Multi-population parallel genetic algorithm using a new genetic representation for the euclidean traveling salesman problem

Abstract

This paper introduces a multi-population genetic algorithm (M-PPGA) using a new genetic representation, the kth-nearest neighbor representation, for Euclidean Traveling Salesman Problems. The proposed M-PPGA runs M greedy genetic algorithms on M separate populations, each with two new operators, intersection repairing and cheapest insert. The M-PPGA finds optimal or near optimal solutions by using a novel communication operator among individually converged populations. The algorithm generates high quality building blocks within each population; then, combines these blocks to build the optimal or near optimal solutions by means of the communication operator. The proposed M-PPGA outperforms the GAs that we know of as competitive with respect to running times and solution quality, over the considered test problems including the Turkey81.