eight years ago, The Pentagon’s Defense Advanced Research Projects Agency organized a harrowing competition in which robots slowly struggled (and often failed) to perform a series of human tasks, including opening doors, operating power tools and driving golf carts. Videos of them clumsily performing the Darpa Robotics Challenge went viral.
Today, the descendants of those hapless robots are more capable and graceful. Several startups are developing humanoid robots that they claim can find jobs in warehouses and factories in just a few years.
Jerry Pratt, a senior research scientist at the Institute for Human and Machine Cognition, a nonprofit research organization in Florida, led the team that won second place in the 2015 Darpa Challenge. He’s now a co-founder of Figure AI, a company building a humanoid robot designed for warehouse work that today announced $70 million in funding.
If Darpa’s challenge took place today, the robot would be able to complete the challenge in about a quarter of the 50 minutes it takes his robot to complete the course, with very few accidents, Pratt said. “From a technology standpoint, there’s been a lot of enabling technology coming out recently,” he said.
Advances in machine learning over the past decade have enabled more advanced computer vision, which makes it easier for machines to navigate complex environments and perform tasks such as climbing stairs and grasping objects. Higher power-dense batteries due to the development of electric vehicles have also made it possible to pack enough energy into a humanoid robot to move its legs quickly to achieve homeostasis—that is, to achieve homeostasis without slipping and falling. Or stay steady when slipping and falling. Misjudge a step like a human being.
Pratt said his company’s robots are taking their first steps around a mock warehouse in Sunnyvale, California. Figure CEO Brett Adcock believes that humanoid robots could be built for the same cost as cars, as long as there is enough demand to ramp up production.
If Adcock is correct, the field of robotics is approaching a pivotal moment. You’re probably familiar with the dancing Atlas humanoid robots that have been getting likes on YouTube for years. They’re made by Boston Dynamics, a pioneer in legged locomotion, which built some of the humanoids used in the Darpa competition and showed that it’s possible to make capable robots in a human shape. But these robots were expensive—the original Atlas was worth millions of dollars—and lacked the software to make them autonomous and useful.
Figure isn’t the only company betting that humanoid robots are coming of age. Others include 1X, Apptronik and Tesla. Tesla CEO Elon Musk visited the original Darpa Robotics Challenge in 2015. The fact that he’s now keen to build a humanoid himself suggests that some of the technology needed to build such machines is finally feasible.
Jonathan Hurst, a professor at Oregon State University and co-founder of Agility Robotics, also participated in the Darpa Challenge, demonstrating a walking robot he built. Agility has been working on legged robots for some time, but Hurst said the company took a physics-first approach to locomotion rather than replicating the mechanics of human limbs. While its robots are humanoid, their legs look like they were inspired by ostriches.