Resource Planning in Cross-Docking Platforms - Models and Algorithms

Cross-docking, a relatively new warehouse strategy that has its roots in the industry, can improve the efficiency of a company's logistics and distribution processes. Specifically, it can minimize the costly storage and order picking function of traditional warehouses by efficiently coordinating (i.e., synchronizing) incoming freight flows and outgoing freight flows. Companies from various industries such as the retailing industry, the less-than-truckload logistics service industry, the express and small parcel delivery industry, and the automotive industry operate cross-docking terminals in their transportation networks and benefit from improved service levels, reduced transportation costs, reduced inventory holding costs, reduced handling costs, etc. Besides its practical relevance, cross-docking has also received a lot of academic attention in the last 30 years. Many academic studies have addressed a wide range of strategic, tactical, and operational cross-docking decision problems. Most studies, however, have neglected resource planning aspects and hence failed to address two major concerns of cross-docking practitioners: - Determining the number of resources needed; - Scheduling internal resources in an efficient way. This thesis sets out to bridge this theory-practice gap in the cross-docking domain by proposing new models that combine two interdependent operational problems faced by cross-docking practitioners, namely the scheduling of internal resources and the scheduling of trucks. Three novel problems are introduced in this thesis. First, the resource and truck scheduling problem, denoted as TSFD-RC-F, is proposed. It allows scheduling both resources and trucks when the resource requirements of trucks are given and known in advance. The TSFD-RC-F aims to determine a truck schedule that can be executed with a minimum number of resources. Then, the multi-mode resource and truck scheduling problem (TSFD-RC-V) is proposed. It is a model extension of the TSFD-RC-F and offers the additional flexibility of adapting the number of resources for processing trucks. While deploying more operators accelerates truck processing, deploying fewer operators prolongs the processing time. The model aims to determine how many resources should be deployed for truck processing and at what time trucks should be serviced in order to minimize the maximum number of required resources. Lastly, the shift and truck scheduling problem (ISTSFD) is proposed. It considers different operator types (e.g., temporary and regular workers), shift patterns, and work breaks. The ISTSFD seeks to find a truck schedule and employee timetable with minimum labor costs. Two variants of the ISTSFD are presented: a single-mode problem (ISTSFD-F) and a multi-mode problem (ISTSFD-V). As the proposed models' complexity statuses make it challenging to solve large-sized instances with a default solver, tailored column generation-based solution procedures for all three problems are developed. Extensive computational experiments are conducted in order to assess the computational performance of both the mixed-integer programs and the proposed solution procedures. In addition, managerial insights are derived by benchmarking the proposed models against frequently used truck scheduling models. It is shown that the proposed discrete-time MIP formulations clearly outperform the proposed continuous-time MIP formulations in terms of both solution quality and computational time. Moreover, the solution time can be reduced by using the proposed preprocessing parameters for calculating the number of delayed freight units and compelling the service level. While a default solver can solve the discrete-time MIPs for small and medium-sized instances in a reasonable time, it often fails to provide good solutions for very large problem instances with a fine time granularity. The proposed heuristics solution procedures, on the other hand, can provide high-quality solutions for very large problem instances in a short time and clearly outperform commercial solvers. In addition, the following key take-home managerial insights could be derived: - By using the internal resource requirements instead of the frequently used makespan or processing time as the primary performance metrics, the cross-docking platform's operational efficiency can be significantly increased. - By integrating the decision of how many resources should be deployed for truck processing (i.e., considering multi-mode processing), further operational efficiency gains can be realized. - The defined service level has a significant impact on the operator demand. Lowering the required service level can be a reasonable means to improve a cross-docking facility's operational efficiency further. - The work break patterns have a significant impact on the operator requirements. Too low a number of work break patterns may result in a strong surge in operator demand.