Decoupling feature extraction from policy learning: assessing benefits of state representation learning in goal based robotics
This work addresses sample efficiency problems for robotics researchers using reinforcement learning, though it appears incremental as it builds on existing state representation learning approaches.
The paper tackles the challenge of sample inefficiency in vision-based robot control by evaluating state representation learning methods, proposing a new unsupervised stacked model that encodes relevant features and achieves comparable or better performance than end-to-end learning with improved sample efficiency.
Scaling end-to-end reinforcement learning to control real robots from vision presents a series of challenges, in particular in terms of sample efficiency. Against end-to-end learning, state representation learning can help learn a compact, efficient and relevant representation of states that speeds up policy learning, reducing the number of samples needed, and that is easier to interpret. We evaluate several state representation learning methods on goal based robotics tasks and propose a new unsupervised model that stacks representations and combines strengths of several of these approaches. This method encodes all the relevant features, performs on par or better than end-to-end learning with better sample efficiency, and is robust to hyper-parameters change.