User-incentive based Rebalancing for Shared e-Mobility Systems

Urban mobility systems are all about people floating around in space and time. This causes a natural problem of fleet imbalance, i.e. vehicles could gradually concentrate to some spots of system and stuck there.

To address this problem, in this work we consider a rebalancing approach by incentivising users. We offer monetary rewards to the users, in exchange for their cooperation to return vehicles at stations that are not their original destinations, but more beneficial for rebalancing. The rebalancing strategies, i.e., when and how much to offer incentives, are learned via reinforcement learning, where we design cascaded actions to cope with dynamics, e.g. the system expansion, and also directly model constraints such as range limitations and charging times in the states.

For more details please see the following papers:

Rebalancing the Expanding EV Sharing Systems: A Multi-agent Deep Reinforcement Learning Approach.
Man Luo, Wenzhe Zhang, Tianyou Song, Kun Li, Hongming Zhu, Bowen Du, Hongkai Wen.
To appear in IEEE Transactions on Intelligent Transportation Systems (T-ITS), 2022.

Rebalancing Expanding EV Sharing Systems with Deep Reinforcement Learning.
Man Luo, Wenzhe Zhang, Tianyou Song, Kun Li, Hongming Zhu, Bowen Du, Hongkai Wen.
In Proceedings of the Twenty-Ninth International Joint Conference on Artificial Intelligence (IJCAI), pp. 1338-1344, 2020.