Using DTMon to Monitor Transient Flow Traffic

USING DTMON TO MONITOR TRANSIENT FLOW TRAFFIC

Hadi Arbabiand Michele C. Weigle

Department Of Computer ScienceOld Dominion University

Second IEEE Vehicular Networking Conference, December 2010, NJ

Hadi Arbabi and Michele C. Weigle {marbabi, mweigle}@cs.odu.edu 2

Motivation

Real-time monitoring of traffic Accurate estimation of travel time and speed

Required in transient flow traffic (e.g., congestion) Fixed point sensors and detectors cannot estimate

travel time and space mean speed

Trends toward probe vehicle-based systems

Dynamic points of interest Augment current technologies

3

Content

INTRODUCTION Traffic Monitoring

Dynamic Traffic Monitoring (DTMon) Task Organizer Vehicles Virtual Strips Methods of Message Delivery

APPROACH Monitoring Traffic Data in Rural Areas

Highways

EVALUATION Transient Flow Traffic

SUMMARYHadi Arbabi and Michele C. Weigle {marbabi, mweigle}@cs.odu.edu

Hadi Arbabi and Michele C. Weigle {marbabi, mweigle}@cs.odu.edu 4

Introduction

Monitoring Vehicle classification Count information

Flow rate Volume Density

Traffic speed Time mean speed (TMS) Space mean speed (SMS)

Travel time (TT)

Traffic Management Center (TMC)

Hadi Arbabi and Michele C. Weigle {marbabi, mweigle}@cs.odu.edu 5

Technologies In Use

Fixed point sensor and detectors Inductive loop detectors (ILD) Acoustic sensors Microwave radar sensors Video cameras

Probe vehicle-based system Automatic vehicle location (AVL) Wireless location technology (WLT) Automatic vehicle identification (AVI)

Hadi Arbabi and Michele C. Weigle {marbabi, mweigle}@cs.odu.edu 6

Dynamic Traffic Monitoring (DTMon)

DTMon - A probe vehicle-based system using VANET and dynamically defined points of interest on the roads Task Organizers (TOs) Vehicles Virtual Strips (VS)

Imaginary lines or points

Methods of Message Delivery

Hadi Arbabi and Michele C. Weigle {marbabi, mweigle}@cs.odu.edu 7

DTMon: Task Organizer & Virtual Strips

TO

Virtual

Strip

Virtual

Strip

Virtual Segment

Hadi Arbabi and Michele C. Weigle {marbabi, mweigle}@cs.odu.edu 8

Task Organizer (TO) Communicates with passing

vehicles Assigns measurement tasks Collects reports from the vehicles Organizes received measurements Informs upcoming traffic conditions

Multiple TOs Centralized

Aggregate information about the whole region

Hadi Arbabi and Michele C. Weigle {marbabi, mweigle}@cs.odu.edu 9

Vehicles

Equipped GPS and DSRC communications device CPU and Required Applications

Record Speed GPS Position Travel Direction Timestamp Classification, Route Number, and …

Receive tasks from a TO Triggered at a specific time, speed, or location

Report Forwarded to the listed TOs Stored and carried to the next available TO

Hadi Arbabi and Michele C. Weigle {marbabi, mweigle}@cs.odu.edu 10

Multiple TOs Multiple VS Multiple VS and Segments

Dynamically Defined Multiple TOs

A Sample Task From TO to Vehicles

Hadi Arbabi and Michele C. Weigle {marbabi, mweigle}@cs.odu.edu 11

Methods of Message Delivery Regular Forwarding (RF) Store-and-Carry (SAC) [if multiple

TOs] Hybrid

RF+SAC

Hadi Arbabi and Michele C. Weigle {marbabi, mweigle}@cs.odu.edu 12

Evaluation

Several experiments using VANET modules that we developed for the ns-3 simulator

•H. Arbabi, M. C. Weigle, "Highway Mobility and Vehicular Ad-Hoc Networks in ns-3," In Proc. of the Winter Simulation Conference. Baltimore, MD, December 2010•Highway Mobility for Vehicular Networks (Project and Google Code)• http://code.google.com/p/ns-3-highway-mobility/

Hadi Arbabi and Michele C. Weigle {marbabi, mweigle}@cs.odu.edu 13

(Overview)

In our previous work Message Reception Effect of Traffic Density, Flow Rate, Speed Effect of Market Penetration Rate Effect of Transmission Range Effect of Traffic In Opposite Direction Distance From TOs Latency and Message Delay Comparison among Methods of Message

Delivery

Hadi Arbabi and Michele C. Weigle, “Monitoring Free-Flow Traffic using Vehicular Networks,” In Proceedings of the IEEE Intelligent Vehicular Communications System Workshop (IVCS). Las Vegas, NV, January 2011

Hadi Arbabi and Michele C. Weigle {marbabi, mweigle}@cs.odu.edu 14

Evaluation

Factors that can affect the Quality of Data Market penetration rate (PR) Method of Message Delivery

Message Reception Rate (MRR) Information Reception Rate (IRR)

IRR ≈ MRR x PR

Latency and Message Delay Methods that can collect more

information from vehicles with less latency are preferred in up-to-date traffic monitoring

Hadi Arbabi and Michele C. Weigle {marbabi, mweigle}@cs.odu.edu 15

Simulation Setup Bi-directional four-lane highway

TO1 is located at 1 km away

TO5 is located at 5 km away (optional secondary TO)

Vehicles enter the highway with Medium flow rate (average 1800 veh/h)

Uniform Distribution

Desired speed 65±5 mph (29±2.2 m/s) Normal Distribution

20% of vehicles are Truck Uniform Distribution

Hadi Arbabi and Michele C. Weigle {marbabi, mweigle}@cs.odu.edu 16

Simulation Setup Non-recurring congestion (5 min

stoppage) Transient Flow Traffic Stopping a vehicle in the first lane after

5 minutes for 5 minutes Between VS1 and VS2 outside the

communication range (300 m) of TO1

Stopped vehicle starts moving, allowing traffic flow to gradually return to normal

Hadi Arbabi and Michele C. Weigle {marbabi, mweigle}@cs.odu.edu 17

Comparison

The performance of DTMon compared with Actual simulation status (ground truth) Fixed point sensors and detectors

Actual simulation data sampled from VS1 and VS2

AVL Equipped Trucks

10 runs of the simulation (20 min each) for each experiment

Test with penetration rates of 5, 10, 25, 50, and 100%

Hadi Arbabi and Michele C. Weigle {marbabi, mweigle}@cs.odu.edu 18

MRR and IRR

PR Actual

RFRF+w/

opp

RF+SAC

5% 100 0.00 0.00 525% 100 1.43 1.43 2550% 100 26.01 28.10 50

100% 100 79.50 91.01 100

PR Actual

RFRF+w/

opp

RF+SAC

5% 100 0.00 0.00 10025% 100 5.71 5.71 10050% 100 52.03 56.20 100

100% 100 79.50 91.01 100

MRR

IRR

15%

7%

Message Reception:

RF+SAC > RF

Hadi Arbabi and Michele C. Weigle {marbabi, mweigle}@cs.odu.edu 19

MRR

VS2

RF + SAC = RF + Rest

Higher Penetration = Higher RF = Less Delay

Hadi Arbabi and Michele C. Weigle {marbabi, mweigle}@cs.odu.edu 20

Estimated Travel Time (ILDs vs. Actual)

Fixed Point Sensor and Detector’s Poor Estimation of TT and SMS

Hadi Arbabi and Michele C. Weigle {marbabi, mweigle}@cs.odu.edu 21

Travel TimeVS2

VS2

Quality of DataRF+SAC >= RF > AVL

Hadi Arbabi and Michele C. Weigle {marbabi, mweigle}@cs.odu.edu 22

Space Mean Speed (SMS)VS2

VS2

Hadi Arbabi and Michele C. Weigle {marbabi, mweigle}@cs.odu.edu 23

Flow Rate

VS2

Count Information (e.g., Flow Rate and Volume)

Only in High PR

Hadi Arbabi and Michele C. Weigle {marbabi, mweigle}@cs.odu.edu 24

Message Delay

TO1VS2TO5

RF Delay Very Low

RF+SAC Delay 1. Amount of Carried Messages2. TTMore RFLess Delay

More SAC More Delay

Hadi Arbabi and Michele C. Weigle {marbabi, mweigle}@cs.odu.edu 25

Quality of Data

Good Estimate?

Sensors and Detectors AVL DTMon

Flow Rate and Density Yes No

See Next Table

TMS YesUnderestimat

e Yes

Travel Time Not Available Overestimate Yes

SMS Not AvailableUnderestimat

e Yes

Vehicle Classification Not Accurate Limited Yes

t-test Alpha = 0.05 (Confidence

> 95%)

Hadi Arbabi and Michele C. Weigle {marbabi, mweigle}@cs.odu.edu 26

Quality of Data

High Quality Estimation Conf. ≥ 95%

Traffic Density

orPenetration

Rate

Message Delivery Method

Flow Rate and Density High Any

Classification,TMS

Travel Time,or

SMS

LowSAC, RF+SAC,

or DTR+SAC

Medium or High Any

t-test Alpha = 0.05 Confidence

> 95%

Hadi Arbabi and Michele C. Weigle {marbabi, mweigle}@cs.odu.edu 27

Summary

DTMon can estimate good quality Travel Time and Speed

DTMon can detect transition in traffic flow using estimated Travel Time and Speed

DTMon can estimate good quality flow rate and density in higher penetration rates RF and RF+SAC have similar performance in higher

penetration rates Using RF+SAC is an improving option in low

penetration rates DTMon can augment current technologies

and monitoring systems

Hadi Arbabi and Michele C. Weigle {marbabi, mweigle}@cs.odu.edu 28

Questions?

Hadi Arbabi and Michele C. Weigle Department of Computer Science at

Old Dominion University Vehicular Networks, Sensor Networks, and

Internet Traffic Research http://oducs-networking.blogspot.com/ {marbabi, mweigle}@cs.odu.edu

This work was supported in part by the National Science Foundation under grant CNS-0721586.