Public transportation scheduling problem is a complex problem considering various influencing factors (such as passenger flow, etc.). However, the effect is not ideal when using the traditional machine learning method. Therefore,this paper aims to propose a Deep Reinforcement Learning Proximal Policy Optimization (DRL-PPO) algorithm for public transport scheduling according to different actual passenger flow. At the end of the paper, we make bus arrival time records and passenger OD data of Guangzhou route 560, for three months from July 1, 2019, to September 30, 2019 for training. Compared with GA algorithm, PPO algorithm is better than GA in convergence speed and final effect, which verifies the effectiveness of PPO in bus scheduling with complex passenger flow characteristics.
Deep Reinforcement Learning for Bus Scheduling Considering Complex Passenger Flow
22nd COTA International Conference of Transportation Professionals ; 2022 ; Changsha, Hunan Province, China
CICTP 2022 ; 326-335
2022-09-08
Conference paper
Electronic Resource
English
Passenger delay minimization signal control method based on deep reinforcement learning
European Patent Office | 2023
|