摘要
为实现通勤走廊内机动公交接驳方案的科学设计,将其韧性恢复过程视为复杂环境中接驳方案经过探索迭代实现韧性提升的双层规划。引入深度强化学习算法构建上层规划,采用价值函数神经网络拟合突发事件与出行者集群行为对接驳方案调整的反应函数,训练接驳方案达到决策目标。下层规划运用元胞神经网络模拟数据智能背景下的集群出行行为。实例研究表明,方法可以使通勤走廊韧性得到有效提升,而集群行为会对韧性恢复产生负面影响。
In order to realize the scientific design of motor bus transferring scheme in the commuter corridor,the resilience recovery process of commuting corridor is regarded as a bi-level programming in which the resilience is improved through continuous exploration and iteration of ground bus transferring scheme in complex environment.The deep reinforcement learning algorithm is introduced to form the upper level planning,and the value function neural network is used to fit the response function of emergencies and travelers’cluster behavior to the adjustment of ground bus transferring scheme.The decision-making objective is achieved by training the transferring schemes.In the lower level planning,the cellular neural network model is introduced to simulate the cluster travel choice behavior under the background of data intelligence.The case study shows that this method can effectively improve the resilience of the commuter corridor,and the cluster behavior will have a negative impact on the resilience recovery.
作者
李雪岩
张同宇
祝歆
LI Xueyan;ZHANG Tongyu;ZHU Xin(School of Management,Beijing Union University,Beijing 100101,China;School of Urban Rail Transit and Logistics,Beijing Union University,Beijing 100101,China)
出处
《复杂系统与复杂性科学》
CAS
CSCD
北大核心
2024年第1期92-99,108,共9页
Complex Systems and Complexity Science
基金
北京市社会科学基金(21GLC046)。
关键词
通勤走廊
韧性
接驳方案
深度强化学习
集群行为
commuting corridor
resilience
transferring scheme
deep reinforcement learning
cluster behavior
