The 20th and 21st joint COTA International Conference of Transportation Professionals (CICTP 2020-21)

Traditional traffic control algorithms have been focused on isolated road intersections or regional road intersections. Based on the estimated interaction of two adjacent intersections, this paper proposes a coordination algorithm for two traffic intersections with the smallest queue length difference by using the reinforcement learning algorithm. The algorithm first optimizes the green signal ratio of the two intersections, and then optimizes the phase difference between the two intersections, which achieves a good control effect compared with the traditional timing control. Compared with other reinforcement learning algorithms, the learning efficiency of the method is better. high. Simulations with synthetic data indicate that the traffic controller based on the proposed algorithm can improve the traffic conditions at both intersections.

CICTP