A Comparison Of Memetic Recombination Operators For The Traveling Salesman Problem

Several memetic algorithms (MAs) – evolutionary algorithms incorporating local search – have been proposed for the traveling salesman problem (TSP). Much effort has been spent to develop recombination operators for MAs which aim to exploit problem characteristics to achieve a highly effective search. In this paper, several recombination operators for the TSP are compared. For the purpose of identifying the important properties of a recombination operator, a new generic recombination operator (GX) is proposed which is comprised of four phases. These phases can be controlled by parameters reflecting the most important properties of recombination operators. It is shown that GX recombination is superior to MPX and DPX when all common edges are preserved in the offspring.