InvariantCloud: Globally Invariant Point Cloud Registration for High-Precision 6DoF Tactile Pose Tracking

Pengfei YE, Yuxiang Ma, Yi Zhou, Molong Duan

Keywords: 6-DoF Pose Estimation, Globally Invariant Point Cloud Registration

TL;DR: InvariantCloud is a 6-DoF tactile object pose estimation framework using globally invariant surface marker distributions, enabling one-shot point cloud registration with sub-2° yaw error and robust, drift-free performance for precise manipulation.

Abstract: Recent advances in imitation learning and vision–language models highlight the need for high-fidelity tactile perception, with $6$-DoF tactile object pose estimation providing a crucial foundation for precise robotic manipulation. We introduce InvariantCloud, a $6$-DoF pose estimation framework that leverages the global invariance of surface marker arrangements on vision-based tactile sensors. In contrast to recent approaches, our one-shot globally invariant point cloud registration suppresses cumulative drift and overcomes long-standing limitations in accurately estimating yaw ($z$-axis) rotation. Empirical evaluations show that InvariantCloud achieves sub-$2^\circ$ yaw tracking error and sub-$1.5^\circ$ yaw re-localization repeatability, demonstrating its superior precision and robustness.

Submission Number: 23