Aula Collegezaal A
Setup: two research talks (academia / industry), and cookies and drinks afterwards, sponsored by Delft Data Science.
Dr. Victor Pereboom, Dutch Analytics
Dr. Wan-Jui Lee, a lead AI researcher of Dutch Railways, will present her work on “Shunting Trains with Deep Reinforcement Learning”.
The Train Unit Shunting Problem (TUSP) is a difficult sequential decision making problem faced by Dutch Railways (NS). Current heuristic solutions under study at NS fall short in accounting for uncertainty during plan execution and do not efficiently support re-planning. Furthermore, the resulting plans lack consistency. In this talk we will explain how the TUSP can be formulated as a Markov Decision Process. By shaping its state space and rewards, it allows us to develop a Deep Reinforcement Learning (DRL) solution. The Deep Q-Network efficiently reduces the state space and develops an on-line strategy for the TUSP capable of dealing with uncertainty and delivering significantly more consistent solutions compared to approaches currently being developed by NS.