Real-time Multi-period truckload routing problems

Abstract

In this thesis we consider a multi-period truckload pick-up and delivery problem dealing with real-time requests over a finite time horizon. We introduce the notion of postponement of requests, whereby the company can postpone some requests to the next day in order to improve its operational efficiency. The postponed requests must then be served on the next day. The daily costs of operation include costs associated with the trucks' empty travel distances and costs associated with postponement. The revenues are directly proportional to the length of job requests. We evaluate the profits of various re-optimization policies with the possibility of postponement. Another important notion of trucking operation corresponds to repositioning strategies which exploit probabilistic knowledge about future demands. A new repositioning strategy is proposed here to provide better decisions. For both notions, extensive computational results are provided under a general simulation framework.