How to make Robot move like a Human with Reinforcement Learning - Couverture souple

Ishuov, Timur

 
9798864120811: How to make Robot move like a Human with Reinforcement Learning

Synopsis

Have you ever imagined a robot learning non-stop? What about a humanoid robot? It learns each step and between trials, and its learning does not stall. Our world is made of waves. We turned a fully connected neural network to work in Fourier domain instead of polynomial Taylor Series. It is a world explorer by its architecture.

It was an advertisement part. What about results? It solves Bipedal Walker-v3 within 70 episodes, Bipedal Walker Hardcore v3 version within 1000 episodes, within 1200 episodes it produces human like moves in Humanoid-v4.

We call it Symphony, music of gestures.

Les informations fournies dans la section « Synopsis » peuvent faire référence à une autre édition de ce titre.