交通信息与安全2017，Vol.35Issue(4)：52-62,11.DOI:10.3963/j.issn.1674-4861.2017.04.007
基于强化学习的限行政策下双模式出发时间选择仿真研究
A Simulation Study of Departure Time Selection in Dual-modal with Impacts of Vehicle Restriction Policies Based on Reinforcement Learning
摘要
Abstract
Combining multi-agent technology with a reinforcement learning model, a computer simulation model of travel modes and choice of departure time of commuters in peak hours is established.In a simulation, travel choice behaviors of commuters are studied with the consideration of impacts of vehicle restriction policies, and the formation of commuting equilibrium in peak periods is also reproduced.Based on simulation results, the effects of different measures for improving public transportations are analyzed.The results show that the number of commuters by bus increases by 18% after the implementation of restriction policies, which eases congestions in peak periods to a certain extent.Meanwhile, the probabilities that commuters travel by bus in unrestricted days become smaller, which means the effects of adopting restriction policies exclusively are fairly limited.Under the influences of restriction policies, if departure frequencies of public transport increase, the number of commuters travel by bus increases by 17.5%, and drivers′ waiting time in congestions decreases by 85%, which can effectively improve the traffic situations.Compared with that, reducing ticket price of public transport is less effective.The multi-agent approach applied in this study shows the richness in individual behaviors which can be realized intuitively and conveniently.It also has advantages in describing interactions between individuals and traffic systems, which provides an effective way to explore formation and evolution of complicated traffic phenomena.关键词
城市交通/通勤/多Agent仿真/强化学习Key words
urban transport/commute/multi-Agent simulation/reinforcement learning分类
交通工程引用本文复制引用
吴学新,凌帅,李庚..基于强化学习的限行政策下双模式出发时间选择仿真研究[J].交通信息与安全,2017,35(4):52-62,11.基金项目
国家自然科学基金项目(71431005)资助 （71431005）
