A matching algorithm with reinforcement learning and decoupling strategy for order dispatching in on demand food delivery

The surge in online food ordering has amplified the demand for efficient and intelligent order dispatching systems.Traditional dispatching strategies, often based on rule-based heuristics, fail to adapt to the dynamic nature of urban delivery environments.This paper proposes a novel matching algorithm that combines reinforcement learning (RL) with a decoupling strategy to address the complexities in real-time order-courier assignment.The reinforcement learning component enables the system to learn optimal dispatching policies through continuous interaction with the delivery environment.Meanwhile, the decoupling strategy allows the system to isolate subcomponents of the decision-making process, thereby reducing computational complexity and enhancing scalability.The proposed method is validated through simulation using real-world delivery data, showing significant improvements in delivery efficiency, order completion rates, and courier utilization when compared with conventional algorithms.This hybrid approach demonstrates potential for deployment in large-scale commercial food delivery platforms.