Effective truckload dispatch decision methods with incomplete advance load information

We investigate the following question of relevance to truckload dispatchers striving for profitable decisions in the context of dynamic pick-up and delivery problems: ``since not all future pick-up/delivery requests are known with certainty (i.e., advance load information (ALI) is incomplete), how effective are alternative methods for guiding those decisions?'' We propose a simple intuitive policy and integrate it into a new two-index mixed integer programming formulation, which we implement using the rolling horizon approach. On average, in one of the practical transportation network settings studied, the proposed policy can, with just second-day ALI, yield an optimality ratio equal to almost 90 percent of profits in the static optimal solution (i.e., the solution with asymptotically complete ALI). We also observe from studying the policy that second-day load information is essential when a carrier operates in a large service area. We enhance the proposed policy by adopting the idea of a multiple scenario approach. With only one-day load information, the enhanced policy improves the ratio of optimality by an average of 6 percentage points. That improvement declines with more ALI. In comparison to other dispatching methods, our proposed policy and the enhanced version we developed were found to be very competitive in terms of solution quality and computational efficiency.