Cassie, the bipedal robot with all legs, successfully ran five kilometers without a leash and on one charge. The machine serves as the basis for the delivery robot from Agility Robotics Digit, as. TechCrunch notes, although you can also remember this for “blind” walking up the stairs. Oregon State University engineers were able to train Cassie on a simulator to enable her to climb and descend stairs without the use of cameras or lidars. Now engineers from the same team were able to train Cassie to run using a deep reinforcement learning algorithm.
According to the team, Cassie, by teaching herself the technique, enabled her to stay upright without being tethered, shifting her balance while running. The robot had to learn how to make endless fine-tuning in order to accomplish the feat. Yesh Godse, a student at OSU’s Dynamic Robotics Lab, explained: “Deep Reinforcement Learning is a powerful technique in artificial intelligence that unlocks skills such as running, jumping and walking up and down stairs.”
The team first tested Cassie’s capabilities by making her run five kilometers around the corner, which was 43 minutes 49 seconds. Cassie finished her run across the OSU campus in 53 minutes 3 seconds – it took a little longer because it took six and a half minutes to solve technical problems. The robot fell once due to overheating of the computer, and then again after turning too quickly. Jeremy Dao, another member of the lab team, however, said they were able to “push the limits of the equipment and show what it was capable of.” The work the team is doing will help expand understanding of leg movement and could help make bipedal robots more prevalent in the future.
All Engadget recommended products are selected by our editorial team, independent of our parent company. Some of our stories contain affiliate links. If you buy something from one of these links, we may receive an affiliate commission.