Title

Reinforcement Learning-based Metro Train Tracking Control to Overcome Input Time Delay

Alternative Title

입력 시간 지연 극복을 위한 강화학습 기반 지하철 열차 추적 제어

DGIST Authors

Kyungbae Lee ; Yongsoon Eun ; Sehoon Oh

Advisor

은용순

Co-Advisor(s)

Sehoon Oh

Issued Date

2024

Awarded Date

2024-08-01

Citation

Kyungbae Lee. (2024). Reinforcement Learning-based Metro Train Tracking Control to Overcome Input Time Delay. doi: 10.22677/THESIS.200000804947

Type

Thesis

Description

Reinforcement learning, Deep deterministic policy gradient, Input time delay, Automatic train operation, Metro train

Table Of Contents

1 Introduction 1
1.1 Motivation 1
1.2 Related Work 2
1.3 Contribution 3
1.4 Thesis Outline 3
2 ATO Simulator and RL 5
2.1 ATO simulator 5
2.2 RL algorithms 8
3 Prediction-based input time delay compensation 11
3.1 Step 1 11
3.2 Step 2 13
4 Training 17
4.1 Training Episode 17
4.2 Neural Networks and HyperParameter 19
5 Results 21
5.1 Impact of Prediction 21
5.2 Comparison with Existing Controller in Selected Sections 24
5.3 Comparison with Existing Controller in Every Section 30
6 Conclusion 49
bibliography 54
국문초록 56

URI

http://hdl.handle.net/20.500.11750/57638
http://dgist.dcollection.net/common/orgView/200000804947

DOI

10.22677/THESIS.200000804947

Degree

Master

Department

Department of Electrical Engineering and Computer Science

Publisher

DGIST

Show Full Item Record