An algorithmic view of how best to organize stock in an automated warehouse without prior knowledge of the popularity (and thus, request frequency) of items stored therein.
As the prevalence of E-commerce continues to grow, the efficient operation of warehouses and fulfillment centers is becoming increasingly important. To this end, many such warehouses are adding automation in order to help streamline operations, drive down costs, and increase overall efficiency. The introduction of automation comes with the opportunity for new theoretical models and computational problems with which to better understand and optimize such systems.
These systems often maintain a warehouse of standardized portable storage units, which are stored and retrieved by robotic workers. In general, there are two principal issues in optimizing such a system: where in the warehouse each storage unit should be located and how best to retrieve them. These two concerns naturally go hand-in-hand, but are further complicated by the unknown request frequencies of stored products. Analogous to virtual-memory systems, the more popular and oft-requested an item is, the more efficient its retrieval should be. In this paper, we propose a theoretical model for organizing portable storage units in a warehouse subject to an online sequence of access requests. We consider two formulations, depending on whether there is a single access point or multiple access points. We present algorithms that are O(1)-competitive with respect to an optimal algorithm. In the case of a single access point, our solution is also asymptotically optimal with respect to density.