Local Optimization Approaches to Solve Dynamic Commercial Fleet Management Problems

Abstract

Strategies are presented that a dispatcher may use to take advantage of real-time information to manage a fleet of trucks to provide truckload pickup-and-delivery service under time windows for service and computing time constraints. These strategies are hybrid in that they combine very efficient dynamic local heuristic rules for quick initial assignment, with formal optimization-based procedures for subsequent load reassignment decisions within the time requirements of the dynamic problem. Simulation experiments were performed to test these strategies under alternative specifications and parameter values. In the first stage, the acceptance decision and initial schedule of trucks were determined so as to respond rapidly to incoming customer requests for service. Then, some time is allowed to improve the schedule of trucks by swapping the assigned loads using local optimization at the reassignment stage. Two strategies are described to control the computation time of the reassignment optimization problem by controlling the problem size. The simulation results confirm the significant potential for reassignment strategies under real-time information to improve system performance and suggest that considerable improvement is attainable even under restricted problem formulations.