A discrete squirrel search optimization based algorithm for Bi-objective TSP

Abstract

This paper presents a novel discrete squirrel search optimization algorithm for the bi-objective traveling salesman problem (TSP). Firstly, the squirrel search algorithm, a single-objective optimization algorithm, needs to be improved to a multi-objective optimization algorithm. This paper designs a mapping fitness function according to the Pareto sorting level and grid density to evaluate all the feasible solutions and applies a selection probability based on the roulette wheel selection. Then, this paper implements this algorithm and other algorithms on classic multi-objective test functions to analyze solutions’ convergence and diversity. It is concluded that it has a good performance in solving multi-objective problems. Moreover, based on this multi-objective squirrel search algorithm, this paper then designs an encoding method to initialize solutions, applies a crossover operator to the squirrel migration process, and utilizes a mutation operator to the squirrel mutation stage. In this case, a discrete squirrel search optimization for the bi-objective traveling salesman problem (TSP) is finally designed. And this paper analyzes the results of this algorithm and other algorithms running on classic bi-objective TSPs. As a result, the presented algorithm’s solutions are also superior to other algorithms for convergence and spread.