Track2Act: Predicting Point Tracks from Internet Videos enables Generalizable Robot Manipulation

TL;DR

Track2Act leverages web videos to predict point tracks and infer manipulation plans, enabling zero-shot generalizable robot manipulation across unseen objects and scenes with minimal robot-specific data.

Contribution

The paper introduces a novel framework that predicts point tracks from web videos to generate manipulation plans, reducing reliance on large demonstration datasets.

Findings

Enables zero-shot manipulation of unseen objects and scenes.

Combines web video-based predictions with minimal robot demonstrations.

Achieves diverse real-world manipulation tasks with minimal in-domain data.

Abstract

We seek to learn a generalizable goal-conditioned policy that enables zero-shot robot manipulation: interacting with unseen objects in novel scenes without test-time adaptation. While typical approaches rely on a large amount of demonstration data for such generalization, we propose an approach that leverages web videos to predict plausible interaction plans and learns a task-agnostic transformation to obtain robot actions in the real world. Our framework,Track2Act predicts tracks of how points in an image should move in future time-steps based on a goal, and can be trained with diverse videos on the web including those of humans and robots manipulating everyday objects. We use these 2D track predictions to infer a sequence of rigid transforms of the object to be manipulated, and obtain robot end-effector poses that can be executed in an open-loop manner. We then refine this open-loop plan by predicting residual actions through a closed loop policy trained with a few embodiment-specific demonstrations. We show that this approach of combining scalably learned track prediction with a residual policy requiring minimal in-domain robot-specific data enables diverse generalizable robot manipulation, and present a wide array of real-world robot manipulation results across unseen tasks, objects, and scenes. https://homangab.github.io/track2act/