Hybrid Multi-agent Deep Reinforcement Learning for Autonomous Mobility on Demand Systems
This addresses the operational efficiency problem for autonomous mobility service providers, representing an incremental improvement over existing methods.
The paper tackles the problem of proactive request assignment and rejection for profit-maximizing autonomous mobility on demand systems by formalizing it as a Markov decision process and proposing a hybrid method combining multi-agent Soft Actor-Critic and weighted bipartite matching. Experiments on real-world taxi data show it outperforms state-of-the-art benchmarks in performance, stability, and computational tractability.
We consider the sequential decision-making problem of making proactive request assignment and rejection decisions for a profit-maximizing operator of an autonomous mobility on demand system. We formalize this problem as a Markov decision process and propose a novel combination of multi-agent Soft Actor-Critic and weighted bipartite matching to obtain an anticipative control policy. Thereby, we factorize the operator's otherwise intractable action space, but still obtain a globally coordinated decision. Experiments based on real-world taxi data show that our method outperforms state of the art benchmarks with respect to performance, stability, and computational tractability.