Learning Navigation Subroutines from Egocentric Videos
This work addresses the challenge of efficient planning and learning in robotics and AI by enabling the acquisition of reusable subroutines from passive video data, which is incremental as it builds on hierarchical and self-supervised methods.
The paper tackles the problem of improving sample and computational efficiency in navigation by learning hierarchical abstractions or subroutines from egocentric videos of experts, resulting in successful learning of consistent and diverse visuomotor subroutines that can be used for exploration and in hierarchical RL frameworks for reaching goals, with real-world deployment on a robotic platform.
Planning at a higher level of abstraction instead of low level torques improves the sample efficiency in reinforcement learning, and computational efficiency in classical planning. We propose a method to learn such hierarchical abstractions, or subroutines from egocentric video data of experts performing tasks. We learn a self-supervised inverse model on small amounts of random interaction data to pseudo-label the expert egocentric videos with agent actions. Visuomotor subroutines are acquired from these pseudo-labeled videos by learning a latent intent-conditioned policy that predicts the inferred pseudo-actions from the corresponding image observations. We demonstrate our proposed approach in context of navigation, and show that we can successfully learn consistent and diverse visuomotor subroutines from passive egocentric videos. We demonstrate the utility of our acquired visuomotor subroutines by using them as is for exploration, and as sub-policies in a hierarchical RL framework for reaching point goals and semantic goals. We also demonstrate behavior of our subroutines in the real world, by deploying them on a real robotic platform. Project website: https://ashishkumar1993.github.io/subroutines/.