Ph.D. Thesis

In December 1996, Jeroen van den Berg earned a Ph.D. degree with his thesis entitled: "Planning and control of warehousing systems". Below we give the summary of the thesis.

Read the Management Outlook Report based on this thesis.

Summary

This thesis discusses the planning and control of warehousing systems. Planning relates to the warehouse management decisions that affect an intermediate time period (one or multiple months), while control concerns those decisions that affect a short time period (minutes, hours).

We start this thesis with a classification of subproblems that may be distinguished in warehouse management. We review methods and models that have been presented in the literature. Subsequently, we investigate several planning and control problems more thoroughly.

First we study the forward/reserve problem. This is a planning problem, where we distinguish the forward area, which is used for efficient order-picking, and the reserve area, which is used for replenishing the forward area. We present a method that determines which products in which quantities should be stored in the forward area, such that the expected amount of work involved in order-picking and replenishing the forward area is minimized. The remaining goods are stored in the reserve area.

The second planning problem that we investigate, concerns finding a class-allocation which minimizes the expected travel time for the storage and retrieval of goods. We address a situation where storages and retrievals are performed in single command cycles. We develop a solution method and demonstrate possible savings, using order data from the automated storage/retrieval system in the Yamaha Spare Parts Distribution Center, located in The Netherlands.

In the remainder of the thesis we focus on (short term) control issues. First we look at the problem of finding a pick sequence for multiple orders in a carousel system, such that the rotation time of the carousel during the order-picking process is minimized. We develop two solution methods. One method concerns the situation where the sequence of the orders is fixed and the other one deals with the situation where the sequence of the orders may be chosen arbitrarily.

Subsequently, we consider a number of control problems that concern automated storage/retrieval systems. The first control problem that we investigate, is the problem where the storage/retrieval (S/R) machine should reside when the system is idle. We select a position which minimizes the expected travel time to the position where the first request after the idle period will take place. This will reduce the response times (the time until an incoming request has been fulfilled) of the storage and retrieval requests.

The second problem with respect to the control of automated storage/retrieval systems that we address in this thesis, is the problem of finding a sequence in which storage and retrieval requests should be performed in order to minimize the travel time of the S/R machine. Minimizing travel time may lead to an improved throughput capacity and reduced response times. We distinguish two approaches. The first approach is to submit incoming requests directly to the system, after which these may be performed immediately. The second approach is to distinguish blocks of requests which are supplied to the system one after the other. The requests in a block have to be completed before one may commence with the next block. We study the first approach in a simulation study in which we compare old and new control policies. For the second approach, we present a method that finds a sequence of the storage and retrieval requests that minimizes the travel time when using the dedicated storage policy.

For the solution methods to be applied in practice using (personal) computers, it is necessary that these do not require excessive computation time. We have shown for all methods that these find a solution in polynomial time. This implies that the computation time will not escalate when the sizes of the problems increase. Furthermore, the quality of the solutions is guaranteed. For some problems the developed methods find optimal solutions, while for the other problems it has been shown that the solutions will be no more than a pre-specified amount from the optimum. We also evaluated the possible savings of the methods by comparing these with procedures that are typically used in practice. Finally, to facilitate the implementation of the developed methods in practice, it is important that the required input data is easily obtained. This holds for all methods. Examples of the required input data are: historical demand data, machine velocities, positions of goods in the warehousing system, et cetera.