A Binary Particle Swarm Optimization-based algorithm to Design a Reverse Logistics Network

As recognized by several authors, the design of a reverse logistics network is a complex problem and still relatively unexplored and underdeveloped. We propose a binary particle swarm optimization (BPSO)-based scheme for solving a NP-hard remanufacturing network design problem. The algorithm combines a traditional stochastic search with an optimal solution method for solving to optimality a relaxed LP problem. We divide the swarm into two elementary groups. The first swarm group guides the search for the best location of remanufacturing facilities, while the second group defines the optimal flows between the facilities. We solve to optimality a relaxed LP problem obtained from the original problem and then we project the solution into the swarm space. The algorithm was coded in GAMS and we report computational results for 10 network instances generated randomly with up to 350 sourcing facilities, 100 candidate sites for locating reprocessing facilities and 40 remanufacturing facilities. Computational results regarding gap and computing times are promising.