InvariantCloud: A Globally Invariant, Uniquely Indexed Point Cloud Framework for Robust 6-DoF Tactile Pose Tracking
Pengfei Ye, Yuxiang Ma, Yi Zhou, Wei Chen, Wenzhen Dong, Molong Duan
AI summary
Problem
Current vision-tactile 6-DoF pose estimation methods struggle with cumulative drift, registration ambiguity, and unreliable yaw rotation, hindering precise robotic manipulation without external cameras or CAD models.
Approach
The framework builds a dense, globally invariant reference point cloud with unique IDs on the tactile sensor, enabling direct one-to-one ID matching for closed-form XY rotation and PCA-based silhouette analysis for robust Z-axis yaw estimation.
Key results
- Eliminates cumulative drift in long-sequence tactile tracking
- Achieves superior yaw (Z-axis) rotation accuracy and stability
- Reduces repeatability error across multiple household objects
- Enables robust inter-contact registration for long-horizon tactile SLAM
Why it matters
Provides a drift-free, high-precision tactile pose estimation framework essential for reliable robotic manipulation and tactile SLAM without external sensors or prior object models.
Abstract
Recent advances in imitation learning and vi- sion–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 frame- work that leverages the global invariance of surface marker constellations 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) ro- tation. Experimental verifications show that InvariantCloud achieves superior yaw tracking accuracy and re-localization repeatability compared to existing benchmarks, demonstrating its precision and robustness in long-sequence manipulation tasks.