关键词: 云模型, 交通信号控制, 前件云发生器, Q-学习

Abstract: In order to reduce the delay of cars passing through intersections, control strategies are set up by cloud model and some parameters of the control model are improved by Q-learning method. The signal control agent of intersections gets the information of vehicles arriving through its perception system. The current signal state is changed according to the set of signal control rules and the information of vehicles arriving. The key rule of the set of signal control rules is described by two-dimensional half normal clouds. The signal control strategies facing different traffic states are generated by two-dimensional forward cloud generators. The main parameters of cloud model are optimized by Q-learning algorithm. Regarding the total vehicle stop delay as objective function, the optimal control schemes based on the cloud model under different traffic volume are generated by iteration. Finally, the signal control method based on the cloud model is compared with the traditional approaches through simulation. The result of the simulation indicates that the effect of the control method based on the cloud model is better than the fixed and actuated control approaches.

Key words: cloud model, traffic signal control, forward cloud generator, Q-learning