Deep reinforcement learning for robotic manipulation with asynchronous off-policy updates

TLDR

Reinforcement learning promises autonomous robots to acquire diverse skills with minimal human input, yet practical robotic applications often rely on hand‑engineered policies and demonstrations to reduce training time, and deep RL has been largely confined to simulation or simple tasks because of its high sample complexity. The study shows that an off‑policy deep Q‑function algorithm can scale to complex 3D manipulation tasks and train deep neural network policies efficiently on real robots. The algorithm trains deep Q‑functions off‑policy and further shortens training time by asynchronously pooling policy updates across multiple robots. Experiments demonstrate that the method learns diverse 3D manipulation skills in simulation and a complex door‑opening task on real robots without prior demonstrations or hand‑crafted representations.

Abstract

Reinforcement learning holds the promise of enabling autonomous robots to learn large repertoires of behavioral skills with minimal human intervention. However, robotic applications of reinforcement learning often compromise the autonomy of the learning process in favor of achieving training times that are practical for real physical systems. This typically involves introducing hand-engineered policy representations and human-supplied demonstrations. Deep reinforcement learning alleviates this limitation by training general-purpose neural network policies, but applications of direct deep reinforcement learning algorithms have so far been restricted to simulated settings and relatively simple tasks, due to their apparent high sample complexity. In this paper, we demonstrate that a recent deep reinforcement learning algorithm based on off-policy training of deep Q-functions can scale to complex 3D manipulation tasks and can learn deep neural network policies efficiently enough to train on real physical robots. We demonstrate that the training times can be further reduced by parallelizing the algorithm across multiple robots which pool their policy updates asynchronously. Our experimental evaluation shows that our method can learn a variety of 3D manipulation skills in simulation and a complex door opening skill on real robots without any prior demonstrations or manually designed representations.