I am pleased to announce the forthcoming publication of a new paper in the European Journal of Operational Research entitled A constructive aisle design model for unit-load warehouses with multiple pickup and deposit points. I am co-author with my colleagues Ömer Öztürkoğlu and Russell D. Meller. This is the second paper from Ömer’s dissertation.
This paper generalizes previous work on aisle designs for unit-load warehouses to ask how designs should change in the presence of multiple points of access or “pickup and deposit points.” Ömer’s first paper produced optimal designs for a single point of access, which would be appropriate when travel in the picking space begins from and ends at a central point—a stretch wrap machine or a small group of shipping doors, for example. Our new work allows multiple points located anywhere on the perimeter of the warehouse. The paper describes a computational approach that determines the orientations of one or two cross aisles and the angles of picking aisles in order to minimize expected travel distance to the access points. Here is an example design:
Notice now the model produces a “fishbone” structure from each access point, with the middle region being shared by both. Notice also that locations in the farthest picking regions from each point have distances less than rectilinear (the standard distance in a traditional warehouse). Expected distance to a randomly selected point in this warehouse is about 11% lower than in a traditional design, but at the cost of about 10% more floor space.
The design above also shows a fishbone like structure from both access points, but distances appear to be much closer to rectilinear here. In fact, expected distances in this design are only about 6% lower than in a traditional design.
This is a design for “crossdocking” operations, in which access points are on opposite sides of the warehouse. An actual crossdock would have many more access points, of course, representing dock doors. Again, the model produces two fishbone like structures, but in a way that outlying regions still have lower expected travel distances than rectilinear. However, the extra space required for the two cross aisles overpowers the slight benefit of the aisles, so this design is actually worse than a simple traditional design with one cross aisle.
The most pleasing part of this work to me was watching the model produce designs that we never imagined. That is, the model was teaching us about good designs—designs that defied our intuition. Three cheers for model building.