Simulated Annealing for the Multi-Vehicle Cyclic Inventory Routing Problem

Abstract

This paper studies the Multi-Vehicle Cyclic Inventory Routing Problem (MV-CIRP) as the extension of the Single-Vehicle CIRP (SV-CIRP). The objective is to minimize both distribution and inventory costs at the customers and to maximize the collected rewards simultaneously. The problem is treated as a single objective optimization problem. A subset of customers is selected for each vehicle including the quantity to be delivered to each customer. For each vehicle, a cyclic distribution plan is developed. We construct a mathematical programming model and propose a simulated annealing (SA) metaheuristic for solving both SV-CIRP and MV-CIRP. For SV-CIRP, experimental results on benchmark instances show that SA is comparable to the state-of-the-art algorithms and it is able to improve 12 best known solutions. For the MV-CIRP, the results show that SA performs better than an Iterated Local Search algorithm.

The Multi-Vehicle Cyclic Inventory Routing Problem: Formulation and a Metaheuristic Approach

“…This consideration has not been studied before and is considered explicitly in this paper to fill a gap in the CIRP literature. We further show the advantages of the MV-CIRP discussed in this paper compared to the variant discussed in [16].…”

“…Gunawan et al [16] introduced the MV-CIRP with a fixed number of vehicles. A simulated annealing algorithm is proposed to solve both SV-CIRP and these specific MV-CIRP instances.…”

“…Gunawan et al [16] introduced the MV-CIRP with a fixed number of vehicles in a short conference paper. In the current paper, we present a more difficult, but also more realistic, version by treating the number of vehicles as a decision variable.…”

“…• The mathematical model of the MV-CIRP as the extension of the work of Gunawan et al [16] is presented.…”

“…• A simple yet powerful algorithm, based on Simulated Annealing (SA), that efficiently solves the MV-CIRP, and also the SV-CIRP, is introduced. This algorithm is adopted from the SA proposed by Gunawan et al [16]. However, an intensification strategy is incorporated in order to focus the search once again starting from the best solution if there is no improvement within a certain number of iterations.…”