New benchmarks and optimization model for the Storage Location Assignment Problem

The Storage Location Assignment Problem (SLAP) is of primary significance to warehouse operations since
the cost of order-picking is strongly related to where and how far vehicles have to travel. Unfortunately, a
generalized model of the SLAP, including various warehouse layouts, order-picking methodologies and
constraints, poses a highly intractable problem. Proposed optimization methods for the SLAP tend to be
designed for specific scenarios and there exists no standard benchmark dataset format. We propose new SLAP
benchmark instances on a TSPLIB format and show how they can be efficiently optimized using an Order
Batching Problem (OBP) optimizer, Single Batch Iterated (SBI), with a Quadratic Assignment Problem
(QAP) surrogate model (QAP-SBI). In experiments we find that the QAP surrogate model demonstrates a
sufficiently strong predictive power while being 50-122 times faster than SBI. We conclude that a QAP
surrogate model can be successfully utilized to increase computational efficiency. Further work is needed to
tune hyperparameters in QAP-SBI and to incorporate capability to handle more SLAP scenarios.