Stable bipedal movement can be quite challenging to pull off, especially for those who have a point-foot design. Researchers have now come up with a framework that enables bipedal point-footed, arm-less robots like Bolt to move with agility. It uses constrained reinforcement learning to address any instability, lack of arms, and limited foot actuation.
With this approach, Bolt can jump on its feet and maintain its balance and even survive a rug pull.
[HT]