Typical parameters used in the study of GA include but not limited to genetic operators [such as crossover and mutation], population size, elitism, selection pressure, replacement strategy, fitness evaluation, etc. In this work, a wide range of parameters were varied on the VRPTW to empirically establish best combinations of evolutionary parameters that produce desirable results.
Weighted sum was used as the fitness measure for testing the various parameters. The presence of multiple design objectives renders the VRPTW as a multi objective problem (MOP). Thus, the resulting outcome from the above experiment is used to run the combinatorial problem using a Multi Objective GA (MOGA) fitness approach and the results compared with the Weighted Sum fitness measure. It was shown that MOGA approach of fitness evaluation [using sum-of-ranks and Pareto ranking] did not outperform Weighted Sum (WS) at the 95% confidence level, however it presented broad range of acceptable solutions to choose from. The MOGA solutions, are not biased towards any of the considered dimensions unlike WS.