基于强化学习的高速铁路列车运行调整方法研究  

Research on High-Speed Railway Timetable Rescheduling Based on Reinforcement Learning

作　　者：杜心怡 邵长虹 Du Xinyi;Shao Changhong(Key Laboratory of Railway Industry for Operational Active Safety Assurance and Risk Prevention and Control,Beijing Jiaotong University,Beijing 100044,China;School of Traffic and Transportation,Beijing Jiaotong University,Beijing 100044,China;Beijing-Shanghai High Speed Railway Co.,Ltd,Beijing 100038,China)

机构地区：[1]北京交通大学运营主动安全保障与风险防控铁路行业重点实验室,北京100044 [2]北京交通大学交通运输学院,北京100044 [3]京沪高速铁路股份有限公司,北京100038

出　　处：《铁道技术标准（中英文）》2024年第6期35-43,共9页Railway Technical Standard(Chinese & English)

基　　金：国家重点研发计划(2022YFB4300603);中央高校基本科研业务费(2023JBZY005)。

摘　　要：我国高速铁路发展迅速,路网规模扩大、列车数量增加以及列车速度提升使得列车间运行关系耦合性强,运行调整难度大。为解决上述问题,本文构建了以调整后的列车时刻表与计划时刻表偏差最小为目标的列车运行调整模型,设计了交互环境、智能体、状态、动作、奖励函数等要素,采用强化学习方法中经典的Q-learning算法进行求解,最后通过算例对本文所提出的模型和算法的有效性进行验证。结果表明,该算法的求解质量比先到先发(FCFS)减少了43.87%的总偏差时间。The rapid expansion of China's high-speed railway network,coupled with an increase in the number and speed of trains,has significantly strengthened the interdependence of trains,presenting challenges in adjusting operations.To address these issues,this paper introduces a train operation adjustment model focused on minimizing deviations between adjusted train timetable and planned timetable.It encompasses the design of an interactive environment,along with components such as agents,states,actions,reward functions,and more.The study employs the classical Q-learning algorithm within a reinforcement learning framework to resolve the aforementioned operational challenges.Finally,an illustrative example is provided to validate the effectiveness of both the proposed model and algorithm.The findings reveal that the algorithm's solution is 43.87%lower in total deviation than the First-Come-First-Serve(FCFS)approach.

关 键 词：高速铁路 列车运行调整 调度 强化学习 Q-LEARNING 

分 类 号：U292.4[交通运输工程—交通运输规划与管理]