Kiva’s Robots Hit Their Stride…er, Slide
Kiva Systems may be winning its battle against the science-fiction robots.
The Woburn, MA, company is a newcomer to a hidebound business. It builds “mobile fulfillment systems” that are overturning all the traditions of warehousing by making the shelves move around, rather than the people. The moving is done by squat wheeled robots that maneuver under the shelves, pick them up, and trundle them over to a person at a picking station, who grabs whatever items are needed to fulfill the next order and boxes them up to be shipped to a retail store or a mail-order customer.
It’s amazing to watch dozens of these little orange robots working in concert, lifting and spinning and scooting without once colliding or spilling anything. (For a lighthearted look at Kiva’s system in action, see this YouTube video of the robots dancing the Nutcracker ballet.) But warehousing is a centuries-old industry in which the last major innovation was the forklift. Convincing a warehouse manager that he should hire a fleet of R2-D2s to staff his next facility is a tough sell.
“If you just go to the people who run these factories and warehouses and say ‘We’re going to have a bunch of robots help your workers,’ they are going to say ‘Obviously this is an East Coast pointy-head who doesn’t understand what a real warehouse is like,” says Mitch Rosenberg, Kiva’s vice president of marketing. To them, the word “robot” is likely to evoke visions of Robbie the Robot from Lost in Space. But show these same people a Kiva warehouse in action, Rosenberg says, and they start to understand the practical potential, he says—including potential productivity increases on the order of 200 to 300 percent.
There’s certainly room for improvement in an industry where most pick workers spend hours each day walking up and down miles of warehouse aisles. This isn’t a terribly efficient process, as online grocer Webvan learned the hard way. After raising some $800 million in venture capital, the onetime dot-com darling went bust in 2001 because—among other things—it never found a low-cost way to fulfill orders.
After all, if every customer’s order sends a pick worker scurrying around a warehouse just as if they were at the supermarket, most of the fabled efficiencies of Internet-based commerce shrivel away. “When you go grocery shopping, what percentage of the time are you actually putting what you’re looking for into your basket?” asks Rosenberg. “Most of the time you’re just walking and waiting to encounter it.”
It’s recognizing and grabbing items that supermarket shoppers—and warehouse stock-pickers—are really good at. Webvan executive Mick Mountz realized that if he could somehow make the bins full of stock come to the pick workers, rather than sending the workers to the bins, the humans in a warehouse would be able to spend more of their time being smart and less just walking around.
The idea came too late to save Webvan. But after the company folded, Mountz went enlisted Raffaello D’Andrea, an expert on automatic controls at the Swiss Federal Institute of Technology, and Peter Wurman, a computer scientist at North Carolina State University, to help him put the concept into action. The result was Distrobot, founded in 2003. (The company changed its name to Kiva Systems in 2005.)
The system Wurman, D’Andrea, and CEO Mountz have now spent five years perfecting is as much about software as it is about robots. A master inventory database knows which shelves in a warehouse hold which items. When an order comes in at the Staples warehouse for a box of green pens and a ream of paper, the system wirelessly instructs two robots to find the appropriate shelves, pick them up (a platform atop a large screw lifts up the shelf a couple of inches as the robot spins underneath), and carry them to a designated stock-picking station.
If both units arrive at the same time, the system tells one robot to wait until the other moves out of the way. The robots have infrared sensors to warn of nearby objects—but collisions are unheard of, because the units are continually reporting their positions to the master system. They keep track of their own locations by watching for small bar code stickers attached to the warehouse floor.
At the station, a laser pointer shows the stock-picker where to look for the green pens or the paper. The worker grabs the needed item out of the indicated bin, places it into a box, and punches a button to let the system know the action is done. When the whole order is assembled, the box coasts off to the loading dock on conveyor belts—or, in some cases, is carried there by other robots. Meanwhile, the system keeps track of which shelves are running low on supplies, and sends robots to carry them to replenishment stations as needed.
Rosenberg let me play stock-picker for a few minutes when I visited Kiva’s test warehouse a couple of weeks ago, and the process is fast-moving, like a video game. “Fifty percent of our customers use videos of Kiva in action in their recruitment videos that they put on local cable,” Rosenberg says. “People who are pick workers from other warehouses look at this and say ‘That looks a lot more relaxing and fun than running up and down those rows.’”
But when logistics managers responsible for equipping new warehouses first look at Kiva’s system, says Rosenberg, they tend to have a different reaction: shock. “The biggest challenge is that our system doesn’t look like anything they’ve ever seen,” he says. To head off misperceptions, Kiva engineers and salespeople try to downplay the fact that the company builds robots at all, frequently calling them “mobile drive units” instead. And during my visit Rosenberg went out of his way to draw a distinction between Kiva and other robot makers, especially Burlington, MA-based iRobot.
Unquestionably the Boston region’s best-known robotics company, iRobot sells gadgets that are famous for doing things, more or less autonomously: mopping and vacuuming floors, cleaning gutters and swimming pools, climbing over rubble piles and scouting out explosive devices in Iraq. Kiva’s robots, by contrast, work in groups of hundreds, almost like a hive of ants or bees; the individual robots have very little intelligence and no initiative.
“I describe it a parallel processing system for material,” says Rosenberg. “Our robots are just another piece of that system. That’s very different from the iRobot perspective, which is “These are rrrrobots!”
Gradually, Kiva is winning supporters in the material handling and logistics industry. Zappos.com announced in February that it’s installing a Kiva system at its Shepherdsville, KY, warehouse. More than 5 percent of all stock at Staples passes through a Kiva fulfillment system in Chambersburg, PA. Walgreens is ramping up its own Kiva-robot-run facilities, and Rosenberg says his firm has several other customers it’s not allowed to name. (Warehousing efficiency “is a subject of paramount strategic importance” to big retailers and e-commerce companies, he explains. “If you know that your competitor has a weakness in distribution, you know how to kill them.”) And today Kiva announced it has shipped its 1000th robot—I mean drive unit.
Kiva—which has raised more than $18 million in venture capital, mainly from Boston’s Bain Capital Ventures—has the capacity to build about 2,000 drive units per year from its Woburn facility. And if it can keep evading the robotics label, it may just get the opportunity to do so. “This isn’t a science experiment,” Rosenberg says. “We do real work here.”