Oregon State University stated for the first time that its biped robot Cassie successfully changed from a sofa to a 5K in 53 minutes. This is an impressive feat of machine learning and robotics, but when it comes to 5K, I won’t worry about leaving killer humanoid robots behind for the time being.
Cassie has been around for a while and is the idea of Agility Robotics (a spin-off company of OSU). The important thing here is not that Cassie completes a speed of 5 kilometers. For example, an average healthy person can walk 5 kilometers in a similar or less time, while most beginners complete that distance in 30-40 minutes. (Even in OSU’s video, you can see people staying in sync with Cassie while walking.) What’s impressive is that Cassie can “run” such a long distance with a single charge.
From a biomechanical point of view, running is very complicated. There are many websites dedicated to analyzing your running gait, but most people just run instinctively. It is difficult to reproduce this ability in robots, especially biped robots because it requires something called “dynamic balance”, or the ability to change position without tipping over while moving. The team said that thanks to the “deep reinforcement learning algorithm,” Cassie was able to teach herself how to make subtle adjustments 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, an OSU robotics professor and co-founder of Agility Robotics. “This type of holistic approach will enable animal-like levels of performance. It’s incredibly exciting.”
However, compared with quadruped robots, balance is a special challenge for biped robots. For example, Boston Dynamics’ Cheetah robot can run at 28 mph, while Unitree’s robotic dog Go1 can run side by side with someone at a moderate speed of 5.6-6 mph. These four-legged robots have been able to run at high speeds for some time, although they may not be able to run long distances by themselves. At the same time, Cassie’s 53-minute running time includes approximately 6.5 minutes, in which the robot’s computer overheated and another instance of the robot falling over too fast while turning. However, the main argument in favor of biped robots is that, in theory, they can be more easily integrated into daily life.
OSU predicts that biped robots will one day handle logistics tasks, such as delivering packages, but will also help people at home. “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,” Hirst said. To this end, Cassie is also able to use machine learning to learn how to go up and down stairs without lidar or cameras.
Although this is a future goal, you won’t see biped robots anytime soon. At least not outside school. Before these robots can truly become consumer devices, they must also overcome many mechanical and engineering challenges. But at the same time, can anyone do something about the nightmarish appearance of these highly advanced robots? No one would invite these things into their home like that.
