Three essays on the location routing problem with intermediate storage facilities.

摘要

The Location Routing Problem with Intermediate Storage Facilities (LRP-IF) is a complex product distribution first encountered by a large magazine publisher and distributor located in the southeastern United States. In the first article of this dissertation, a mathematical programming formulation for the LRP-IF is developed and the intractability of the problem is demonstrated by a brief set of computational experiments using a commercial solver. In the dissertation's second article, two decomposition-based heuristic approaches are developed for the LRP-IF and tested via a series of computational experiments against a CPLEX-generated lower bound.;A simulated annealing improvement procedure is constructed to improve upon the solutions developed by the three-phase heuristic approach. This procedure yields modest improvements for several problem instances. In the third article of this dissertation, the three-phase heuristic and the simulated annealing improvement procedure are also applied to a real-world LRP-IF problem instance. This application results in significant improvement to the solutions developed manually by the magazine publishing and distribution company.;Opportunities for future work are presented. These opportunities include the application of the three-phase heuristic and the simulated annealing improvement procedure to problems similar to the LRP-IF. A conceptual application of the heuristic approach to the Pickup-and-Delivery Problem is described. Other possible extensions include consideration of multiple time periods, strengthening of the LRP-IF lower bound, and consideration of additional facets of the real-world LRP-IF problem instance.;The two-phase heuristic approach has proven to be impractical for large problems due to the difficulty of developing solutions during one of the phases. The three-phase heuristic develops solutions quickly for each problem instance tested. These solutions are of good quality in all but a few problem configurations. These difficult problem configurations have been examined and possible explanations for poor heuristic performance are presented.