Google DeepMind teaches AI agent to navigate Parkour track

Posted on Thursday, July 13 2017 @ 13:12 CEST by Thomas De Maesschalck
Google's DeepMind unit shows off the result from another of its AI research projects. This time the researchers trained an AI agent to navigate a parkour course by using reinforcement learning to teach itself to navigate as far and as fast as possible. Full details at ExtremeTech.
DeepMind looked at non-human walkers as well. The “ant” walker above was able to learn how to leap across chasms in a way the human simulations never would. Again, it learned to do this via trial and error. Actually, there’s nothing that dictates the movement of human-like simulations must look human-like. Some of the emergent behaviors include amusing quirks, like the stick figure’s tendency to flail its arms about to keep its balance. Then there’s the way the simpler “planar” walking legs used its knee to lever itself over tall walls.

Loading Comments