On the dynamic management of order pickers: modelling, solution approach and case studies

With the current trends towards faster response through e-commerce, there has been increased pressure on efficient management of urban warehouses which operate in an environment with stringent lead times and dynamic patterns of orders. In this study, we consider a warehouse setting that makes use of an IoT platform, which provides the ability to track and dynamically route or re-route the order pickers. Due to continuously arriving orders and the need for short lead times, such a dynamic picking system calls for prompt order batching, picker assignment, and rerouting decisions with frequent decision epochs. To address this challenging problem, we propose exact and heuristic solution approaches both utilizing decomposition and set partitioning formulation. For the instance sizes that the exact approach is not reliable in terms of solution time, the heuristic satisfies both solution quality and agile response requirements. We also develop a simulation tool to compare static and dynamic decision making environments in terms of different KPI's such as order fulfilment time and minimum number of pickers needed. We conduct comprehensive experiments using randomly generated instances and two case studies based on an e-commerce and a grocery warehouse. Our experiments show that a dynamic system always outperforms a static one in e-commerce/dark store environments with fewer order lines and smaller warehouses. We also show that in a grocery warehouse, where orders and warehouse size are much larger, effectiveness of a dynamic system depends on a dynamic management of the picking capacity.