A similarity-based mechanism to control genetic algorithm and local search hybridization to solve traveling salesman problem

Abstract

A big shortcoming of the simple genetic algorithm is that whenever it converges to a local optimum, its performance is continuously deteriorating, especially in the highly nonlinear problems such as traveling salesman problem (TSP). Therefore, some heuristics such as local search are needed to help genetic algorithm (GA) loops escaping such situations. The critical point in such hybridization is the determining a suitable time for applying local search to the GA population. In this study, a new hybridization of GA and local search based on a new similarity-based control mechanism is proposed, and its behavior on different TSP instances is compared with simple GA. The experimental results show that the proposed hybrid algorithm yields the optimal tour length in most of the cases, especially in the TSP instances with higher complexity.