Experimental investigation of shared storage assignment policies in automated storage/retrieval systems

IIE Transactions, August, 1999 by Sadan Kulturel, Nur E. Ozdemirel, Canan Sepil, Zafer Bozkurt

Thus, when precise information is available concerning the timing of storages and retrievals for individual unit loads of products, it is guaranteed to have substantial benefits from DOS-based policies. However, as stated in Francis et al. [14], it would be incorrect to assume that we cannot benefit from using shared storage in the absence of such precise information or under highly stochastic output functions. The concept can still be applied, but the performance of the DOS-based policy should be assessed in stochastic environments. Also, relatively little experience has been gained so far in applying the DOS-based shared policy to large scale problems "optimally". Hence, in contrast to many previous studies, most of which assume deterministic environments, it would be useful to study this policy experimentally in a stochastic AS/RS environment.

Several other AS/RS design issues have been addressed in the literature. Hwang and Lee [15], Chang et al. [16], Linn and Wysk [17,18] and Lee [19] are among the most recent ones.

3. The AS/RS environment under consideration

The AS/RS under consideration is a rectangular end-of-aisle system as schematized in Fig. 1. The system consists of a single AS/R machine serving two identical racks along the aisle. Although actual systems usually consist of several AS/R machines and aisles, a single m-crane system can be represented by m one-crane systems. Each storage rack has R rows and C columns. All 2 x R x C slots are equal in size, each holding a single unit load. The number of columns in the racks is determined by the inventory policy and varies from one case to another. The number of rows, however, is set as five and remains constant for all cases.

The system is bounded at the point where unit loads are transferred from/to the AS/R machine at a single I/O point. The AS/R machine is capable of moving both vertically and horizontally simultaneously. Its travel time is linearly proportional with the distance traveled. The shuttle time for the AS/R machine to load or unload a unit at the I/O point or a storage location is assumed to be constant. Interleaving is allowed. All storage functions are initiated with the AS/R machine at the I/O point, but a retrieval function may also be initiated upon completion of a storage process. The AS/R machine is assumed to be available during the entire simulation.

Multiple items are stored and retrieved in the system. The interarrival time between retrieval orders is assumed to be exponentially distributed with a known mean for each item. The continuous review, order quantity, reorder point (Q,r) model is employed as the inventory replenishment policy. Inventory status of each product type is initialized as zero.

When shared storage assignment policies are considered, the products and storage racks are both divided into three classes. ABC analysis is used to classify items (for turnover-based policy) or unit loads (for DOS-based policy). While class A items or units represent the fast moving products, classes B and C represent the slower and the slowest moving products. Unit costs of products are set accordingly. The specifics of forming the classes are discussed in subsequent sections. All the products that are in a particular ABC class have the same constant replenishment lead time. This time is taken as the average of the mean interarrival times of all products in that ABC class.