Development and Testing of Field-Deployable Real-Time Laser-Based Non-Intrusive Detection System for Measurement of True Travel Time on the Highway

Problem

Reliable, real-time traffic speed information is invaluable for traffic planners. Current methods for measuring travel time do so by measuring speed at one or more points along a link and then extrapolating speed across the entire length of the link. Loops, radar, and video image processing all use extrapolation to deduct travel time. However, when traffic flow slows because of an accident, speed variations along any given link can be quite large. This is when accurate information for making routing decisions is most critical, and the uncertainties caused by extrapolation are magnified.

Vehicle-as-Probe (VAP) methods determine travel time by identifying vehicles at the start of the link and then re-identifying them at the end of the same link, giving true travel time as the time difference between the two. The cost of VAPs is quite high, however, requiring a large number of vehicle tags and tag readers to be effective.

It is possible to identify some distinguishing characteristic on a vehicle at the beginning of a link and then re-identify that same vehicle at the end of a link without the set-up costs of a VAP. If a characteristic can be found that separates cars into 100 possible classifications, a match between upstream and downstream vehicles can be made with high probability. Even more classifications would make the process even more accurate.

Method

This project developed a laser detection system that can directly and accurately determine origin and destination information for vehicles in the traffic stream non-intrusively and without violating privacy. It produces local vehicle speed, vehicle volume, and vehicle classifications, even under high flow conditions. Point-to-point travel time and incident detection can easily be determined.

This system can be mounted above the road, which allows it to be installed and maintained without tearing up the road or disrupting traffic. It operates on a simple "on/off" basis, requiring much less computation for vehicle detection than video methods. It also produces and senses its own signal, so it does not depend on time of day and weather conditions for accuracy (as do video detection systems). There are no moving parts, and power and bandwidth requirements are very low.

The system projects a laser onto the ground, and then collects the reflected light onto a photodiode array. A vehicle is detected when it passes under the beam, based on the absence of the reflected light. With two identical laser/sensor pairs placed close together, it is possible to measure the speed of both the front and the rear of the vehicle as it triggers each sensor. The length of each vehicle is determined from the speed and the amount of time the vehicle spends under the detector.

Findings

We have built a field prototype detection system and tested it under real traffic conditions. Our test results verify that the principle of this system is technically sound and the software works in most cases. There are a few situations, such as stop-and-go traffic, where the detection of vehicle length does not work well.

We also found that the separating distance of the two laser-sensor detectors was too small in our test, which made the system detect different speeds for the front and the rear of vehicles. Also, the slowness of the analog signal contributed to the error. Both of these problems will be fixed in the next prototype, by putting the pairs farther apart and by using a faster digital signal.