On May 11, 2022, Cassie, a bipedal robot developed at Oregon State University, ran 100 meters in 24.73 seconds at the university’s track, setting a Guinness World Record for the fastest 100 meters by a bipedal robot. The run started from a standing position and ended back in a standing position with no falls.
Cassie has no torso, head, or arms; it is essentially a pair of ostrich-like legs with knees that bend like a bird’s. It carries no cameras or external sensors and balances using only its own internal sensing. The robot was developed in the lab of robotics professor Jonathan Hurst, originally under a DARPA grant, and is the basis for the commercial robots built by Oregon State spinout Agility Robotics.
What made the record notable was the method. Cassie learned to run using machine learning rather than hand-coded gait controllers, and Oregon State described it as the first bipedal robot to use machine learning to control a running gait on outdoor terrain. The same robot had earlier completed an untethered 5K on campus on a single battery charge in about 53 minutes.
For a general reader, the record is a vivid marker of how fast learned controllers closed the gap on dynamic, athletic movement that long stymied two-legged machines, and it traces a direct line from a university lab to a commercial humanoid company.