The Swiss robot dog ANYmal has no problem walking on rocky terrain, but now researchers in Zurich have taught the four-legged robot to also parkour. It is a sport based on the use of complex maneuvers to smoothly overcome obstacles in an urban environment.

ETH doctoral student Nikita Rudin, who works in one of the teams, also practices parkour in his spare time. “Before the start of the project, several of my research colleagues thought that legged robots had already reached the limits of their development potential, but I had a different opinion. In fact, I was sure that much more could be done with the mechanics of legged robots.”

With his parkour experience in mind, Rudin set out to further push the boundaries of ANYmal. He taught the four-legged robot new skills using machine learning. ANYmal can now scale obstacles and perform dynamic maneuvers to jump off them.

In the process, ANYmal learned like a child – by trial and error. Now, when presented with an obstacle, ANYmal uses its camera and artificial neural network to determine which obstacle is in front of it. He then makes moves that seem more likely to succeed based on previous training.

ANYmal will soon also be used on construction sites or other places that are too dangerous for people. For example, to inspect a collapsed house in a disaster area.

Source: Science Daily