Dynamic multi-agent deep reinforcement learning-based pricing and incentivization approach in multimodal transportation networks

The paper presents a multi-agent deep reinforcement learning framework designed to optimize pricing and incentivization strategies in multimodal transportation networks. It employs two RL agents: one representing a public authority focused on enhancing equity and efficiency, and another representing a shared mobility service provider aiming to maximize revenue. Numerical experiments indicate that the proposed approach can reduce congestion peaks, decrease commuter costs by approximately 20%, lower emissions by about 10%, and nearly double public transport profits, thereby facilitating a more equitable distribution of transportation benefits. This framework offers a valuable decision-support tool for practitioners engaged in sustainable mobility planning.