High-Performance Local Search for Solving Real-Life Inventory Routing Problems

Abstract

In this paper, a real-life routing and scheduling problem encountered is addressed. The problem, which consists in optimizing the delivery of fluids by tank trucks on a long-term horizon, is a generalization of the vehicle routing problem with vendor managed inventory replenishment. The particularity of this problem is that the vendor monitors the customers’ inventories, deciding when and how much each inventory should be replenished by routing tank trucks. Thus, the objective of the vendor is to minimize the logistic cost of the inventory replenishment for all customers over the long run. Then, an original local-search heuristic is presented for solving the short-term planning problem. The engineering of this algorithm follows the three-layers methodology for “high-performance local search” recently introduced by some of the authors. A computational study demonstrates that our solution is both effective, efficient and robust, providing long-term savings exceeding 20 % on average, compared to solutions computed by expert planners or even a classical greedy algorithm. The resulting software is now exploited in North America by one of the French industry leaders.