Enhancing Smart, Sustainable Mobility with Game Theory and Multi-Agent Reinforcement Learning

This work proposes the use of game-theoretic solutions and multiagent reinforcement learning in the mechanism design of smart and sustainable mobility services. In particular, we focus on applications to ridesharing as an example of a cooperative cost game. As such, we firstly solve the coalition formation problem and propose algorithms to allocate riders into cars in a socially optimal way. Secondly we propose a mechanism to share the cost in an equitable way so that ridesharing is incentivized. For the proposed methods, we study properties of individual rationality and stability. Lastly, we discuss future work, where we plan to compare centralized solutions with decentralized algorithms based on multi-agent reinforcement learning.