A hybrid genetic algorithm for the heterogeneous dial-a-ride problem

Abstract

This paper investigates the Heterogeneous Dial-A-Ride Problem (H-DARP) that consists of determining a vehicle route planning for heterogeneous users’ transportation with a heterogeneous fleet of vehicles. A hybrid Genetic Algorithm (GA) is proposed to solve the problem. Efficient construction heuristics, crossover operators and local search techniques, specifically tailored to the characteristics of the H-DARP, are provided. The proposed algorithm is tested on 92 benchmarks instances and 40 newly introduced larger instances. Computational experiments show the effectiveness of our approach compared to the current state-of-the-art algorithms for the DARP and H-DARP. When tested on the existing instances, we achieved average gaps of only 0.47% to the bestknown solutions for the DARP, and 0.05% to the optimal solutions for the H-DARP, compared to 0.85% and 0.10%, respectively, obtained by the current state-of-the-art algorithms. For the 40 newly generated instances, average gaps of the hybrid GA are 0.35% smaller compared to the current stateof-the-art method. Besides, our method provides best results for 31 of these instances and ties with the existing method on 8 other instances.