Delft University of Technology
Faculty Mechanical, Maritime and Materials Engineering
Transport Technology / Logistic Engineering



R. de Feijter Operational procedures for order picking
Literature survey, Report 99.3.LT.5258, Transport Technology, Logistic Engineering.


Warehousing involves all movement of goods within a warehouse. These movements are affected by management decisions. Management decisions within a warehouse can be divided into three categories: The control of warehousing operations covers the following decision types on the operational level: Closely related to these decision types on the operational level are some decision types on the tactical and strategic level: The layout of a warehouse (in this context: The way in which storage and retrieval are organized) can be subdivided into: The part-to-picker systems can be subdivided into systems performing single command cycles, systems performing dual command cycles, systems performing multi command cycles or carousel systems. Picker-to-part systems are always multi command cycle systems. Single command cycles perform one storage or retrieval per cycle. Dual command systems perform one storage and one retrieval per cycle. Multi command systems perform multiple storages or retrievals per cycle.

Single command systems do not need batching or sequencing methods. The dwell-point is preferably dynamically located at the location that minimizes the expected S/R-machine travel time from the dwell point to the points of need. In most situations the input location is also a good location for the dwell-point. The storage policy can be dedicated, class-based or random. In balanced systems the duration-of-stay based dedicated policy prevails, whereas in other systems the turnover (or cube-per-order index) based class policy provides better results.

Dual command systems, like single command systems, do not need batching methods. Sequencing methods are used to determine the best way to combine a storage and a retrieval request. In most systems this means choosing a retrieval request from a list to link it to a given storage request, because the storage requests have to be performed in a first-come-first-serve manner.
The optimal dwell-point location in a dual command cycle is the input point. This is especially true when all performed cycles are dual command cycles. As for most single command systems, the best storage policy for dual command cycles is the turnover based class policy.

In multi command part-to-picker systems, most batching methods are composed of two major parts: the seed selection rule and the additon rule. There are many alternatives for both rules. When creating a batch, including all assigned orders in the seed provides better results than keeping the original seed. An order batching rule which uses the economic convex hull in the seed selection and in the order addition rule shows the best results. Routing / sequencing heuristics try to determine a minimal tour passing through all the locations that have to be visited. Some possible heuristics are the nearest-neighbour, spacefilling curve, ½-band insertion and the hull heuristic. The best performing heuristic is the hull heuristic. The results of this heuristic can be further improved by an improvement algorithm.
The desirable dwell-point position and the storage policy are the same as in the case of a dual command cycle. The class-boundaries however do not have to be L-shaped as in the case of a dual command system.

The construction of most batching algorithms for a multi command picker-to-part system uses the same basic rules as a part-to-picker system: the seed-selection rule and the order addition rule. An order batching algorithm which uses a minimum additional aisle method in the seed selection rule and in the order addition rule provides the best results in this situation.
The routing of the order picking tour can use a traversal, a return policy or a combination of both. This tour can either be created by a heuristic or optimally determined. A well performing method, especially when multiple cross-aisles are present, is a heuristic that uses a combination of both policies. When the aisles are wider, the largest gap method (return policy) tends to become more rewarding.
The desirable dwell-point position and the storage policy are once again the same. If the orders are picked in a traversal manner, the A-zone (as in ABC-curve) covers the aisles closest to the I/O-point. However, if the orders are picked in a return manner, the A-zone covers the outer end(s) of the aisles.

In carousel systems, the batching of orders can be done with simple algorithms because the distance between the locations is one-dimensional. The sequencing of orders in a carousel system has a special property (a cycle can stop at any position) that provides an opportunity to increase the efficiency of the system. An algoritm called Matchtree effectively performs this task.
When the random storage policy is used, the dwell point can be arbitrarily chosen. The class-based storage policy is, again, the most promising.

Reports on Logistic Engineering (in Dutch)
Modified: 2000.06.24; logistics@3mE.tudelft.nl , TU Delft / 3mE / TT / LT.