A Train Cooperative Operation Optimization Method Considering Passenger Comfort based on Reinforcement Learning
Authors:
- 3 August 2022
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 368-373
Abstract
This paper mainly focuses on the high-speed train cooperative operation problem, considering passenger comfort. To solve this problem, a speed curve optimization method based on Reinforcement Learning algorithm is proposed. First, according to the train dynamics system, we build the speed curve optimization object. For realizing the train cooperative operation and increasing the passenger comfort, trains distance constraint and jerk rate constraint are fully considered in formulating reward function. And then, agent of Twin Delayed Deep Deterministic Policy Gradient (TD3) is established. Based on TD3 algorithm, two groups of Q-value networks and one group of action action networks have the same structure, different initial values and update rates respectively, which improves the stability and convergence speed of the algorithm. After the training. The Agent can generate a desirable speed curve of train based on constraints of vehicle output and jerk rate under cooperative operation.Keywords
Funding Information
- National Natural Science Foundation of China (U1934221,U21A20169)
- Sichuan Science and Technology Program (2021JDJQ0012,2020YFQ0057)
This publication has 6 references indexed in Scilit:
- A Pareto Front Transformation Model for Multi- Objective-Based Constrained OptimizationIEEE Access, 2020
- A Model Predictive Control Approach for Virtual Coupling in RailwaysIEEE Transactions on Intelligent Transportation Systems, 2019
- Applications of Deep Reinforcement Learning in Communications and Networking: A SurveyIEEE Communications Surveys & Tutorials, 2019
- Distributed cooperative control of multiple high-speed trains under a moving block system by nonlinear mapping-based feedbackScience China Information Sciences, 2018
- Coordinated cruise control for high-speed train movements based on a multi-agent modelTransportation Research Part C: Emerging Technologies, 2015
- Learning to predict by the methods of temporal differencesMachine Learning, 1988