A simple detection scheme for delay-inducing freeway incidents

This paper describes a freeway incident detection scheme that does not rely on complicated theories. The procedure compares the occupancy information recorded by two neighboring loop detectors to determine whether an incident has occurred in the intervening segment, using a recipe that is directly related to an intrinsic property of delay-inducing incidents. The method, which can also signal the termination of a detected incident, assumes that any significant bottlenecks are located outside the segment in question: i.e. that the segment is "homogeneous". Independent of complicated theories, the proposed method can be applied with little calibration to any site despite the presence of detector errors and statistical fluctuations. It is also possible to use site-independent (default) parameters, although one should then expect a somewhat inferior performance. Tests with five sets of 30 s data from three different sites in the United States, Canada, and the Netherlands, using the recommended default parameters, were encouraging. The resulting graphical displays included in this paper clearly illustrate the scheme's effectiveness in distinguishing non-recurrent from recurrent congestion.