Agility Robotics' Bipedal Cassie Ran a 5K Without the Need of Tethers
Using deep reinforcement learning to maintain balance and tackle terrain.
Cassie — a bipedal robot from Agility Robotics — has just earned itself a new achievement by running a full five kilometers without needing any sort of tether.
The new feat was achieved through something known as a deep reinforcement learning algorithm, which enables Cassie to teach itself how to maintain balance throughout the run by making small adjustments as it goes. OSU Dynamic Robotics Laboratory’s Yesh Godse explains that “Deep reinforcement learning is a powerful method in AI that opens up the skills like running, skipping and walking up and down stairs.”
Overall, the five-kilometer run took Cassie just 43 minutes and 49 seconds, although it did falter twice, the first time due to computer components overheating and the second time simply because it took a turn too sharply. Despite the two minor setbacks, Agility Robotics is confident that Cassie’s new achievement will help with the future development of even more sophisticated bipedal robots which could ultimately be used as couriers.
Elsewhere in tech, Tesla may be delaying production for the Cybertruck until 2022.
