一种改进的深度确定性策略梯度网络交通信号控制系统

Improved deep deterministic policy gradient network traffic signal control system

摘要:

交通信号系统控制着城市车辆运行秩序,其效率高低直接影响了社会经济的发展.以十字路口的交通信号控制系统为研究对象, 基于深度确定性策略梯度网络DDPG提出了一种改进算法. 结合交通环境的特点设计了特征增强和样本去重算法提高算法的性能. 通过对实际交通系统运行情况进行调研,基于SUMO仿真环境搭建了交叉路口交通仿真平台. 利用FEPG算法控制交通信号,实现了车辆的高效通行. 实验结果表明,该算法能够有效地降低车辆等待时间, 减少车辆的污染排放.

Abstract:

The traffic signal system controls the order of urban vehicles, and its efficiency directly affects the development of the social economy. This paper takes the traffic signal control system at the intersection as the research object, an improved algorithm is introduced based on the deep deterministic policy gradient (DDPG), in which the feature enhancement and sample deduplication algorithms combined with the characteristics of the traffic environment are designed to improve the performance of the algorithm. By analyzing the operation of the actual traffic system, a traffic simulation platform for intersections is set up based on the SUMO simulation environment. The FEPG algorithm is used to control traffic signals to achieve efficient vehicle traffic. Experimental results show that the algorithm can effectively reduce vehicle waiting time and reduce vehicle emissions.