Deep Reinforcement Learning-Based Method for Multi-Stage Resource Allocation in Infectious Disease Emergencies

Different from common one-time natural disasters, infectious disease disasters have the characteristics of persistence and spread. Sudden infectious disease natural disasters often bring greater impact and damage to human beings, which also puts forward higher requirements for the distribution of emergency materials. In addition to considering the special nature of infectious disease disasters, the human suffering caused by disasters should also be taken into account. In this paper, the improved SEIR model of infectious disease dynamics is combined with the material distribution model to establish a real multi-stage emergency material distribution model to reduce the spread of the epidemic and the suffering of the victims. In order to solve the problem of limited ability of accurate algorithm and low quality of heuristic algorithm, a material allocation method based on deep reinforcement learning is designed. Numerical experiments show that the method based on deep reinforcement learning is not limited by the scale of the instance, and its solution quality is obviously higher than that of the heuristic algorithm with the increase of the scale of the instance.