CordViP: Correspondence-based Visuomotor Policy for Dexterous Manipulation in Real-World

TL;DR

CordViP introduces a novel framework that leverages 6D object pose estimation and correspondence learning to enhance dexterous robotic manipulation, overcoming limitations of traditional 3D point cloud methods.

Contribution

The paper presents CordViP, a new approach that constructs interaction-aware point clouds with correspondences for improved dexterous manipulation in robots.

Findings

Achieves state-of-the-art performance in six real-world tasks.

Demonstrates superior generalization across objects and viewpoints.

Shows robustness to different scenarios and occlusions.

Abstract

Achieving human-level dexterity in robots is a key objective in the field of robotic manipulation. Recent advancements in 3D-based imitation learning have shown promising results, providing an effective pathway to achieve this goal. However, obtaining high-quality 3D representations presents two key problems: (1) the quality of point clouds captured by a single-view camera is significantly affected by factors such as camera resolution, positioning, and occlusions caused by the dexterous hand; (2) the global point clouds lack crucial contact information and spatial correspondences, which are necessary for fine-grained dexterous manipulation tasks. To eliminate these limitations, we propose CordViP, a novel framework that constructs and learns correspondences by leveraging the robust 6D pose estimation of objects and robot proprioception. Specifically, we first introduce the interaction-aware point clouds, which establish correspondences between the object and the hand. These point clouds are then used for our pre-training policy, where we also incorporate object-centric contact maps and hand-arm coordination information, effectively capturing both spatial and temporal dynamics. Our method demonstrates exceptional dexterous manipulation capabilities, achieving state-of-the-art performance in six real-world tasks, surpassing other baselines by a large margin. Experimental results also highlight the superior generalization and robustness of CordViP to different objects, viewpoints, and scenarios. Code and videos are available on https://aureleopku.github.io/CordViP.