Multi-Objective Vehicle Routing Problems Using Two-Fold EMO Algorithms to Enhance Solution Similarity on Non-dominated Solutions

In this paper, we focus on the importance of examining characteristics of non-dominated solutions especially when a user should select only one solution from non-dominated solutions at a time, and select another solution due to the change of problem conditions. Although he can select any solution from non-dominated solutions, the similarity of selected solutions should be considered in practical cases. We show simulation results on vehicle routing problems that have two demands of customers: Normal Demand Problem (NDP) and High Demand Problem (HDP). In our definition the HDP is an extended problem of NDP. We examined two ways of applying an EMO algorithm. One is to apply it to each problem independently. The other is to apply it to the HDP with initial solutions generated from non-dominated solutions for the NDP. We show that the similarity of the obtained sets of nondominated solutions is enhanced by the latter approach.