Stop Thinking, Just Do!

Sung-Soo Kim's Blog

Deep Reinforcement Terrain Learning

tagsTags

21 May 2016


Deep Reinforcement Terrain Learning

In this piece of work, a combination of deep learning and reinforcement learning is presented which has proven to be useful in solving many extremely difficult tasks. Google DeepMind built a system that can play Atari games at a superhuman level using this technique that is also referred to as Deep Q-Learning. This time, it was used to teach digital creatures to walk and overcome challenging terrain arrangements.


The paper “Terrain-Adaptive Locomotion Skills Using Deep Reinforcement Learning “ is available here:
http://www.cs.ubc.ca/~van/papers/2016…

The implementation of the paper is also available here:
https://github.com/xbpeng/DeepTerrainRL

OpenAI’s Gym project:
https://gym.openai.com/


comments powered by Disqus