Automatic Generation of Traffic Signal Timing Plan
Henry Liu, Jianfeng Zheng
Report no. MnDOT 2014-38
Due to budget constraints, most of the traffic signals in the US are retimed once every 2-5 years. Despite that, traffic delay increases 3-5% per year with outdated timing plans. It would be desirable to reduce the signal retiming costs by automating all or a portion of the manual process. This research takes one step forward in this direction. In this project, we develop a performance visualization and fine-tuning tool for arterial traffic signal systems, aimed at reducing the labor costs for signal retiming. Using high-resolution event-based data from the SMART-Signal system, a set of easy-to-use algorithms are developed to refine traffic signal systems. Specifically, a framework is developed to diagnose operational problems regarding cycle lengths, green splits and offsets. Then, algorithms for offsets and green splits fine-tuning are proposed. To fine-tune offsets, a practical procedure to construct time space diagram (TS-Diagram) to visualize the progression quality on arterials is proposed and validated. For green splits, an adjusted measure of effectiveness (MOE), the utilized green time (UGT), is proposed for performance evaluation. Moreover, a practical procedure for time of day (TOD) transitions is also developed to generate optimal timing plan schedules. Field case studies and simulation experiments are carried out to illustrate and validate the proposed algorithms. The algorithms could be used during the retiming process to help agencies reduce labor costs, or to periodically refine traffic signal systems for coordinated arterials.