Humanoid robots face a major test with Amazon’s Digit pilots

Announced amid a deluge of news at this week’s Delivering the Future event in Seattle was word that Amazon will begin testing Agility’s Digit in a move that could bring the bipedal robot to its nationwide fulfillment centers. It’s baby steps as these things go, and such early-stage deals don’t necessary mean something bigger down the road.

Take, for instance Agility’s Ford pilot, when the startup was exploring last-mile delivery as a potential way forward. Not too long after, the firm began focusing Digit’s output exclusively on warehouse and factory work.

In April of last year, Amazon named Agility one of the first five recipients of the company’s $1 billion Industrial Innovation Fund. While being included in the fund doesn’t guarantee that Amazon will utilize your technology down the road, it’s a pretty clear indicator that the retail giant is — at the very least — interested in its potential.

“The Innovation Fund is really about exploring what’s possible out there,” Amazon Robotics chief technologist Tye Brady told me in an interview this week. “It’s about understanding practical real-world examples, as well.”

The executive adds that, while Amazon Robotics has thus far exclusively traded in wheeled locomotion, legs present a good deal of possibility.

“We are interested in walking robots,” says Brady. “I find that very interesting, the ability to move on different terrains is interesting. We’re also interested in what works — and frankly what doesn’t work — about it. The humanoid form is really interesting. I don’t know if it’s a good thing or a bad thing. We’re experimentalists at heart. We’re gonna figure that out. We’re going to do a pilot and see how that works out.”

The company’s focus on wheeled AMRs (autonomous mobile robots) dates back to its 2012 purchase of Kiva Systems, whose platforms have formed the foundation for the whole of Amazon Robotics. There are currently 750,000 AMRs deployed across the company’s warehouse network. The company has launched non-AMR systems, as well, including picking arms like Sparrow, which was launched during the same event last year.

It’s difficult to overstate how profound an impact Amazon’s efforts have had on the rest of the industrial robotics space. For one thing, the company has turned up the pressure for the competition to automate in order to meet growing customer expectations of same- and next-day deliveries. For another, the decision to stop supporting Kiva customers outside the Amazon ecosystem led directly to the foundation of some of the industry’s biggest names, including Locus Robotics and 6 River Systems.

A system needs to demonstrate an increase in productivity in order for Amazon to integrate it into its growing robotics ecosystem. It’s less innovation for innovation’s sake, and more scoping out any possible advantage that will get goods to customers in less time. Including drones.

How, precisely, humanoid robots specifically and bipedal robots more generally might slot in remains to be seen. The other big hurdle there is that any new system needs to match the company’s almost unthinkable scale.

There are a number of startups vying to own the humanoid robotics crown at the moment, including 1X, Figure and Tesla. Agility’s Digit is the least human-looking of the bunch, but it’s got a ton of funding and a massive head start. The company also recently opened a new factory in Salem, Oregon, which it claims can produce up to 100,000 Digits a year once fully online.

There’s no shortage of excitement around the category, but proving things out at scale is another question entirely. Whether Digit succeeds or fails at the tasks laid out for it could have a profound impact on the trajectory of humanoid robots generally. Much like the Kiva Systems have proven a major catalyst for AMRs, if Amazon successfully rolls out Digit at scale, suddenly everyone will want to get their hands on some humanoid workers.

The biggest talking point around the form factor is the fact that humans build workspaces for other humans. That includes shelving heights, terrain, aisle width and the staircase, the bane of the ARM’s existence. From this standpoint a humanoid robot suddenly makes a lot more sense. The reality of things is that most companies operate in brownfield sites. That is to say their warehouses and factories generally aren’t built with specific automation solutions in mind. Humanoid robots slot nicely into a brownfield site.

Of course, Amazon has the resources to build any facility it wants, so it’s logical that many of its own robots are effectively working in greenfield sites. Those limitations are less of a concern for Amazon than much of the competition, but obviously if an effective system can slot into the existing workflow with minimal friction, that’s certainly ideal.

Brady confirms, however, that Digit isn’t the end-all, be-all of Amazon’s plans for mobile manipulation.

“When you start to bring [sensing, compute and actuation] together in interesting combinations, really unique things start to happen,” he says. “We’re world leaders when it comes to mobile robots. And now we are very much in the business of manipulating not only packages, but also objects. And to bring them together, it’s exciting to see all of the possibilities.”

That could mean alternate ways in. For instance, Amazon knows how to build both an AMR and a robot arm. If one were to effectively mount the latter to the former, they would have a kind of mobile manipulation on their hands.

“You see with the Agility robot — you can think of that as a mobile manipulator,” says Brady. “That has interest to us. The mode of mobility has particular interest to us, because we don’t happen to have done a lot of work in bipedal robots. But absolutely, we could combine that with identification systems, manipulation systems, sortation systems. Anything and everything we’ll do to innovate for our customer and improve safety for employees. We’re getting there with the core fundamentals.”

If for any reason Digit fails to stick the landing, that certainly doesn’t mean the end of it or bipedal robots generally. Perhaps it simply doesn’t sit comfortably in Amazon’s existing work flows. Maybe the robot’s not quite ready for Amazon scale or Amazon’s not quite in a place where Digit makes sense.

Regardless, it would be smart for anyone remotely interested in bipedal robots to sit up and take notice here. The pilots could well have a profound impact on the way we think of the category, going forward.