Digging Deep: Finding and Maximizing the Throughput Capacity of Multideep Storage Systems
研究了多深存储系统中通过存储分配、重排和取货策略提高吞吐能力的方法，使用马尔可夫链和排队网络建模，为零售、备件和医药物流等行业提供优化建议。
Multideep storage systems are space-efficient storage solutions for a variety of industries and applications, such as in retail, spare parts and pharmaceutical logistics, and container terminals. They include robotic compact storage and retrieval (RCS/R) and multideep automated storage and retrieval (AS/R) systems. In such systems, multiple loads can be stored behind or above each other in a single lane, which leads to high space utilization. However, loads must be reshuffled if they block a requested load. This increases the command cycle time. We use Markov-chain models to estimate the steady state of the storage system and derive the travel time, which is then used in a closed queueing network to estimate the throughput capacity with a given number of robots. We built these models using four storage assignment strategies, three load reshuffling strategies, and two retrieval load selection strategies, incorporating the access frequency of the products and allowing multiple stored loads per product. Four strategy combinations are analyzed, including the current AutoStore strategy. We find that when information about the access frequency and number of loads per product is available, the throughput capacity can be increased significantly by properly storing and reshuffling loads to better positions. Based on the throughput models, we optimize the rack layout yielding maximum throughput capacity for two industry cases. Furthermore, we provide managerial insights on storage assignment, reshuffle, and retrieval load selection strategies for multideep storage systems. Supplemental Material: The online appendix is available at https://doi.org/10.1287/trsc.2024.0690 .