Logic-based incident detection on signalized streets with heterogeneous data

This paper presents a new logic-based method of detecting incidents on signalized streets. The concept of the method is based on the fact that incidents, when reducing capacity of intersections, change the sequence of congested sections (congestion patterns) along an arterial street. Rules are derived that use discrepancies between expected and actual congestion patterns to detect the presence of incidents and to identify their location. The feasibility of the method was checked with computer simulation. The following features of the proposed method distinguish it from the existing methods: (1) any data that carries congestion information can be used; (2) all the rules applied to infer incident information are rigorously derived from the fundamental principles of traffic operations and not from subjective human knowledge; (3) the method estimates the range of possible values of capacities affected by the incident; and (4) the method is able to detect incident clearance. This paper presents a novel idea of detecting incidents rather than a fully verified method.