In this paper we present a simple framework for matching vehicles in congestion using data available from a sequence of existing loop detector speed traps. Most loop based travel time estimates only use local features (flow, occupancy, average speed) to infer travel time. Such strategies can nog quantify conditions beyond the detection regions. In particular, they can only respond to disturbances between detectors after the effects propagate to one or both detectors and have been correctly identified. By matching vehicles that pass a sequence of speed traps we can estimate travel times, independent of unknown conditions between the speed traps. The scope of this paper is to illustrate the principles of extracting identifiable features at successive detection one mile apart. Techniques for automating the process are not addressed in this paper; they include statistical methods, digital signal processing and artificial intelligence. (A)
Abstract