Population Based Equilibrium in Hybrid SA/PSO for Combinatorial Optimization

This article introduces a hybrid algorithm combining simulated annealing (SA) and particle swarm optimization (PSO) to improve the convergence time of a series of combinatorial optimization problems. The implementation carried out a dynamic determination of the equilibrium loops in SA through a simple, yet effective determination based on the recent performance of the swarm members. In particular, the authors demonstrated that strong improvements in convergence time followed from a marginal decrease in global search efficiency compared to that of SA alone, for several benchmark instances of the traveling salesperson problem (TSP). Following testing on 4 additional city list TSP problems, a 30% decrease in convergence time was achieved. All in all, the hybrid implementation minimized the reliance on parameter tuning of SA, leading to significant improvements to convergence time compared to those obtained with SA alone for the 15 benchmark problems tested.