基于Q学习算法的两交叉口信号灯博弈协调控制

Q学习和博弈论相结合解决相邻两交叉口信号灯协调控制问题。在基本Q学习算法的基础上引入博弈论,以Q值作为赢得函数建立赢得矩阵。相邻两交叉口之间的协调关系属于二人非零和合作博弈,采用Nash公理方法求得其谈判解,并以此作为Q学习策略选择的依据实现两交叉口协调控制。应用Paramics交通仿真软件进行算法仿真,结果表明该方法的有效性。

Game Coordination Control of Two Intersections Based on Q-Learning Algorithm

Traffic signal coordination control for two adjacent intersections was studied by hybrid Q-learning and game theory. Based on the general Q-learning algorithm,the game theory was introduced and the payoff function of Q-values was used to build the payoff matrix. The problem of coordination control for two adjacent intersections is two-player Non-zero-sum cooperative game in Game theory. From Nash Negotiation solution by Nash Theorem,the result of the game which was the basis for making decision of Q-learning was obtained. Furthermore,coordination control for two adjacent intersections based on Q-learning and game theory was realized and illustrated by Paramics simulation software. The validity of this method is proved.