Asymmetric Actor Critic for Image-Based Robot Learning

Deep reinforcement learning (RL) has proven a powerful technique in many sequential decision making domains. However, robotics poses many challenges for RL, most notably training on a physical system can be expensive and dangerous, which has sparked significant interest in learning control policies using a physics simulator. While several recent works have shown promising results in transferring policies trained in simulation to the real world, they often do not fully utilize the advantage of working with a simulator. In this work, we propose the Asymmetric Actor Critic, which learns a vision-based control policy while taking advantage of access to the underlying state to significantly speed up training. Concretely, our algorithm employs an actor-critic training algorithm in which the critic is trained on full states while the actor (or policy) is trained on images. We show that using these asymmetric inputs improves performance on a range of simulated tasks. Finally, we combine this method with domain randomization and show real robot experiments for several tasks like picking, pushing, and moving a block. We achieve this simulation to real-world transfer without training on any real-world data. Videos of these experiments can be found in www.goo.gl/b57WTs.