Observed arrival · 2026-09-04
TactileLab Simulates the Feel of Robotic Touch
An IsaacLab-based framework for training dexterous robots with simulated contact forces, depth images, positions, and tactile-flow signals.
Field notes
The framework uses rigid-body simulation to estimate dense tangential motion at contact interfaces, aiming to capture shear-relevant cues without expensive soft-body physics. Its documented observation set ranges from low-dimensional forces and positions to depth maps and pixel-level or sparse tactile flow. The page also describes PETS-Net for combining tactile observations with proprioception, and reports comparisons averaged across five random seeds. Code is identified as forthcoming, while the page presents the project as an anonymous submission.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue