problem

Metric Definition

Traffic Status Estimation

Performance Evaluation

Future Work and Conclusion

19

Performance Evaluation

Large-scale field testing on arterial and inferior roads

20

The testing results showed VBA-based is better than LBA-based algorithms due to the data feature. More specially, the average error of VBA-Avg can be within only 17.3%

21

Lessons Learned

Map-matching

Poor map-matching performance degrades the accuracy of traffic status estimation

22

Traffic light

The mean speed of whole trip of 56 km is 21.1 km/h.

traffic light delays: 82 minutes

total time cost: 159 minutes

23

Outline Introduction

Motivation and Problem

Metric Definition

Traffic Status Estimation

Performance Evaluation

Future Work and Conclusion

24

Conclusion

A performance evaluation study has been carried out in Shanghai by utilizing the sensors installed on 4000 taxis for traffic monitoring

Two types of traffic status estimation algorithms, the link-based and the vehicle-based, are introduced based on such data basis.

The results from large-scale testing cases demonstrate the feasibility of such an application in most of cities

25

thanks!