Cassie is the base of Agility Robotics' delivery robot, Digit. You've probably been following the tech's progress closely for some time and know that Digit started as Cassie, an ostrich-inspired Cassie. The robot is purely a robot with legs.The robot is proving to be a strong platform for bipedal exploration, and has helped Agility achieve its commercial goals. Cassie was announced by Jonathan Hunt, an Oregon State University professor in 2017, with the goal of receiving a $1million grant from the DoD. This is a familiar story in robotics.Today, the Dynamics Robotics Laboratory at OSUs College of Engineering showed Cassie how they are pushing Cassie to the limits. According to the team, the robot could run 5K without tethering on one charge. Although Cassies is unlikely to break any human world records anytime soon the robot's 53-minute and three-second runs were still impressive.The robot's time was approximately 6.5 minutes. They had to troubleshoot a computer that was overheating and a turn that went wrong, which caused it to fall off its feet.Cassie is a highly efficient robot due to how it was designed and built. We were able really to reach the hardware limits and show what it could do, Ph.D. student Jeremy Dao stated in a statement.The team claims that Cassie learned to run by using deep reinforcement learning algorithms. This allowed Cassie to learn how to balance and stay upright while running.Yesh Godse, a graduate student, adds that deep reinforcement learning is an important method in AI. It opens up abilities like running, skipping, and walking up and down steps.Cassie was also able to walk up and down stairs with no assistance from lidar or onboard cameras.
