Applying Matheuristic To Solve Inventory-Routing Problem In Vendor-Managed Inventory Context

Abstract

This thesis studies a multi - depot distribution system. In this system, the products must be delivered from a group of depots to several retailers in a finite time horizon—the shipments from the depots to retailers of the homogeneous vehicle with limited carrying capacity. Customer demand at each retailer is dynamic with different periods. Decisions in this system include replenishment quantities at each retailer during considered periods and its corresponding delivery route with minimum total inventory and routing cost during the considered time horizon. This thesis considers the problem as a Multi-Depot Inventory Routing Problem (MDIRP). In general, the MDIRP is an NP-hard problem that optimizes the trade-off between inventory and transportation management in an integrated way. A Mixed-Integer Linear Programming model is formulated to solve it, and then this thesis proposes a three-phase decomposition matheuristic to solve the problem in real-size problem efficiently. In the first phase of the matheuristic, an integer program is solved to build clusters, while the second phase generates routes. Finally, in the third phase, a routebased formulation of the problem is solved to obtain the final a feasible MDIRP solution. More emphasis is devoted to simultaneously balancing several factors that impact the clustering and route construction phases: distances, demand and inventory levels, time horizon extension and vehicle capacity. Computational experiments show that the matheuristic is very effective.