Emergence of locomotion behaviours in rich environments
Jul 14, 2017 by CGPress Staff
2
|
Researchers at Cornell University Library have made available their research into AI trained locomotion. By reinforcing learning using simple rewards and introducing a procedurally generated and variable terrain, they were able to train AI agents to develop their own complex locomotion techniques. You can read the paper on Cornell University Library’s website and see a visual depiction of the learned behaviour on YouTube.
made me laugh tears 🙂
This it an old technology:
https://www.youtube.com/watch?v=wippooDL6WE