Designing efficient order picking systems: combining planning problems and integrating real-life features

Abstract

Complex market conditions and new developments make a warehouse manager's job hard. E-commerce and globalisation intensify competition among warehouses. The high expectations of customers to provide unique products and quick deliveries force warehouses to increase storage capacity, while at the same time reducing pick times. Additionally, expensive industrial land and high labour costs put pressure on the warehouse costs. To cope with these challenges, a wide range of order picking planning problems need to be optimised. Previous academic research focusses mainly on individual planning problems, without accounting for existing real-life features. Optimizing order picking planning problems sequentially may yield a suboptimal overall warehouse performance. Furthermore, excluding real-life features when developing algorithms and decision support tools prevents managers from using the academic findings in practice. Therefore, the objective of this thesis is to design efficient manual order picking systems by combining order picking planning problems and accounting for real-life features (e.g., safety constraints, due time constraints, workload peaks).

The main contributions of this PhD research are as follows. First, a classification of existing literature on tactical and operational order picking planning problems identifies interesting and relevant research directions to narrow the gap between academic research and practice. Second, an interaction analysis explains how and why the four main order picking planning problems (i.e., picker zoning, storage assignment, order batching and routing) are related. It also provides insights into the relevance and importance of incorporating real-life features (i.e., picker blocking, safety constraints and high-level storage) while planning order picking operations. Third, the value of incorporating workload related features is demonstrated by presenting a proof of concept of time series forecasting models in a warehouse context and by introducing a new mathematical programming model that balances the workload of order pickers over a short term planning horizon. Fourth, the benefits of optimising the integrated order batching, routing and job assignment problem are demonstrated, while coping with resource and due time constraints as well as high-level storage locations. Finally, the research provides future research opportunities that will be highly relevant to practice and which are largely unexplored in literature, thereby further reducing the research-practice gap.

Results show that the total order pick time can be substantially reduced by combining order picking planning problems. Combining existing order picking policies may yield performance benefits of 60-75% compared to the current operation in practice. Moreover, this PhD research illustrates the relevance and importance of incorporating real-life features in academic modelling approaches. Results show that safety constraints induce wait times, and cause additional travelling, picker blocking turns out to be minimised at the expense of additional setup time, and slow vertical travel results in additional travel and wait times. Consequently, ignoring these real-life features causes substantial performance inefficiencies. Robust policies for organizing operations efficiently are provided for a wide range of practical order picking systems, thereby including the effect of real-life features. Finally, time series forecasting techniques and the operational workload balancing model supports managers to define the daily resource capacity and how to allocate these resources. On average, these decision support tools are able to strongly reduce the daily over- or underestimated resources compared to the individual gut feeling and experience of supervisors. These insights and results can be used to integrate operational order picking planning problems, which may result in additionally reduced pick times of 15-20% in the real-life warehouse. The provided managerial insights and decision support tools increase the control and efficiency of order picking operations and reduces the research-practice gap.