From Pixels to Torques: Policy Learning with Deep Dynamical Models

Data-efficient learning in continuous state-action spaces using very high-dimensional observations remains a key challenge in developing fully autonomous systems. In this paper, we consider one instance of this challenge, the pixels-totorques problem, where an agent must learn a closed-loop control policy from pixel information only. We introduce a data-efficient, model-based reinforcement learning algorithm that learns such a closed-loop policy directly from pixel information. The key ingredient is a deep dynamical model that uses deep autoencoders to learn a low-dimensional embedding of images jointly with a prediction model in this low-dimensional feature space. This joint learning ensures that not only static properties of the data are accounted for, but also dynamic properties. This is crucial for long-term predictions, which lie at the core of the adaptive model predictive control strategy that we use for closedloop control. Compared to state-of-the-art reinforcement learning methods, our approach learns quickly, scales to high-dimensional state spaces and facilitates fully autonomous learning from pixels to torques.