交叉口Agent间的博弈学习协调方法

Method of Intersection Agent Coordination Based on Game-learning

将Multi-agent应用在城市交叉口交通协调控制中。通过引入交通信号控制Agent,分析Agent的协调过程;以交通信号控制Agent为角色,建立了相互关联的城市交叉口资源动态协调配置抽象模型;应用对策论作为协调实现途径,并且以分布式Q强化学习中Q值更新作为其效用函数。通过对两交叉口协调实例分析,证明了采用博弈学习协调方法的有效性。

Coordination based on Muli-agent constitutes a focus of open,distributed,complicated systems. In this paper, this technique was applied to urban intersection traffic coordination control. The coordination mechanism of traffic signal control Agent was analyzed. The abstract dynamic model of coordinated distribution of resources about interactive intersections was put forward used traffic signal control Agent as a role. The method of Game Theory was applied to realize the coordination ,and the renewed Q-value in the distributed reinforcement Q-learning was used to build the payoff values in the method of game theory. By comparing the method with and without coordination by an experiment of two intersections,the result indicates that the coordination method based on Game-learning is effective.