Abstract: A Chaotic Fuzzy Q-Learning（C-FQL） method which is used to solve the problem of urban area traffic coordinated control is put forward. It adds chaos disturbance into the fuzzy Q-learning to improve the traditional way of agent choosing action, and embeds the forgetting factor to balance the relationship between exploration and utilization in fuzzy Q-learning. It applies this algorithm to urban area traffic coordinated control to optimize the cycle length, split, offset of each signalized intersection. This paper builds a classic urban traffic network and makes simulation based on the traffic simulation platform of TSIS. The results of simulation show that the method provided can greatly improve the whole efficiency of area traffic.

Key words: area traffic control, Q-learning, chaotic variable, fuzzy control

摘要： 提出了一种解决城市区域交通协调控制问题的混沌模糊Q学习（C-FQL）方法。在模糊Q学习的过程中添加混沌扰动，以改进传统的Agent选择动作的方式，并通过遗忘因子以平衡模糊Q学习中探索和利用之间的关系。将该算法应用于城市区域交通协调控制中优化各信号交叉口的周期、绿信比和相位差。利用TSIS交通仿真平台，建立典型的城市区域交通网络并进行仿真。仿真结果表明该方法可以大大提高区域交通的整体效率。

关键词: 区域交通控制, Q学习, 混沌变量, 模糊控制