Abstract:

Aming at the problem of lower coordination efficiency level and the dimension disaster of traditional distributed adaptive traffic signal control, a model of urban regional traffic signal control system is established which is formulated as a local cooperation game of signalized intersections. On this basis, the learning algorithm based on the local information game exchange is proposed. It can adjust the control policy independently so that it can gradually learn the optimal traffic signal control policy. The performance evaluation index is constructed by weighting the average delay and parking times of the road network. The simulation experiment shows that the game-learning achieves better control effect and convergence compared with the actuated control and genetic algorithm in different traffic demand cases. It has better ability to manage and control overflow.

Key words: traffic engineering, coordinated control, game, learning, traffic signal

摘要：

针对传统分布式自适应交通信号控制协调效率受限，并且存在维数灾难问题，建立了城市区域交通信号控制系统模型，将其优化问题建模为局部交叉口交通信号博弈协调控制，提出基于交叉口交通信号控制agent局部信息博弈交互的学习算法。在学习过程中交叉口交通信号控制agent进行局部信息博弈交互，自主调整交通信号控制策略使其逐步学习到最优策略。通过设计不同的交通需求情景，对路网平均延误和平均停车次数进行加权构建性能评价指标，相对于遗传算法和感应控制方法，博弈学习取得更好的交通信号控制效果，其能收敛到最优性能评价指标，其具有更好的交通需求管控能力。

关键词: 交通工程, 协调控制, 博弈, 学习, 交通信号