Collaborative optimization of last-train timetables for metro network to increase service time for passengers. (March 2023)

• Record Type: Journal Article
• Authors: Wang, Fangsheng; Xu, Ruihua; Song, Xuyang; Wang, Pengling
• Is Part Of: Computers & operations research. Volume 151(2023)
• Publication Date: 2023-03
• Subjects: Last-train timetable -- Collaborative optimization -- Latest departure time -- Genetic algorithm -- Q-learning; Operations research -- Periodicals; Electronic digital computers -- Periodicals; 004.05
• DOI: 10.1016/j.cor.2022.106091
• Languages: English
• ISSNs: 0305-0548

Abstract

Highlights: Passengers' latest time to use metro service is optimized by timetable coordination. An improved genetic algorithm based on Q-learning is developed to solve the problem. Analyses with benchmarks prove more passengers can be served with optimized solutions. Abstract: Last-train timetables are significant in metro systems and directly influence transportation organization efficiency and passenger service levels. This study focuses on the collaborative optimization of last-train timetables for a large-scale metro network. Different from most studies, which often focus on improving the ability of passengers to transfer between different metro lines (transfer accessibility) during the last-train period, or the ability of passengers to reach their destinations after boarding the last train service from the origin station (origin-destination (OD) accessibility), this work aims to optimize the latest time for passengers (LTP) to reach their destinations using the metro services. A mixed-integer linear programming (MILP) model is established to optimize the last-train timetables with maximizing LTPs. For comparison, two MILP models respectively aim at maximizing transfer accessibilities and OD accessibilities, are adopted as benchmarks. An improved genetic algorithm based on Q-learning (QGA) is developed to solve the proposed MILP models for optimizing last-train timetables for a large-scale metro network. The proposed method is validated by optimizing the last-train timetable of the metro network of Chengdu, China. The results indicate that compared with optimizing the transfer and OD accessibilities, optimizing LTPs can consider both single-line and multiple-line passenger benefits, and directly increase their accessibilities and feasible times to use the metro service.