基于强化学习算法原则的出行选择行为建模与仿真

Modeling and Simulation of Travel Choice Behavior based on Reinforcement Learning Principles

针对目前高峰时段交通拥堵的问题,基于强化学习算法原则,提出了一种对出行者出行选择行为建模及仿真的方法.首先对出行者的认知更新过程进行建模,然后利用Logit模型来描述出行者的决策过程,最后采用MATLAB软件仿真了在不同小汽车出行成本费用下,出行者出行时间和出行方式选择的变化规律.仿真结果表明:随着小汽车出行成本的增加,部分小汽车出行者逐渐选择公共交通出行,并且当出行成本增加到15元以上时,出行方式选择变化趋于平缓;同时,出行者倾向于提前出行且高峰时段出行需求有所减少,进而有利于减缓高峰拥堵状况.

Aiming at the issue of traffic congestion during peak hours,a method for modeling and simulating the travel choice behavior of commuters is proposed.The process of perception updating is modeled based on principles of reinforcement learning,and Logit model is used to describe the decision-making process of the commuters.A simulation on the choices of departure time and mode under different car travel cost is conducted using MATLAB.The simulation results indicate that most car commuters gradually choose public transport,and mode choice changes quickly with the increase of car travel cost until the car travel cost is over 15 yuan.Meanwhile,commuters are inclined to depart earlier than usual,and travel demand during peak hours is decreased,which releases peak hour congestion.