The MOEA/D algorithm with gaussian neighbourhoods for the multiobjective travelling salesman problem

Krzysztof Michalak

Abstract

In this paper the MOEA/D-G algorithm is proposed which is a modification of the MOEA/D algorithm using Gaussian distributions to determine the probability with which neighbours of a given subproblem are selected as parents of new specimens assigned to this subproblem. The proposed method is applied to the Multiobjective Travelling Salesman Problem (MOTSP). Solutions found by the MOEA/D-G algorithm have a better quality than those found by the original MOEA/D version. Also, given equal computation time, all versions of the MOEA/D outperform the NSGA-II algorithm