Adaptive Memory Procedure for Solving Real-world Vehicle Routing Problem

TL;DR

This paper introduces an adaptive memory metaheuristic for solving complex real-world vehicle routing problems, achieving significant cost and time savings over existing algorithms.

Contribution

It presents a novel adaptive memory procedure combined with local search tailored for complex industrial VRPs with multiple constraints.

Findings

Average savings of 2.03% in delivery time

Average savings of 20.98% in total delivery costs

Validated on real industrial case study and benchmarks

Abstract

Logistics and transport are core of many industrial and business processes. One of the most promising segments in the field is optimisation of vehicle routes. Scientific effort is focused primarily on algorithms developed in simplified environment and cover a fraction of real industrial application due to complex combinatorial algorithms required to be promptly executed. In this paper, a real-world case study in all its complexity is observed and formulated as a real-world vehicle routing problem (VRP). To be able to computationally cope with the complexity, we propose a new procedure based on adaptive memory metaheuristic combined with local search. The initial solution is obtained with Clarke-Wright algorithm extended here by introducing a dropout factor to include a required stochastic attribute. The procedure and corresponding algorithms are tested on the existing benchmarks and further on the real industrial case study, which considers capacities, time windows, soft time windows, heterogeneous vehicles, dynamic fuel consumption, multi-trip delivery, crew skills, split delivery and, finally, time-dependent routes as the most significant factor. In comparison with the current state-of-the-art algorithms for vehicle routing problem with a large number of constraints, we obtain an average savings of 2.03% in delivery time and 20.98% in total delivery costs.