When it comes to athletic bipedal robots, nothing compares to Boston Dynamics’ Atlas robot. We’ve seen Atlas run, jump, dance, and do parkour. Researchers at Oregon State University (OSU) are getting in on the fun, showcasing its Cassie bipedal robot completing a 5K.
Cassie was invented at OSU and produced by OSU spinout company Agility Robotics. While Cassie’s accomplishment pales in comparison to some of Atlas’ athletic feats, it was able to complete the 5K at the OSU campus in just over 53 minutes on a single battery charge. You can watch a video of Cassie completing the 5K below.
A human appears to be controlling the direction in which Cassie walks, but Cassie takes each step on its own. During the 5K, there was about six and a half minutes of down time due to two resets. The robot’s computer overheated once, and Cassie fell once after it was asked to execute a turn at too high a speed.
With knees that bend like an ostrich’s, Cassie taught itself how to run using deep reinforcement learning. Running requires dynamic balancing – the ability to maintain balance while switching positions or otherwise being in motion – and Cassie has learned to make infinite subtle adjustments to stay upright while moving.
“The Dynamic Robotics Laboratory students in the OSU College of Engineering combined expertise from biomechanics and existing robot control approaches with new machine learning tools,” said Jonathan Hurst, who co-founded Agility Robotics in 2017 and is a robotics professor at OSU. “This type of holistic approach will enable animal-like levels of performance. It’s incredibly exciting.”
Hurst has often said walking robots will one day be a common sight – much like the automobile, and with a similar impact. The limiting factor has been the science and understanding of legged locomotion, but OSU researchers continue to make breakthroughs.
“Cassie is a very efficient robot because of how it has been designed and built, and we were really able to reach the limits of the hardware and show what it can do,” said Jeremy Dao, a Ph.D. student in the Dynamic Robotics Laboratory.
“Deep reinforcement learning is a powerful method in AI that opens up skills like running, skipping and walking up and down stairs,” added Yesh Godse, an undergraduate in the lab.
“In the not very distant future, everyone will see and interact with robots in many places in their everyday lives, robots that work alongside us and improve our quality of life,” Hurst said.
Back in June 2020, Hurst was the first guest on The Robot Report Podcast. Agility launched Cassie in mid-2017 and is now on the third and commercialized version of its Digit humanoid robot. We talked to Hurst about how passive dynamics steered the direction of Agility’s robots, the role of simulation and reinforcement learning, technical challenges and business opportunities of legged robots, comparisons to Boston Dynamics, his favorite robots, and much more. You can listen to the interview below.