Robots for the Warehouse: Options Other than Kiva Systems

In previous postings, I’ve written about Amazon’s acquisition of Kiva Systems and on how the Kiva system works. At the CSCMP Annual Conference last week, Mike Roth, the VP of North American Fulfillment Services at Amazon, talked about using automation sparingly to preserve flexibility. He pointed, for example, to the exploding number of product categories Amazon has added over the last several years. If Amazon had built distribution centers (DCs) just for books, it would have been in trouble as e-readers have shrunk sales of printed books. Meanwhile, a DC designed for books would not have been well designed for other product categories like appliances or pet supplies. Amazon’s acquisition of Kiva Systems validates the idea that a new generation of more flexible distribution automation systems has a strong business case.

But the acquisition also raises a question: What now? Kiva is going to be very busy implementing its solution at Amazon distribution centers; it is not looking for new customers.

Fortunately, there are other choices. Swisslog offers a solution called AutoStore that, like Kiva, is also a robotic solution that automatically moves goods to pick/pack stations. You can watch a YouTube video of this solution, but below is a short description of the system along with my views of its pros and cons.

AutoStore makes use of standard-sized totes. The totes may contain just one stock keeping unit (SKU) or may be subdivided so that multiple SKUs are put in a tote. Once loaded with products, the totes are stacked on top of each other in a column. The stacks can be up to 16 bins high and reach a maximum height of 25 feet. Multiple columns of totes are tightly stacked into a rectangular grid.

There is a light infrastructure supporting the columns of totes. On top of the totes is a mezzanine deck that supports the robots. When a tote needs to be selected, a robot goes to the correct grid location on the mezzanine level and extends its clamper hands down through a hole in the deck to pick the top tote. If the required tote is at the bottom of a column with 16 totes, the robot will raise 15 totes to the mezzanine level and place them aside. The robot then selects the required tote and moves it to an induction station for routing to the correct pick station. The robot then puts the other totes back into their grid column stacked upon each other.

Clearly, the process of raising totes to the mezzanine level and placing them aside until the right tote is selected does not sound efficient. But that really does not matter. In comparison to humans, robots work cheap. All that matters is that the software controlling AutoStore correctly calculates how long the process of getting a tote to a pick station will take so that the system can keep the human at a station continuously busy. Further, the very layout that makes the robots work harder is what contributes to very high storage density.

I want to explore the flexibility of this solution in a bit more detail.

1. If throughput needs to increase, you can add more robots. This provides much more flexibility than conveyor/sortation systems, but is probably similar to the flexibility of Kiva.

2. If sales increase and more goods need to be stored, you can add more columns of totes to increase the size of the rectangle. The supporting infrastructure, and the mezzanine layer on top of the totes, is then extended to sit upon the new columns. This type of storage expansion flexibility is far better than traditional automation. It may also be better than what Kiva can achieve. With Kiva, only about 8 feet of vertical storage space are used. In theory, Kiva bots could operate on traditional warehouse mezzanine levels (as opposed to an AutoStore-style mezzanine deck), but when I’ve asked Kiva Systems if there are any DCs where their bots are doing this, the response I’ve received is, “Not at this point.” In contrast, I’ve heard that Kiva was de-installed at one site because of its vertical storage limitations. In theory, Amazon would have done thorough due diligence. Perhaps this problem has been solved.

3. Another form of flexibility is network-based flexibility. If a DC where the automation system is used has to close, can the equipment be reused at a different DC in the network? With traditional conveyors and sortation equipment, that is usually not possible. When the DC closes, the conveyors and sortation equipment are junked. With Kiva, you can march the bots into a truck and move them to a new DC at a very small cost. The AutoStore solution would fall somewhere in the middle. You could take down the columnar infrastructure and mezzanine deck and rebuild it at a different DC, but that does not sound easy or cheap.

The Kiva acquisition will lead companies to pay new attention to flexible material handling solutions, which I expect will benefit other solution providers like Swisslog.