A New Approach for Solving Location Routing Problems with Deep Reinforcement Learning of Emergency Medical Facility

The Location Routing Problem (LRP) has extensive applications in emergency medical facility selection. It involves a complex spatial optimization problem of selecting the optimal facility locations from a set of demand points and candidate facility locations and planning the most efficient routes. This study introduces a two-stage solution approach based on deep reinforcement learning, which fully leverages the potential of deep reinforcement learning in handling sequential decision problems. Firstly, in the location selection stage, through the interaction between an agent and the environment, the agent needs to choose facility locations from numerous demand points to either maximize overall profit or minimize overall costs. This stage determines central points and allocates demand points to the nearest facility locations. Subsequently, in the route planning stage, the agent decides how to distribute deliveries among the pre-selected facility locations to minimize overall route length or costs. Experimental results demonstrate the effectiveness of this method in addressing location routing problems. It may provide valuable support for responding to emergencies and disaster events, helping decision-makers make more intelligent choices for facility locations and route planning, ultimately enhancing service response speed and efficiency. Furthermore, this method holds broad application potential in other domains, further advancing the development and utilization of deep reinforcement learning in spatial optimization problems.