Multi-objective Emergency Resource Dispatch Based on Coevolutionary Multiswarm Particle Swarm Optimization

Abstract

Emergency resource dispatch plays a significant role in the occurrence of emergency events. An efficient schedule solution not only can deliver the required resources in time, but also can reduce the loss in the disaster area. Recently, many scholars are dedicated to dealing with emergency resource dispatch problem (ERDP) by constructing a model with one objective (e.g., the cost to transport resources or the satisfaction degree of people in the disaster area) and solving the model with single objective optimization algorithms. In this paper, we build a multi-objective model that considers both the cost objective and the satisfaction objective, which takes into account multiple retrieval depots and multiple kinds of resources. We propose to solve this multi-objective ERDP optimization model via the recently famous coevolutionary multiswarm particle swarm optimization (CMPSO) algorithm. Based on multiple populations for multiple objectives (MPMO) framework, the CMPSO algorithm uses two populations to optimize the above two objectives respectively, and leads particles to find Pareto optimal solutions by storing information of different populations in a shared archive. We construct ERDP with various scales to validate the feasibility of the applied CMPSO algorithm. Moreover, by setting the satisfaction objective as the constraint, we also compare the results obtained by CMPSO with those obtained by constrained single objective particle swarm optimization (PSO) algorithm. Experimental results show that: 1) the nondominated solutions obtained by CMPSO perform well in both convergence and diversity on two objectives; 2) the results on the cost objective obtained by CMPSO are generally superior to those of PSO under same degree of satisfaction. © 2021, Springer Nature Switzerland AG.