TrafficView: A Scalable Traffic Monitoring System

Tamer Nadeem, Sasan Dashtinezhad, Chunyuan Liao, Liviu Iftode*

Department of Computer Science

University of Maryland, College Park*Now with Rutgers University

2

TrafficView

• Enable drivers to see vehicles in front of their cars, farther than they can see, in real-time

• Use vehicle-to-vehicle ad hoc networks

3

How TrafficView Works

• Each vehicle has an embedded system– GPS receiver (location, speed, time)– Short-range wireless NIC– On-Board Diagnostics interface (optional)

Receive data fromremote vehicle

Non-validateddataset

Validate

Validateddataset

Local data

Display

Broadcast data

4

Need for Data Aggregation

• Ad hoc networks of vehicles are dynamic

• Data propagation must be simple

• Send all data in one packet (up to MTU)

• Use data aggregation to put as much

information as possible in one packet

5

How Far Can You See?

• Problem

– How to aggregate data to see vehicles as far as

possible with “acceptable” accuracy loss

• Natural Solution

– Aggregate data for vehicles that are close to

each other

– Perform more aggregation as distance

increases

6

Outline

• Motivation and Problem Definition

• Data Representation

• Aggregation Algorithms

• Evaluation

• Conclusions and Future Work

7

Data Representation

• Vehicles store records:

– Vehicle ID (ID), position (POS), speed (SPD),

broadcast time (BT)

• Broadcast time: the time at which the originating

vehicle sent out the record

• An aggregated record contains more

than one ID

8

Aggregated Records

• Having n records

• Calculate the aggregated record’s fields:

},,,}...{,,,{ 1111 nnnn BTSPDPOSIDBTSPDPOSID

}...min{ 1 na BTBTBT

nodecurrent the to vehicleof distance idi

n

i i

ia d

POSPOS

1

n

i i

ii d

SPDSPD

1POS and SPD are weighted averages.

9

Aggregation Algorithms

• Ratio-based

• Cost-based

10

Ratio-based Aggregation

Current Vehicle

Aggregation ratio: inverse of the number of records that would be aggregated in one record

Portion value: amount of the remaining space in the broadcast message

3. In each region, each two consecutive records that are closer than the merge threshold, are merged

1. Calculate region boundaries

2. Calculate merge thresholds

11

Cost-based Algorithm• The Ratio-based algorithm selects the records to be

aggregated blindly!

• Assign a cost to merging two records, select

records corresponding to lowest cost

• Cost function:

– High cost to close vehicles

– Minimize error due to merging records

– Minimize number of cars in merged records

a

aa

d

sddsdd 2211 ||||cost

12

Information Aging• Problem

– Vehicles move and change speed– Records can be out-of-date– Received information might be invalid

• Solution– Delete record if no information about that vehicle

is received in a while– Compute expected delay for each record received– Store record only if

|actual delay – expected delay| < threshold

13

Evaluation

• Metrics

• Road Scenarios

• Simulation Results

14

Metrics• Visibility

– Average distance ahead about which a vehicle has

information

• Accuracy

– Average position error introduced due to aggregation

• Knowledge Percentage

– Average percentage of vehicles in each region ahead

about which a vehicle has information

15

• Evenly distributed entries and exits• Random constant speed during

time intervals• Changing lanes randomly

Traffic Model

0

5

10

15

20

25

30

35

40

45

50

0 1 2 3 4Average Number of Lane Changes

Perc

en

tag

e o

f C

ars

(%

)

0

5

10

15

20

25

30

20 25 30 35 40

Average Speed (m/s)

Per

cen

tag

e o

f C

ars

(%)

16

Simulations• NS-2 simulations

– 802.11b with 11Mbps bandwidth

– transmission range of 250m

– MTU = 2312 bytes

• 15,000m road, 4 lanes

• 300s duration of simulation

• Algorithms:– Ratio-based, Cost-based, Non-aggregation, and Brute-force Cost-

based

• Selected parameters using preliminary simulations

17

Scenarios

Name # of nodes Avg. speed (m/s) Avg. gap (m)

Rush-hour 690 10 100

City 780 20 100

High-density highway 870 30 100

Low-density highway 548 40 175

18

Effect of Road Parameterson Visibility (1)

Ratio-basedAggregation

Cost-basedAggregation

19

Effect of Road Parameterson Visibility (2)

Non-aggregation Brute-forceCost-based

20

Visibility (High-density Highway)

21

Accuracy (High-density Highway)

22

Knowledge Percentage(High-density Highway)

23

What We Learned

• Intuitively, cost-based algorithm appeared

to be a better choice

• Cost-based algorithm is only marginally

better for relatively closer distances

• Ratio-based algorithm is better for farther

away distances and is more flexible

24

Conclusions

• TrafficeView provides drivers with real-

time view of vehicles in front of their cars

• Designed and evaluated two aggregation

algorithms using realistic road scenarios

• Ratio-based algorithm is a good algorithm

– Good visibility and small position error

25

Future Work

• Working on prototype implementation

• Linear programming model to

automatically calculate the aggregation

parameters

• Privacy and Trust

26

Thank You!

http://www.cs.umd.edu/~nadeem/projects/trafficview