| Presentation | 2022-12-16 Learning of train control measures by means of Deep Q-Network Shogo Igarashi, Takumi Fukuda, Sei Takahashi, Hideo Nakamura, Tetsuya Takata, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | Although the predictive fuzzy control technique has been put to practical use as a train control strategy for automatic train operation systems, it is difficult to control trains considering the speed limit and gradient of all running sections due to the complexity of the model. In this paper, we propose a method of learning single train control strategies for automatic train operation systems using Deep Q-Network, which learns control strategies based on the experience of simulation in advance, and confirm that the control strategies obtained by this method provide good control in terms of on-time performance, energy saving, and good ride quality. The control strategy obtained by this method is confirmed to provide good control in terms of punctuality, energy efficiency, and good ride quality. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | Auto Train Control / Train Control / Machine Learning / Reinforcement Learning / Deep Q-Network |
| Paper # | DC2022-77 |
| Date of Issue | 2022-12-09 (DC) |
| Conference Information | |
| Committee | DC |
|---|---|
| Conference Date | 2022/12/16(1days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | Safety, etc. |
| Chair | Tatsuhiro Tsuchiya(Osaka Univ.) |
| Vice Chair | Toshinori Hosokawa(Nihon Univ.) |
| Secretary | Toshinori Hosokawa(Nihon Univ.) |
| Assistant | |
| Paper Information | |
| Registration To | Technical Committee on Dependable Computing |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Learning of train control measures by means of Deep Q-Network |
| Sub Title (in English) | Preliminary study with a single train control |
| Keyword(1) | Auto Train Control |
| Keyword(2) | Train Control |
| Keyword(3) | Machine Learning |
| Keyword(4) | Reinforcement Learning |
| Keyword(5) | Deep Q-Network |
| 1st Author's Name | Shogo Igarashi |
| 1st Author's Affiliation | Department of Computer Science, Graduate School, Nihon University(Nihon Univ) |
| 2nd Author's Name | Takumi Fukuda |
| 2nd Author's Affiliation | Department of Computer Engineering, College of Science and Technology, Nihon University(Nihon Univ) |
| 3rd Author's Name | Sei Takahashi |
| 3rd Author's Affiliation | Department of Computer Engineering, College of Science and Technology, Nihon University(Nihon Univ) |
| 4th Author's Name | Hideo Nakamura |
| 4th Author's Affiliation | Professor Emeritus of Nihon University(Nihon Univ) |
| 5th Author's Name | Tetsuya Takata |
| 5th Author's Affiliation | Kyosan Electric Manufacturing Co.,Ltd.(Kyosan Electric Manufacturing) |
| Date | 2022-12-16 |
| Paper # | DC2022-77 |
| Volume (vol) | vol.122 |
| Number (no) | DC-318 |
| Page | pp.pp.26-29(DC), |
| #Pages | 4 |
| Date of Issue | 2022-12-09 (DC) |