Monday, August 28, 2006

What is a V-shaped cross aisle?

Our research began by asking the following simple question, "Why does it have to be straight?" -- meaning, why does a cross aisle have to be straight? Of course, the answer is, "It doesn't!" After some modeling and experimentation, we developed designs that look like the one above. On the left is a traditional warehouse; on the right, a warehouse with an optimal cross aisle. Expected travel distance to make a pick in this warehouse is 10% less than in the traditional warehouse, even accounting for the lost space due to inserting the cross aisle.

The main insight here is that the V-shaped cross aisle allows workers to travel into the picking space as quickly as possible, where they can then branch off and get their items. There are three travel paths: along the bottom aisle and up (this is the fastest way to get some items), along the cross aisle and down, and along the cross aisle and up.

We would like to hear your opinions on

1. Safety issues for forklift drivers. In particular, some picks would require workers to have to make more than a 90 degree turn (travel path #2 above).

2. Congestion concerns? How wide would the cross aisle have to be to remove this concern?

3. Would the reduced storage density ever be worth it? This design offers the classic tradeoff: lower labor costs for retrieval, in exchange for a slightly larger warehouse.


Modeling assumptions

Many of the questions we have gotten from industry folks are in response to a press release by the University of Arkansas, which necessarily did not contain many of the assumptions of the research. Here, for the record, are our assumptions:

1. These designs are for unit-load picking only, not for order picking environments in which workers must retrieve more than one item per tour. We are working on the latter (more difficult) problem now, and will post results when we have them.

2. All travel begins and ends at a single pickup and deposit point, which may be the point of entry to a picking module (flow rack or bin shelving, for example) or a stretch-wrap machine. We are working on more general material flow models that will account for multiple P&D points, and will post results when we have them.

3. "Percent improvement" records only the percent reduction in expected travel distance to a randomly chosen location. It does NOT consider time to load/unload, time to retrieve pick lists, or time to visit the potty! You can estimate your level of savings by multiplying the percent improvement by the percent of time your workers spend traveling.

4. Every location is equally likely to be visited, which is approximately equivalent to randomized or closest open location.

5. We have ignored the following issues: congestion, spacing of building columns, worker disorientation, and numbering of rack locations. All are real issues that deserve consideration: I will address them in another post.


What are fishbone aisles?

Fishbone aisles comprise a new design for unit-load warehouses that significantly (on the order of 20%) reduces the expected distance to travel to make a pick or stow. The idea is a simple one: Imagine a warehouse with single-deep selective rack having a single pickup and deposit (P&D) point at the bottom center. Each pick or stow begins and ends at the P&D point. The "spines" of the fishbone aisles extend diagonally from the P&D point with "bones" radiating from them, at appropriate angles.

The idea is to approximate travel "as the crow flies," which would be straight-line distance from the P&D point to the picking location. This cannot be done, of course, but a little thought and experimentation with the figure above should convince you that fishbone travel is very close to "travel by flight." By contrast, a traditional warehouse forces rectilinear travel (travel at right angles), which is far from optimal.

Our mathematical models suggest that, for typical warehouse sizes in industry, expected distance to make a pick in a fishbone warehouse is about 20% lower than for a traditional warehouse.

Your thoughts?

- What issues do you see with such a design in practice? (Please read my post on Modeling Assumptions before responding.)

- Congestion concerns? How might they be handled?

- We have thought about worker disorientation --- might they simply get lost?! Keep in mind that we assume they are going to store or retrieve only a single item.

- Construction issues? Spacing of building columns has been mentioned at some of our presentations. Why couldn't the columns be built around the racks, instead of the other way around?