Evolutionary Algorithm with Cross-Generation Environmental Selection for Traveling Salesman Problem

Abstract

To enhance the evolution of classical evolutionary algorithms (EAs) in solving the traveling salesman problem (TSP), this paper devises a cross-generation environmental selection mechanism to pick promising yet diversified individuals for the next iteration. Specifically, instead of only using parental individuals or offspring individuals, such an environmental selection strategy combines the offspring in two consecutive generations and then selects the half best individuals as the parent population for EAs to evolve in the next iteration. In this way, EAs with this approach not only preserve rich diversity but also have rapid convergence, enabling the population to discover optimal solutions effectively and efficiently. Particularly, this paper embeds this environmental selection strategy into the classical genetic algorithm (GA) along with three different crossover strategies to solve TSP. Experiments have been conducted on 8 TSP instances of different scales from the TSBLIB benchmark set. Experimental results demonstrate that the proposed environmental selection scheme is very helpful for EAs to solve TSP. © 2024 held by the owner/author(s).