When to use simulated annealing for solving CVRP? A case study of fuel deliveries in Poland

TL;DR

This paper presents a novel SA-based heuristic for solving the CVRP tailored to fuel delivery networks, demonstrating its effectiveness through computational tests in Poland's gas station network.

Contribution

It introduces a customized CVRP model for fuel delivery and integrates it with a Simulated Annealing algorithm, showing improved solution efficiency over traditional methods.

Findings

SA-based heuristic outperforms Gurobi in solution quality

Customized model effectively incorporates real-world constraints

Computational tests validate approach in Poland's gas station network

Abstract

The paper addresses Capacitated Vehicle Routing Problem (CVRP) in the context of fuel delivery to gas stations. The CVRP aims to minimize total travel distance for a fleet with limited capacity. Fuel delivery, however, introduces unique complexities within the CVRP framework. We propose a novel approach that integrates the Simulated Annealing (SA) algorithm with a customized CVRP model specifically designed for gas station networks. This model incorporates real-world constraints like vehicle capacity, fuel demands at each station, and road network distances. The paper outlines the design of SA-based CVRP model for fuel delivery. We detail the objective function (minimizing distance) and the SA's exploration mechanism for generating candidate solutions. To assess its effectiveness, the proposed approach undergoes computational tests in Poland's gas station network serviced by the Samat transportation company. We compare the performance of our SA-based CVRP model with the conventional Mixed Integer Programming model for CVRP powered by Gurobi. The results aim to demonstrate the efficacy of the proposed SA-based heuristic in finding efficient routes for fuel deliveries.