Learning Real-World Robot Policies by Dreaming

TL;DR

This paper introduces a dreaming model that learns a realistic world representation for robots, enabling policy learning in simulation that successfully transfers to real-world robot control from images.

Contribution

It presents a novel action-conditioned dreaming model that captures scene dynamics, allowing efficient policy learning without extensive real-world interaction.

Findings

Dreaming model accurately emulates real environment sequences

Policies learned in dreaming transfer effectively to real robots

Reduces real-world data requirements for robot learning

Abstract

Learning to control robots directly based on images is a primary challenge in robotics. However, many existing reinforcement learning approaches require iteratively obtaining millions of robot samples to learn a policy, which can take significant time. In this paper, we focus on learning a realistic world model capturing the dynamics of scene changes conditioned on robot actions. Our dreaming model can emulate samples equivalent to a sequence of images from the actual environment, technically by learning an action-conditioned future representation/scene regressor. This allows the agent to learn action policies (i.e., visuomotor policies) by interacting with the dreaming model rather than the real-world. We experimentally confirm that our dreaming model enables robot learning of policies that transfer to the real-world.