基于Q-学习和粒子群算法的区域交通控制模型

Area Traffic Control Model Based on Q-Learning and PSO

针对城市交通系统的动态性和不确定性,提出了基于Q-学习和粒子群算法相位差优化算法,对区域交通动态实时控制进行了研究。根据不同的交通流情况确定不同的区域控制目标函数,将Q-学习的奖惩机制引入粒子群算法的选优过程中,通过改进的粒子群算法实时优化区域控制策略。编制该控制方法的仿真程序,应用AIMSUN仿真软件验证算法的控制效果。结果表明,该方法对不同交通量下可保持较高的控制效率,控制效果明显优于感应控制。

Considering the dynamics and uncertainty in urban transportation system, the traffic signal offset optimization algorithm was proposed based on Q-learning and Particle Swarm algorithm, and the real-time dynamic area traffic control was studied. According to different traffic status, the system applied different area control objective function, introduced reward mechanism of Q-learning into the optimization process of PSO algorithm, optimized area traffic control strategy by improved PSO in real-time. Programming the simulation program of this control model, AIMSUN simulation software was used to validate the control effect. The simulation result shows the proposed method has high control efficiency in different traffic scenarios, and is obviously better than the traditional ones.