Traffic light optimization with low penetration rate vehicle trajectory data

Traffic light optimization is a cost-effective method for reducing congestion and energy consumption. This paper presents a large-scale traffic signal re-timing system using a small percentage of vehicle trajectories as input, eliminating the need for vehicle detectors. A probabilistic time-space diagram connects a stochastic point-queue model with vehicle trajectories, enabling reconstruction of recurrent traffic states. Optimization algorithms update traffic signal parameters, and real-world testing in Birmingham, Michigan demonstrated up to 20% delay and 30% stop reduction. This system offers a scalable and efficient solution for traffic light optimization.

Xingmin Wang, Zachary Jerome, Zihao Wang, Chenhao Zhang, Shengyin Shen, Vivek Vijaya Kumar, Fan Bai, Paul Krajewski, Danielle Deneau, Ahmad Jawad, Rachel Jones, Gary Piotrowicz, Henry X. Liu