Incident detection algorithm development on signalized urban arterial streets

Congestion due to incidents on urban freeways and arterial streets is an important issue in urban mobility. It is also an area with potential for successful implementation of ITS (Intelligent Transportation Systems) technology. For many years, traffic engineers have studied techniques to detect an incident on urban freeways. However, since the traffic flow pattern on arterial streets is more dynamic than on freeways, algorithms to detect incidents on arterials have been more difficult to define.; Existing freeway and signalized arterial street incident detection algorithms have been investigated to determine the merits of each in their potential use on urban arterial streets. Based on the literature review, a Kalman Filtering Predictor Algorithm was modified and used to recursively filter and update aggregate traffic control variables and to eventually determine an incident state.; A test using non-incident arterial street data showed good tracking ability of selected traffic variables over time. A second test, using data from an incident on an arterial street, confirmed that the algorithm developed in this study has good potential for arterial street incident detection.; A simulation using TRAF-NETSIM was then conducted to test the algorithm for incident-free and various incident states. The algorithm performance measures were a false alarm rate (FAR) of 0.319 percent per hour per station, a detection rate (DR) of 100 percent with a mean detection time (MDT) of 4.93 minutes for one-lane blockages. For the high traffic volume direction, the FAR was reduced to 0.111 percent per hour per station. The MDT varied from 4.8 minutes for long links to 2.7 minutes for short links.; A discussion of the algorithm and recommendations for further research to improve incident detection on urban arterial streets is also included in this report.