Comparing the performance of genetic operators for the vehicle routing problem

This paper compares the performances of genetic operators for the Capacitated Vehicle Routing Problem (CVRP). Two representations and some crossover and mutation operators are considered. Design of Experiment (DoE) is used to identify the best combination of the studied GA components. The goal is to determine the combination of features that maximizes the efficiency of the algorithm. Results provided from a computational experiment over common benchmarks showed which parameters and genetic operators affect the quality of provided solutions.