Biased random-key genetic algorithm for warehouse reshuffling

Abstract

Due to its strategical importance, the efficient stock management in a warehouse presents several challenges that can be approached using optimization methods. In this universe, frequently explored problems are ambient dimensioning, department organization and layout, stock organization and layout, pilling design, product storage and recovery methodology. Design and operation imprecisions and failures can result in large delays in the product delivery or even in missing items in final client stocks. Among the main causes of missing items in inventories, there are the incongruity between storage capacity and refilling frequency; infrequency, delay, or nonexistence of product restitution in shelves; inexact or wrong inventories; storages with the inadequate organization, package disruption and scarce availability; poor storage layout and inefficient operational services. To determine the optimized product stocking is a problem frequently approached in the literature throughout the decades. However, the increasing need or changes in the storage, increase the importance of other problem: the sequence of movement to obtain a particular stock organization, given the current organization of the items. This problem is known as stock rearrangement, stock shuffling, or stock reshuffling. The optimization of package reshuffling in large warehouses directly impacts the profits. Large warehouses need, very frequently, to reorganize stock because of: seasonality, market changes, logistics, and other factors. Certain types of products have higher demand during specific periods of the year. Products on sale may leave the stock faster, new products may have higher output. All these are examples that justify a frequent stock reshuffling. Warehouse stock reshuffling consists of repositioning items by moving them sequentially. Several studies aim to solve reshuffling problems by applying exact methods. However, due to the complexity of the problem, only heuristics result in practical solutions. This study investigates how to optimize unit-load warehouse reshuffling in multiple empty locations scenarios. Traditional heuristics are reviewed and an evolutionary programming approach is proposed for the unit-load warehouse reshuffling problem. Experimental results indicate the proposed heuristic perform satisfactorily in terms of computational time and is able to improve solution quality upon benchmark heuristics.