Comparison of Genetic Crossover Operators for Traveling Salesman Problem

The traveling salesman problem (TSP) is an NP-hard problem that has been the subject of intensive study by researchers and academics in the field of optimization for many years. Genetic algorithms (GA) are one of the most effective methods for solving various NP-hard problems, including TSP. Recently, many crossover operators have been proposed to solve the TSP problem using GA. However, it remains unclear which crossover operator performs better for the particular problem. In this study, ten crossover operators, namely; Partially-Mapped Crossover (PMX), Cycle Crossover (CX), Order Crossover (OX1), Order Based Crossover (OX2), Position Based Crossover (POS), Edge Recombination Crossover (ERX), Maximal Preservative Crossover (MPX), Extended Partially-Mapped Crossover (EPMX), Improved Greedy Crossover (IGX), and Sequential Constructive Crossover (SCX) have been empirically evaluated. 30 TSP data sets have been used to comprehensively evaluate the selected crossover operators, and the experiments have been repeated 30 times to make our results statistically sound. Likewise, how successful the operators are, has been found through critical diagrams and statistical tests. Among tested operators, the IGX and SCX methods were the best operators in terms of convergence rate. On the other hand, PMX outperformed other operators in terms of computational cost.