A knowledge-based initialization technique of genetic algorithm for the travelling salesman problem

Genetic Algorithm (GA) is efficient for the travelling salesman problem, but it has the defect of slow convergence and is easily trapped in local optima. Because the initialization has a profound impact on the optimization, this study proposed to improve the performance of GA by applying a knowledge-based initialization technique (KI). KI learns the features of evolved population and uses them to guide the generation of initial population. Advanced initial solution without path crossover can be fast generated with this method. Instances in TSPLIB were used to test different initialization methods. The results proved that this proposed technique helped GA get better initial population and performance.