基于MA2C方法优化和Sumo仿真实时推演的信号控制策略研究

Research on Signal Control Strategy Based on MA2C Method Optimization andSumo Simulation Real-time Inference

当前,交管部门多依赖人工经验解决城市交通车流拥堵问题,然而此种方法缺乏数学理论支持与实践验证,并且无法从全局性和整体性的角度提供最优方案。本文基于Sumo交通仿真,采用MA2C多智能体强化学习模型,制定具有针对性的信号灯配时强化学习规则,经多轮训练,获得对应车流运输效率最高的信号灯配时。同时,将Sumo仿真建模和数据处理过程进行了代码化与模块化,以提升运转效率,最终实现交通信号灯配时优化与过程的全自动化。

Currently,traffic management departments rely heavily on manual experience to solve urban traffic congestion problems.However,this method lacks mathematical theoretical support and practical verification,and cannot provide optimal solutions from a global and holistic perspective.This article is based on Sumo traffic simulation and uses the MA2C multi-agent reinforcement learning model to develop targeted signal timing reinforcement learning rules.After multiple rounds of training,the signal timing with the highest corresponding traffic transportation efficiency is obtained.At the same time,Sumo simulation modeling and data processing processes were coded and modularized to improve operational efficiency,ultimately achieving full automation of traffic signal timing optimization and processes.