The biped robot designed by Oregon State University (OSU) made history after completing 5 kilometers in just 53 minutes. The robot Cassie was developed under the supervision of robotics professor Jonathan Hurst and produced by Agility Robotics, which focuses on leg movements. The robot was first launched in 2017 in collaboration with artificial intelligence professor Alan Fern. Since then, students at Oregon State University have been exploring the machine learning of the robot.
It uses machine learning to control the running gait and completes a 5-kilometer journey on the campus of Oregon State University after a battery charge. Professor Hurst said that the students combined biomechanics expertise and existing robot control methods with new machine learning tools to shape Cassie. “This holistic approach will bring performance to animal-like levels. This is very exciting,” Professor Hirst said in a statement.
The robot’s knees bend like an ostrich, and it teaches itself to run through a so-called “deep reinforcement learning” algorithm. The team stated that dynamic balance is a key part of running, and Cassie has made infinite fine adjustments to stay upright when moving.
Jeremy Dao, a PhD student in the Dynamic Robotics Laboratory at Oregon State University, said the robot is efficient because of the way it is designed and manufactured, adding that the team can reach the limits of the hardware and demonstrate its capabilities.
Regarding “deep reinforcement learning”, Yesh Godse, another student in the lab, said that this is a powerful method in artificial intelligence that can unlock skills such as running, jumping, and going up and down.
Professor Hirst said that walking robots will become a common sight, similar to cars, and even have similar effects. Professor Hirst added that the limiting factor is the science and understanding of leg movement. However, he said that Oregon State University’s research has achieved multiple breakthroughs.
In Cassie’s 5km run, the robot completed in 53 minutes and 3 seconds, and it took 6-1/2 minutes to reset it after it fell twice. The team stated that it fell down because the computer overheated for the first time, and then after being asked to perform a turn at too high a speed.