R3M: A Universal Visual Representation for Robot Manipulation
This addresses the challenge of sample efficiency in robot manipulation for robotics researchers and practitioners, offering a significant but incremental advance over existing visual representations.
The paper tackles the problem of data-efficient learning for robotic manipulation by pre-training a visual representation (R3M) on human video data, resulting in over 20% improvement in task success compared to training from scratch and over 10% compared to state-of-the-art methods across 12 simulated tasks, and enabling real-world learning with just 20 demonstrations.
We study how visual representations pre-trained on diverse human video data can enable data-efficient learning of downstream robotic manipulation tasks. Concretely, we pre-train a visual representation using the Ego4D human video dataset using a combination of time-contrastive learning, video-language alignment, and an L1 penalty to encourage sparse and compact representations. The resulting representation, R3M, can be used as a frozen perception module for downstream policy learning. Across a suite of 12 simulated robot manipulation tasks, we find that R3M improves task success by over 20% compared to training from scratch and by over 10% compared to state-of-the-art visual representations like CLIP and MoCo. Furthermore, R3M enables a Franka Emika Panda arm to learn a range of manipulation tasks in a real, cluttered apartment given just 20 demonstrations. Code and pre-trained models are available at https://tinyurl.com/robotr3m.