At OSU’s Whyte Track and Field Center, Cassie, a bipedal robot created by the Oregon State University (OSU) College of Engineering and manufactured by an OSU spinoff company called Agility Robotics, recently ran 100 meters without falling in 24.73 seconds. The robot set a Guinness World Record for running a bipedal robot’s fastest 100 meters.
The bipedal robot’s average speed was just over 4 m/s, which was a little slower than its top speed because Cassie had to start from a standing position and end the sprint in the same place, which proved to be difficult to develop.
As with how taking off and landing are harder than actually flying a plane, starting and stopping while standing is more difficult than running, according to OSU AI Professor and project collaborator Alan Fern. This 100-meter result was only possible because advanced artificial intelligence for hardware control and mechanical hardware design worked closely together.
Damion Shelton, co-founder and CEO of Agility Robotics, will serve as the keynote speaker at RoboBuiness on October 19 and 20, which is organized by WTWH Media, the parent company of The Robot Report. Building Human-Centric Robots for Real-World Tasks is the title of Shelton’s keynote address on October 20 from 9 to 9:45 AM. Digit will be shown off by Agility Robotics during the session and on the expo floor. They will also give a sneak peek at the next version of Digit.
With a top speed of 43 mph, ostrich-like knees, and no cameras or external sensors, Cassie is blind to its surroundings and is not autonomous. Ostriches are the world’s fastest runners.
Since Cassie’s debut in 2017, OSU students have been researching machine learning options in the Dynamic Robotics and AI Lab at Oregon State, where Cassie has been employed. They learn how to move up and down stairs as well as run and walk. The Dynamic Robotics and AI Lab combined physics with AI techniques that are typically used with data and simulation to develop robot control.
Using a computer technique called parallelization, which allows multiple processes and calculations to take place simultaneously, the team was able to reduce Cassie’s simulated training, which is equivalent to a year, to just one week. This enables Cassie to experience several training scenarios at once.
In 2021, Cassie traversed the OSU campus untethered and on a single battery charge, covering 5 kilometers in just over 53 minutes. Cassie controlled her running gait on the outdoor terrain while she was out for a run using machine learning.
The bipedal robot was made with the help of a $1 million grant from the Defense Advanced Research Projects Agency (DARPA) and more money from the National Science Foundation. Jonathan Hurst, a professor of robotics at Oregon State University and the chief technology officer and co-founder of Agility Robotics, oversaw the project.
It may be the first bipedal robot to learn to run, but Hurst predicts that it won’t be the last. “I think that control strategies like these will play a significant role in the development of robotics. This race’s potential is what makes it exciting. This 100-meter dash outperforms other control techniques in the relatively new field of using learned policies for robot control. I believe things will move along more quickly from here.
