Performance Comparison of Steady State GAs and Generational GAs for Capacitated Vehicle Routing Problems

Abstract

This paper presents a comparison on performances between the Coarse-Grained Steady-State Genetic Algorithm (SSGA) and the Generational Genetic Algorithm (GGA) on benchmark problems of the Capacitated Vehicle Routing Problem (CVRP). A statistical fractional multi-factorial design of experiments is done to find optimal parameter settings for the SSGA, while the best settings for the GGA were taken from aprevious study. The GAs were compared pairwise on problems of various sizes, with results indicating the SSGA outperforms the GGA on all the problems. A pooled statistical test further support this, with a p-value less than 0.05%, further proving the SSGA is significantly better than the GGA.