Optimisation of order picking operations in a spare parts warehouse with dynamic order arrivals

Abstract

Warehouses are an important part of many supply chains. To improve operational performance, optimisation of the order picking operations is considered indispensable. Our research focuses on a manual, picker-to-parts, order picking system. In this setting, the order picking operations can be divided into serveral planning problems, i.e., the order batching, the picker routing and the batch scheduling problem. The order batching problem looks at which orders should be combined in a single pick tour throughout the warehouse. The picker routing problem decides on the most efficient path an order picker should follow to visit the storage locations of all items in his batch. Finally, the batch scheduling problem handles to assignment of batches to pickers, and the sequence in which the batches are picked. In the past years, research efforts shifted from the optimisation of these individual problems towards the optimisation of the integrated order batching, picker routing and batch scheduling problem (IBRSP). Since these planning problems are interrelated, solving the integrated problem leads to better overall results.

So far, studies have focused on a static problem setting, in which all orders are known at the start of the planning horizon. In practice, however, new and possibly urgent orders arrive throughout the day. To offer a good customer service level, these new orders should be included in the picking schedule as soon as possible. Therefore, our study extends the static IBRSP to account for dynamically arriving orders. By re-optimising previous schedules when new orders arrive, urgent orders can be handled very quickly in a cost-effective way.

A new metaheuristic algorithm, based on a large neighbourhood search, is developed to solve this optimisation problem. To test the algorithm, a series of problem instances is created. Discussions with practitioners indicated that companies only want to pick orders with due dates in the near future, to make sure these and any dynamically arriving orders are handled in time. Therefore, the effect of limiting the length of the planning period is investigated. Results show that there is indeed a trade-off between operational efficiency, expressed as the total order pick time, and service level, expressed as the total tardiness. By allowing the handling of orders with rather late due dates already at the start of the planning period, the order pick time can be reduced, with an increasing tardiness as side-effect. This shows that reducing the planning horizon, as currently occurs in practice, may indeed be required to guarantee a high service level. However, further experiments reveal that this issue can be largely mitigated by anticipating on future order arrivals. By including these expected orders from the beginning of the planning period, the service level actually improves when orders further into the future can be picked and scheduled very early, with only a minor reduction in operational efficiency.