An Order Dispatch Algorithm for Ride-Hailing Based on Quantum Annealing and Multi-Agent Reinforcement Learning

The popularity of ride-hailing has grown significantly, making it one of the mainstream modes of transportation. Its core value and competitive advantage largely lie in the precise and efficient allocation of orders to drivers and passengers on the ride-hailing platform. In this paper, we use multi-agent reinforcement learning as the framework to accurately model and solve the state transition process in the dispatch area. For the problem of e-hailing order allocation, it is transformed into a quadratic unconstrained binary optimization (QUBO) model suitable for quantum annealing algorithm processing, and a reinforcement learning model is used to guide the quantum annealing QUBO model to provide multi vehicle order allocation decisions. Deploy the proposed algorithm on a simulation platform based on 100G e-hailing data from Chengdu City within a month, and conduct performance comparison experiments with Hungarian algorithm, greedy algorithm, and simulated annealing. The experimental results show that compared to the three algorithms mentioned above, the algorithm proposed in this paper can increase the average total reward of scheduling ride-hailing by 20% and the order acceptance rate by about 10%.