Dock Assignment and Truck Scheduling at Cross-docking Terminals

Abstract

In a cross docking system products are received at the terminal, occasionally gathered with other products going to the same destination, then shipped at the earliest opportunity, without going into long-term storage. Applying cross docking in the distribution network can reduce transportation cost, the inventory holding cost, cycle time and increase customer satisfaction. In order to obtain such benefits, the company should be able to operate the cross docking effectively and efficiently. For this purpose several scheduling and door assignment procedures have been introduced in recent years, which aim at solving the so called truck scheduling and dock assignment problem. This research deals with both dock assignment and truck scheduling problems. The integration of both problems is discussed in two scenarios: one considers a static dock assignment and truck scheduling, and the other one addresses a dynamic dock assignment and truck scheduling. Mathematical formulations for both problems are first presented as 0-1 integer programming models. Since both dock assignment and truck scheduling problems are NP-hard, the integration of both problems is more difficult to solve. Thus we propose different heuristic algorithms to solve the integrated problem. For the static dock assignment and truck scheduling three different reduced variable neighborhood search (RVNS) algorithm frameworks are proposed, while for the dynamic dock assignment and truck scheduling a RVNS algorithm with three different types of solution initialization is proposed. The experimental results show that the RVNS algorithms for the static dock assignment and truck scheduling are capable of finding good solutions in a much shorter computation time when it is compared with Gurobi optimizer solutions. For the RVNS for the dynamic dock assignment and truck scheduling it was observed that the algorithm converges to the same quality of solution no matter the type of initial solution generated for small instances, however for large instances a better quality of solution is obtained when the algorithm is seeded by a construction heuristic.