Picking Operations in Warehouses With Dynamically Arriving Orders: How Good is Reoptimization?

ABSTRACT E‐commerce operations are essentially online, with customer orders arriving dynamically. However, very little is known about the performance of online policies for warehousing with respect to optimality, particularly for order picking and batching operations, which constitute a substantial portion of the total operating costs in warehouses. We aim to close this gap for one of the most prominent dynamic algorithms, namely reoptimization (Reopt) , which reoptimizes the current solution each time a new order arrives. We examine Reopt in the Online Order Batching, Sequencing, and Routing Problem (OOBSRP) , in both cases when the picker uses either a manual pushcart or a robotic cart. Moreover, we examine the non‐interventionist Reopt in the case of a manual pushcart, wherein picking instructions are provided exclusively at the depot. We establish analytical performance bounds employing worst‐case and probabilistic analysis . We demonstrate that, under generic stochastic assumptions, Reopt is almost surely asymptotically optimal and, notably, we validate its near‐optimal performance in computational experiments across a broad range of warehouse settings. These results underscore Reopt 's relevance as a method for online warehousing applications.