Today, a robotics startup referred to as Figure is unveiling “the world’s first commercially viable normal goal humanoid robotic,” referred to as Figure 01. Shown within the rendering above, Figure 01 doesn’t but exist, however based on this morning’s press launch, it’s going to “have the flexibility to suppose, study, and work together with its setting and is designed for preliminary deployment into the workforce to handle labor shortages and over time cleared the path in eliminating the necessity for unsafe and undesirable jobs.” Which sounds nice, when (or if) it occurs.
We are usually skeptical of bulletins like these, the place an organization comes out of stealth with formidable guarantees and a few spectacular renderings however little precise {hardware} to reveal together with them. What caught our eye in Figure’s case is their exceptionally certified robotics group, led by CTO Jerry Pratt. Pratt spent 20 years on the Florida Institute for Human and Machine Cognition (IHMC), the place he led the group that took second place on the DARPA Robotics Challenge Finals. Working with DRC Atlas, NASA’s Valkyrie, and most not too long ago Nadia, IHMC has established itself as a frontrunner in robotic design and management. And if anybody goes to take a helpful humanoid robotic from an engineering idea to business actuality, these are the parents to do it.
Figure was based in 2022 by Brett Adcock, who additionally based Archer Aviation, which has efficiently constructed and is at the moment flight testing a business passenger eVTOL plane. The firm bought its begin self-funded, however is closing a Series A inside the subsequent couple of weeks. Over the previous yr, Figure has employed greater than 40 engineers from establishments that embrace IHMC, Boston Dynamics, Tesla, Waymo, and Google X, most of whom have important prior expertise with humanoid robots or different autonomous methods.
“It’s our view that that is the very best humanoid robotics group on the market,” Adcock tells IEEE Spectrum. “Collectively, the group has in all probability constructed 12 main humanoid robots,” provides CTO Jerry Pratt. “We’ll have experience in nearly each a part of the 1000’s of issues that it’s essential to do for humanoids.” Pratt says that originally, Figure isn’t anticipating to make use of a lot in the best way of latest know-how with their robotic—it’s not primarily based round some secret actuator know-how or something like that. “It’ll be a brand new design, with actually strong engineering.”
The robotic that Figure is working in direction of is a “commercially viable normal goal humanoid,” which goes to look one thing like this:
https://www.youtube.com/watch?v=b37rQZ4maPo
Obviously, the above video (and the entire robotic photographs on this article) are renderings, and don’t present an actual robotic doing actual issues. However, these renderings are primarily based on a CAD mannequin of the particular robotic that Figure plans to construct, so Figure expects that their last {hardware} will probably be similar to what they’re exhibiting at this time. Which, if that’s the way it finally seems, will probably be spectacular: it’s a really slim type issue, which does put some limits on its efficiency. The last robotic will probably be absolutely electrical, 1.6m tall, weigh 60kg with a 20kg payload, and run for five hours on a cost.
Figure
“Having a humanoid type—it’s actually robust doing the packaging,” explains Pratt. “In normal, with know-how that’s out there at this time, you may hit someplace round 50%-60% on most human specs, like levels of freedom, peak speeds and torques, issues like that. It gained’t be superhuman; we’ll be specializing in real-world purposes and never making an attempt to push the bounds of pure efficiency.” This focus has helped Figure to constrain its design in pursuit of business utility: you want a robotic to be slim as a way to work in areas designed for people. With this design philosophy, you’re not going to get a robotic that can have the ability to do backflips, however you’re going to get a robotic that may be productive in a cramped workcell or stroll safely via a crowded warehouse.
This relates again to the explanation why Figure is constructing a humanoid robotic within the first place. The added complexity of legs needs to be justified by some means, and Figure’s perspective is that constructing a robotic with out legs that has the required vary of movement to do what it must do in a human workspace could be complicated sufficient that you just would possibly as effectively simply construct the robotic with legs anyway. And doing so opens up the chance (or maybe the crucial) to generalize. “If you’re making humanoids, you just about need to get to normal goal,” says Pratt. “For only one software, there’ll in all probability all the time be a devoted robotic that’ll be higher.”
“Whenever individuals say, ‘why are you doing humanoids, why don’t you do one thing higher?’ Maybe we’re simply restricted by our creativeness, however I can not consider what that might be. With at this time’s know-how, it’s unattainable to get to pretty much as good as a human, so I feel the technique of getting as near a human as you may is a wonderfully legitimate one.” —Jerry Pratt
Figure, like most different corporations engaged on business humanoids, sees warehouses as an apparent entrypoint. “The warehouse makes it simpler on us,” says Adcock. “It’s indoors. There aren’t any prospects round. There are already AMRs [autonomous mobile robots] and cobots [collaborative robots] working round people. And there’s a warehouse administration software program system to handle high-level behaviors. Our guess right here is that if we will determine easy methods to get one software that’s sufficiently big, and deploy sufficient robots, we will add new software program as we go to do extra issues and over time manufacture actually excessive volumes and get the robotic to be reasonably priced.” Adcock acknowledges that the robotic should make monetary sense available in the market that it’s concentrating on. That is, if it’s going to take the place of human labor in a warehouse, it should be aggressive in value with human labor, which will probably be a critical problem which will (at the least initially) depend on some choice for human teleoperation to maximise reliability.
Figure believes that it has a sensible shot at being the primary firm to really commercialize a normal goal humanoid robotic, though each Adcock and Pratt identified that there’s a lot potential demand that they’re not particularly fearful about competitors. “I feel it’s only a query of getting there,” Pratt tells us. “There’s room for a number of corporations to only get there, and I feel we might be considered one of them.”
I don’t suppose anyone’s going to dispute that normal goal humanoids will occur. I feel it’s only a matter of after they’ll occur, and what that can appear like. —Brett Adcock
Getting there, as Figure makes express in its grasp plan, “would require important developments in know-how.” Here is what the corporate believes it might want to make occur:
- System Hardware: Our group is designing a completely electromechanical humanoid, together with arms. The purpose is to develop {hardware} with the bodily capabilities of a non-expert human. We are measuring this by way of vary of movement, payload, torque, value of transport and velocity, and can proceed to enhance via speedy cycles of improvement, every cycle as a part of a continuum.
- Unit Cost: We’re aiming to scale back particular person humanoid unit prices via high-rate quantity manufacturing, working in direction of a sustainable financial system of scale. We are measuring our prices via the absolutely burdened working value/hour. At excessive charges of quantity manufacturing, I’m optimistic unit value will come all the way down to reasonably priced ranges.
- Safety: It’s important that our humanoids will have the ability to work together with people within the office safely. We will design them to have the ability to adhere to business requirements and company necessities.
- Volume Manufacturing: We foresee not solely needing to ship a top quality product, but in addition needing to ship it at an exceptionally excessive quantity. We anticipate a steep studying curve as we exit prototyping and enter quantity manufacturing. We are getting ready for this by being considerate about design for manufacturing, system security, reliability, high quality, and different manufacturing planning.
- Artificial Intelligence: Building an AI system that allows our humanoids to carry out on a regular basis duties autonomously is arguably one of many hardest issues we face long-term. We are tackling this by constructing clever embodied brokers that may work together with complicated and unstructured real-world environments.
This all sounds very compelling, nevertheless it’s essential to notice that so far as we’re conscious, Figure has not executed any of it but. They have targets and goals, they’re designing in direction of these targets and goals, they usually can do their finest foresee and anticipate a few of the challenges that lie forward and plan and put together to the extent that it’s doable to take action. However, at this level it’s untimely for us (or anybody) to guage whether or not or not the corporate will probably be profitable, since they nonetheless have quite a lot of issues to determine. To be clear, I imagine that Figure believes that it could actually, ultimately, do what it says it plans to do. My criticism right here is generally that the corporate is doing extra telling than exhibiting—traditionally, this has not been a great technique for robotics, which tends to be susceptible to the underdelivery of overpromises.
Figure does acknowledge that that is going to be a very, actually laborious course of, and that the corporate faces “excessive threat and intensely low possibilities of success,” which is an attention-grabbing assertion within the midst of what’s in any other case relatively quite a lot of uniformly optimistic hype, for lack of a greater phrase. And Adcock understands that the loftier targets (just like the “shopper family” and “off-world” purposes) are prone to take some time, telling us that the corporate “will get actually excited concerning the potential right here over a multi-decade lengthy interval.”
“Part of the guess we’re making is that we’re not too early, and that the know-how to do that exists.” —Jerry Pratt
Figure
So what’s the precise state of Figure’s robotic proper now? “We simply completed our alpha construct,” Adcock says. “It’s our first full-scale robotic. We’re constructing 5 of them. We hope it’s going to begin to take its first steps inside the subsequent 30 days. And now we’ve began on our second era {hardware} and software program model that we’ll have accomplished this summer season.” It’s an aggressive timeline, and Figure hopes to be creating a brand new main model of each {hardware} and software program each 6 months, indefinitely. “We suppose we’re positioned effectively,” Adcock continues. “Hopefully we’ll make our massive milestones this yr, and be ready to be first to market. We’re going to strive. We’re going to maneuver as quick as we probably can to hit that purpose.”
