Multi-Population Genetic Algorithm for Rich Vehicle Routing Problems

Abstract

Genetic algorithm (GA) is a metaheuristic method that has been widely adopted for solving the Rich Vehicle Routing Problems (RVRP) due to its ability to find quality approximate solutions, even for large-scale instances of the problem, in a reasonable time. However, GA is stochastic in nature and does not guarantee a good solution all the time, a problem primarily due to premature convergence. In this paper we present Multi-population Genetic Algorithm for Rich Vehicle Routing Problems (MPGA-RVRP) to provide diversity and delay premature convergence in GA by making use of multiple populations that each evolves independently optimizing a single objective while sharing potential solutions. MPGA-RVRP is applied in RVRP with three objectives:- total route distance, total route duration and total route cost. Results from the experiments show that MPGA-RVRP performs considerably better compared to benchmark, multi-objective Genetic Algorithm (MOGA).