Improved vehicle reidentification and travel time measurement on congested freeways

This paper presents an improved algorithm for matching individual vehicle measurements at a freeway detector,station with the vehicles' corresponding measurements taken at another detector station located upstream. The improvements consist of four new tests to identify incorrect matches and either replace them with the correct matches or discard the errors. The result is a higher number of reidentified vehicles with a lower frequency of errors compared to our earlier work. Although this algorithm is potentially compatible with many vehicle detector technologies, the paper illustrates the method using existing dual loop detectors to measure vehicle lengths. Due to the limited accuracy of the existing loop detector measurements under free flow conditions, this presentation is restricted to matching vehicles during congested traffic conditions. The algorithm is applied to individual lanes and exploits information about the sequence that vehicles pass the detectors. As such, the effect of vehicles changing lanes between stations is also considered in this paper. Of course once a vehicle has been matched across neighboring detector stations, the difference in its arrival time at each station defines the vehicle's travel time on the intervening segment. Thus, the algorithm extracts travel time data without requiring the deployment of new detector technologies. The approach described herein has been extensively tested on I-80 in the Berkeley Highway Laboratory (BHL). This paper presents some of the results obtained over two different segments of the BHL and demonstrates that the algorithm matches a sufficient number of vehicles for the purpose of travel time measurement.