基于交叉口多维状态评估的信号配时优化研究

Research on signal timing optimization based on intersection multi-dimensional state evaluation

交叉口运行状态的准确评估能够为交通管理系统提供可量化的交叉口信息,为优化交叉口信号控制提供依据。针对当前交叉口运行状态评估侧重行车感受,客观性不强,确定了结合机动车和行人需求的多维评估指标体系,利用AHP-变异系数双层集成赋权模型确定各指标权重,得到交叉口运行评估模型。基于评估模型采用含噪声的深度Q学习(NoisyNet DQN)算法进行信号配时优化研究。以合肥市某交叉口为例,基于交通仿真软件证明了该方法能有效减缓交通拥堵,提高行人感受,具备较高的应用拓展性。

The accurate evaluation of intersection operation status can provide quantifiable intersection information for traffic management system and provide basis for optimizing intersection signal control.In view of the current intersection operation status evaluation focuses on driving experience and is not objective,this paper determines multi-dimensional evaluation index systems combined with the needs of motor vehicles and pedestrians,and uses the AHP variation coefficient double-layer integrated weighting model to determine the weight of each index,so as to obtain the intersection operation evaluation model.Based on the evaluation model,a NoisyNet deep Q-learning reinforcement learning algorithm is used to optimize the signal timing.Taking an intersection in Hefei as an example,based on traffic simulation software,it is proved that this method can effectively alleviate traffic congestion,improve pedestrian feeling,and has high application expansibility.