Any-point Trajectory Modeling for Policy Learning
This addresses the high cost of collecting demonstration data for robot learning, offering a method to leverage abundant video data for control tasks, though it is incremental in improving existing video pre-training approaches.
The paper tackles the challenge of learning robot policies from video demonstrations without action labels by introducing Any-point Trajectory Modeling (ATM), which pre-trains a model to predict future trajectories of arbitrary points in videos, enabling robust visuomotor policy learning with minimal labeled data and achieving an 80% average improvement over baselines across over 130 tasks.
Learning from demonstration is a powerful method for teaching robots new skills, and having more demonstration data often improves policy learning. However, the high cost of collecting demonstration data is a significant bottleneck. Videos, as a rich data source, contain knowledge of behaviors, physics, and semantics, but extracting control-specific information from them is challenging due to the lack of action labels. In this work, we introduce a novel framework, Any-point Trajectory Modeling (ATM), that utilizes video demonstrations by pre-training a trajectory model to predict future trajectories of arbitrary points within a video frame. Once trained, these trajectories provide detailed control guidance, enabling the learning of robust visuomotor policies with minimal action-labeled data. Across over 130 language-conditioned tasks we evaluated in both simulation and the real world, ATM outperforms strong video pre-training baselines by 80% on average. Furthermore, we show effective transfer learning of manipulation skills from human videos and videos from a different robot morphology. Visualizations and code are available at: \url{https://xingyu-lin.github.io/atm}.