A NEW PSO-BASED ALGORITHM FOR AN OPERATIONAL MANAGEMENT PROBLEM

Abstract

Operational management issues represent a permanent challenge for the current economic environment and the research activity. This research will model a Travelling Salesman Problem (TSP). The complexity of this fundamental problem (np-hard) allows a chance to apply and develop heuristic methods and evolutionary algorithms along with exact methods (dynamic programming, branch & bound). This paper proposes a new discrete algorithm to solve the TSP based on the Particle Swarm Optimization (PSO) technique. The features of this method are fast determination through an iterative process of the optimal problem, the generalised search in all the solutions, and the avoidance of the local optimal. To avoid premature convergence, we have introduced a new operator with a new mathematical function, and we have proposed new techniques for measuring and maintaining population diversity. We tested the algorithm's performance by applying it to numerical instances and compared it to the solutions and performance provided by other evolutionary algorithms. By delaying the convergence process, the new algorithm PSO offers reasonable solutions in terms of quality comparable to those offered by different evolutionary algorithms tested. At the end of the research, we highlighted the conclusions, limitations, and new techniques based on the PSO algorithm.