Generalization in Dexterous Manipulation via Geometry-Aware Multi-Task Learning
This addresses a fundamental challenge in robotics for enabling autonomous systems to handle diverse objects, though it is incremental as it builds on existing reinforcement learning methods with a new representation.
The paper tackles the problem of poor generalization in dexterous robotic manipulation by showing that a single generalist policy, using multi-task learning with object point cloud representations, can manipulate over 100 diverse real-world objects and generalize to unseen shapes or sizes, even outperforming specialist policies on training and test objects.
Dexterous manipulation of arbitrary objects, a fundamental daily task for humans, has been a grand challenge for autonomous robotic systems. Although data-driven approaches using reinforcement learning can develop specialist policies that discover behaviors to control a single object, they often exhibit poor generalization to unseen ones. In this work, we show that policies learned by existing reinforcement learning algorithms can in fact be generalist when combined with multi-task learning and a well-chosen object representation. We show that a single generalist policy can perform in-hand manipulation of over 100 geometrically-diverse real-world objects and generalize to new objects with unseen shape or size. Interestingly, we find that multi-task learning with object point cloud representations not only generalizes better but even outperforms the single-object specialist policies on both training as well as held-out test objects. Video results at https://huangwl18.github.io/geometry-dex