Abstract

Development of advanced technologies for real-time road traffic congestion detection is required by advanced traffic management systems (ATMS) and advanced traveler information systems (ATIS). However, provision of dynamic intra-lane and inter-lane traffic information such as queuing and lane-changing remains incomplete in emerging technologies. This paper is intended to introduce a prototype of a new framework capable of real-time detection of incident and non-incident congestion on freeways. It employs dynamic estimates of lane traffic characteristics including queue lengths in blocked lanes tin the case of lane-blocking incidents), the number of vehicle in each lane, and mandatory lane-changing fractions in lanes with traffic congestion for the use of real-time road traffic congestion detection. On-line lane traffic count and occupancy data collected from point detectors are used as the major input to the proposed framework. The framework is founded on the basis of 1) nonlinear stochastic system modeling and estimation which involves the use of an extended Kalman filter and 2) the modified sequential probability ratio test technology (MSPRT). Preliminary tests had been conducted, indicating the feasibility of employing the proposed framework for the use in real-time incident detection on freeways. Further tasks will include tests for the case of non-incident congestion. The research presented here may help stimulate research in related areas such as incident management systems, automatic vehicle tracking and monitoring systems, and automatic road congestion warning systems for further use in ATMS and ATIS.