Dray Optimization in Truck/Rail Networks

Abstract

Daily drayage operations involve moving loaded or empty equipment between customer locations and rail ramps. Drayage orders are generally pickup and delivery requests with time windows. The repositioning of empty equipment may also be required in order to facilitate loaded movements. The drayage orders are satisfied by a heterogeneous fleet of drivers. Driver routes must satisfy various operational constraints.

In the first part of the dissertation, our goal is to minimize the cost of daily drayage operations in a region on a given day. We present an optimization methodology for finding cost-effective schedules for regional daily drayage operations. The core of the formulation is a set partitioning model whose columns represent routes. Routes are added to the formulation by column generation. We present numerical results for real-world data which demonstrate that our methodology produces low cost solutions in a reasonably short time.

The second part of the dissertation addresses minimizing total empty mileage when driver capacity is not restrictive and new orders are added to the problem in an online fashion. We present a lower bound for the worst case guarantee of any deterministic online algorithm. We develop a solution methodology and provide results for the performance of different scheduling policies and parameters in a simulated environment.

In the third part of the dissertation, we study a system with one rail ramp and one customer location which is served by a single driver. The problem has discrete time periods and at most one new order is released randomly each time period. The objective is to maximize the expected number of orders covered. With this simple problem, we seek to learn more about route planning for a single driver under uncertainty. We prove that carrying out an order ready to be picked up at the driver's current location is optimal for the case with one customer location. We show that the structure of the optimal policies is not simple and depends on various parameters. We devise a simple policy which yields provably near-optimal results and identify a case for which that policy is optimal.