Dynamic Pricing and Incentivization in Multimodal Transportation Networks using Deep Reinforcement Learning

Abstract

This paper proposes a multi-agent deep reinforcement learning (DRL) framework for coordinating dynamic pricing and incentivization strategies in multimodal transportation networks. Two competing DRL agents are deployed: a shared mobility service provider that maximizes revenue through adaptive fare adjustment, and a public authority that allocates spatio-temporal transit subsidies to improve equity, reduce emissions, and enhance system efficiency. Unlike prior work that treats pricing and incentivization separately, the proposed multi-agent DRL formulation captures their dynamic interdependencies within a multimodal network, enabling more balanced and system-level optimal outcomes. Experiments on the Sioux Falls benchmark network demonstrate that dynamic incentivization reduces commuters’ costs by around 20%, cuts emissions by approximately 10%, and nearly doubles public transport profit during the morning peak, while the joint dynamic pricing-incentivization strategy successfully balances conflicting objectives between private and public actors. The proposed approach can support the analysis of sustainable and equitable multimodal mobility strategies.