Automated Warehouse Systems: A Guideline for Future Research

This study aims to provide a comprehensive tool for the selection, design, and operation of automated warehouse systems considering multiple automated storage and retrieval system (AS/RS) options as well as different constraints and requirements from various business scenarios.

We first model the retrieval task scheduling problem in crane-based 3D AS/RS with shuttle-based depth movement mechanisms. We prove the problem is NP-hard and find an optimality condition to facilitate the development of an efficient heuristic. The heuristic demonstrates an advantage in terms of solving time and solution quality over the genetic algorithms and the other two algorithms taken from literature. Numerical experiments illustrate that when a company tends to have multiple short planning horizons with small task batches (i.e., aims to increase the responsiveness level), adding more shuttles is helpful. However, if a company has a long planning horizon with a large task batch size, having faster cranes is more efficient to reduce the makespan.

We then focus on the impacts of the number of shuttles, operational mode, storage policies, and shuttle dispatching rules on the expected cycle time of a tier-to-tier shuttle-based storage and retrieval system. The system is modeled as a discrete-time Markov Chain to derive the shuttle distribution under each scenario create the expected travel time models. Numerical experiments indicate that class-based storage is always better than the random storage policy. The best shuttle dispatching rule under each combination of the number of shuttles, operational mode, and storage policy can be quickly identified through the expected cycle time models which are simple and computation friendly.

At last, we study the warehouse design problem considering the choice, design, and operation of 2D AS/RS and 3D AS/RS in a systematic way. The warehouse design problem under consideration aims to reduce the investment while satisfying different business needs measured by the desired throughput capacity. We propose a branch-and-bound algorithm to conquer the computational challenges. With the developed algorithm, an optimal warehouse design can be obtained under different application environments, characterized by the desired throughput capacity, inventory level, and demand rate of each SKU.