考虑出行者偏好和经验的路径选择行为研究  

Study on Route Choice Behavior Considering Traveler's Preference and Experience

作　　者：杜玲丽[1] 胡骥[1] 赵怀明[2] 刘海旭[1] 蒯佳婷 DU Ling-li;HU Ji;ZHAO Huai-ming;LIU Hai-xu;KUAI Jia-ting(School of Transportation and Logistics,Southwest Jiaotong University,Chengdu Sichuan 610031 ,China;China Railway Eryuan Engineering Group Co. ,Ltd. ,Chengdu Sichuan 610031 ,China)

机构地区：[1]西南交通大学交通运输与物流学院,四川成都610031 [2]中铁二院工程集团有限责任公司,四川成都610031

出　　处：《公路交通科技》2019年第5期138-144,151,共8页Journal of Highway and Transportation Research and Development

基　　金：国家自然科学基金项目(51278429)

摘　　要：出行者的路径选择行为是包括自身特性在内多种因素共同作用的结果。为分析出行者偏好对路径选择行为的影响,首先假设出行者从路网中获取的信息为不完全历史信息,建立了理解行程时间及其更新模型,然后给出了基于经验-加权吸引力(EWA)学习模型和累计强化学习模型的两种偏好动态更新规则,最后通过Dogit模型将理解行程时间和路径偏好共同纳入出行者的路径选择决策中。在此基础上,对比分析了不考虑路径偏好、路径偏好为固定值、基于EWA学习模型更新路径偏好和基于累计强化学习模型更新路径偏好4种不同偏好情况下网络交通流的演化情况。算例结果表明:相比利用Logit模型不考虑路径偏好的流量分配结果,利用Dogit模型考虑路径偏好的流量分配结果更为均衡,且在考虑偏好时,路径偏好为固定值、基于EWA学习模型更新路径偏好和基于累计强化学习模型更新路径偏好3种情况下路径的均衡流量间差异较小;偏好动态更新时,基于EWA学习模型的路径偏好动态更新规则较累计强化学习模型能更好地捕捉出行个体的路径偏好,但由累计强化学习模型得到的路网流量分配结果更为均衡;偏好为固定值时,路径的均衡流量介于EWA学习模型和累计强化学习模型两种偏好动态更新规则下路径的均衡流量之间。Travelers route choice behavior is the result of multiple factors including their own characteristics. In order to analyze the influence of travelers preference on route choice behavior, first, assuming that the information obtained from the road network is incomplete historical information, the understanding travel time model and its update model are established. Then, 2 dynamical updating rules based on the preference of experience-weighted attraction ( EWA) learning model and the cumulative reinforcement learning model are given. Finally, through the Dogit model, understanding travel time and route preference are integrated into travelers route choice decision. On this basis, the evolutions of network traffic flow under 4 preferences (without considering route preference, route preference as a fixed value, updating route preference based on EWA learning model, and updating route preference based on cumulative reinforcement learning model) are compared and analyzed. The result of the example shows that (1) compared with the traffic distribution using the Logit model without considering the path preference, the traffic distribution using the Dogit model considering path preference is more balanced, and there is little difference among the route equilibrium traffic volumes under the route preference as a fixed value, the updating route preference based on EWA learning model and cumulative reinforcement learning model;(2) when dynamical updating the preference, the route preference dynamical update rule based on the EWA learning model can capture travelers individual route preference better than the cumulative reinforcement learning model, but the road network traffic distribution result obtained by the cumulative reinforcement learning model is more balanced;(3 ) when the preference is a fixed value, the route equilibrium volume is between those under the rules of the EWA learning model and the cumulative reinforcement learning model.

关 键 词：城市交通 路径偏好 Dogit模型 路径选择行为 EWA学习 

分 类 号：U491.17[交通运输工程—交通运输规划与管理]