Researchers at Google have developed a robotic dog and taught it to walk by showing it motion-capture videos of real dogs walking on treadmills. The robot, Laikago, has a body very different from a biological dog's body; a digital version of Laikago used computer simulations to determine how to imitate the motion of the digital version of the biological dog without directly copying its mechanics.
The researchers used a reinforcement learning algorithm to help the robot learn to move as similarly to the original reference motion as possible. The algorithm attempts random movements, and gets a digital "reward" if it gets closer to the dog's reference motion. Over many iterations, this reward systems teaches the simulated robot dog to move like a real dog.
Said Google’s Jason Peng, “The drawback with the kind of manual approach is that it's not really scalable for every skill that we want a robot to perform.”
View Full Article
Abstracts Copyright © 2020 SmithBucklin, Washington, DC, USA