A Modified Discrete Particle Swarm Optimization Algorithm for the Generalized Traveling Salesman Problem

Abstract

The mDPSO algorithm proposed employs the destruction and construction procedure of the iterated greedy algorithm (IG) in its mutation phase. Its performance is enhanced by employing a population initialization scheme based on an NEH constructive heuristic for which some speed-up methods previously developed by authors are used for greedy node insertions. Furthermore, the mDPSO algorithm is hybridized with local search heuristics to achieve further improvements in the solution quality. To evaluate its performance, the mDPSO algorithm is tested on a set of benchmark instances with symmetric Euclidean distances ranging from 51 (11) to 1084 (217) nodes (clusters) from the literature. Furthermore, the mDPSO algorithm was able to find optimal solutions for a large percentage of problem instances from a set of test problems in the literature. It was also able to further improve 4 out of 9 larger instances from the literature. Both solution quality and computation times are competitive to or even better than the best performing algorithms from the literature.