Real-Time Work Zone Travel TimeReal-Time Work Zone Travel Time

ConclusionAutomatic License Plate Recognition (ALPR) cameras,

Bluetooth, and RADAR were investigated for their travel time measurement capabilities. ALPR and RADAR presented a potential lane bias issue in congestion while Bluetooth showed a potential bias to slower vehicles. In addition to the importance of data accuracy given the need for fast, temporary deployments, data collection equipment should also have the following characteristics:

• Real-time data reporting

• Quick and Simple Deployment

• Portable power supply

• Low Maintenance

Finally, as all tested technology is point detection based, it is critical that any deployment covers the full extent of anticipated work zone related congested area.

This research was sponsored by the Georgia Department of Transportation under Project RP 11-15. Opinions expressed here are those of the authors and not necessarily those of the Georgia Department of Transportation.

OverviewWork zones are a major source of non-recurrent congestion.

Providing accurate and timely information to motorists regarding travel time and delays is critical and can improve mobility and safety through work zones. The objective of this research is to investigate the capability of various travel time data collection technologies to produce accurate work zone travel time information in real-time. In this effort vehicle detection and travel time data were collected along freeway work zones in Atlanta, Georgia using multiple data collection technologies, including Automated License Plate Recognition (ALPR) Cameras, Bluetooth, RADAR, and high definition video. The collected high definition video footage was post-processed utilizing a proprietary video-processing program developed at Georgia Tech that allows manual entry of vehicle license plate information. The travel times and vehicle count information from the manual video license plate processing are then used as a baseline for comparison against both the ALPR, Bluetooth, and RADAR data results.

EquipmentBluetooth TechnologyCommon in many standard devices: cell phones, headsets, GPS, and

vehicles.Each active Bluetooth device constantly transmits a unique MAC

(Media Access Control) address: six pairs of two hexadecimal digits separated by colons (e.g., 00:02:72:20:67:2A)

Identify vehicles with Bluetooth devices moving in the traffic stream at multiple locations and calculate travel time

Automatic License Plate Recognition (ALPR) TechnologyALPR camera systems

digitally capture vehicle license plate characters

Identify license plate numbers at multiple locations and calculate travel time

I-285 Data Collection

Location of the I-285 work zone corridorSource: Google Maps

Vehicle detection and travel time data were collected in a freeway work zone corridor along the northwestern section of Interstate-285 in Atlanta, Georgia on six days in Fall 2012 and one day in Spring 2013

Data was collected at two interchange sites

Bluetooth, ALPR, and RADAR data were collected from side-fire locations

High definition video data was collected from overpass locations

RADARRADAR captures speeds and calculates travel time between two

pointsSelf locates using GPS and cell signal to report speeds and locationComputer, battery, and sensors contained in a Type II traffic barrelUtilizes cell modem when available and Satellite modem when not

Date Site A Site BLength of

Data Collection

Direction of Travel

Work Zone?

09-07-12

Paces Ferry Road

Northside Drive 2 hours EB No

09-12-12

Northside Drive

Roswell Road 2 hours EB No

09-14-12

Northside Drive

Roswell Road 2 hours EB No

09-29-12

Riverside Drive

Paces Ferry Road 3 hours WB No

10-20-12

Paces Ferry Road

Northside Drive 2.5 hours EB Yes

11-10-12

Roswell RoadChamblee Dunwoody

Road2 hours EB No

4-13-13 South Cobb Drive

Mt. Wilkinson Parkway

5.5 hours EB Yes

Summary of the six I-285 deployments

Set-up of Bluetooth unitsImage Courtesy: Kathryn Colberg

Set-up of the ALPR cameras and equipmentImage Courtesy: Kathryn Colberg

Setting up a RADAR unitImage Courtesy: Wonho Suh

Travel Time MatchingRaw vehicle detection data from Bluetooth and ALPR equipments are

matched across the two sites using a travel time matching algorithmThis matching algorithm finds exact matches from all of the

equipment data and also finds additional ALPR matches by making all possible plate number combinations for plate reads containing bracketed digits.

Draft Travel Time Results & DiscussionTravel times from the various equipment were compared using travel

time plots and Y-Y plots.

Figure 1: ALPR, Custom Bluetooth, and Baseline comparison travel time plot from October 20th, 2012

Bluetooth units: slower moving vehicles are within the detection zone for longer lengths of time

ALPR system: Adjacent (freeway lanes) are less likely to be obstructed from the side-fire ALPR camera, resulting in potential bias to adjacent lane speed.

Saturday, Octobe

r 20th,

2012

(Active Work

Zone)

Lane at

Site A

Total Vehicle Volume

at Site A

Total Vehicle Volume

at Site B

# of Manual Baselin

e Matche

s (by lane

at site A)

Avg. Manual Baselin

eTravel Time

(by lane at

site A)

# of Unique ALPR

Detections

(Site A)

# of Unique ALPR

Detections

(Site B)

# ofALPR

Matches (by Lane

at site A)

Avg. ALPR Travel Time (by

lane at site A)

Lane 1

2,260 559 338 10:29 unknown unknown 2 17:08

Lane 2 791 928 59 14:56 unknown unknown 6 18:04

Lane 3

802 1,814 78 17:15 unknown unknown 48 19:19

Lane 4

1,243 1,574 166 16:14 unknown unknown 152 17:43

Lane 5 N/A 948 N/A N/A N/A unknown N/A N/A

Total 5,096 5,823 641 N/A 3,157 1,663 208 N/A

Figure 3: Comparison of the average ALPR and video travel times over five minute bins

Figure 4: Comparison of the average Custom Bluetooth and video travel times over five minute bins

Figure 2: ALPR, Bluetooth, and Baseline comparison travel time plot from April 13th 2013

Lane Bias IssueALPR73% of all ALPR matches were detected in Lane 4 (the outside lane

adjacent to equipment) during October 10th, 2012 data collection.Vehicles in lanes further from equipment tended to be occluded.

RADARSimilar to ALPR, high travel time spikes suggest primary vehicle

detection in lane closest to equipment.Low profile may cause it to primarily sense vehicles in the closest

lanes.

Randall Guensler (PI), Michael Hunter (Co-PI), Angshuman Guin, Wonho Suh, James Anderson, Kathryn Colberg, and Stephanie Zinner

Infrared image from ALPR camera used in license plate recognition process.