In this work we approach the dynamic taxi dispatch problem as a Markov Game and solve it using a model free Deep Reinforcement Learning approach. We propose a novel Markov Game formulation, and we address the challenges associated with such formulation and strive to improve the efficiency of the solution. The biggest challenge is the size of the state space which grows exponentially with the size of the taxi fleet, the number of passengers, and the number of locations. We tackle this problem by representing the problem state as an image and using a Convolutional Neural Network as a function approximator, making the state space dependent only on the size of the map. The proposed algorithm was validated against a rule-based heuristic under different supply-demand ratios, and it was found to outperform the rule-based technique by a large margin when there is a lack of supply.
A Scalable Multi-agent Reinforcement Learning Approach for the Dynamic Taxi Dispatch Problem
Lect. Notes Electrical Eng.
International Conference on Intelligent Transportation Engineering ; 2021 ; Beijing, China October 29, 2021 - October 31, 2021
2021 6th International Conference on Intelligent Transportation Engineering (ICITE 2021) ; Kapitel : 101 ; 1145-1156
2022-06-01
12 pages
Aufsatz/Kapitel (Buch)
Elektronische Ressource
Englisch
Europäisches Patentamt | 2023
Europäisches Patentamt | 2022
|