Tactile-JEPA: topology-aware SSL cuts robot force error 6.3%, orientation error 20.8%
Elizaveta Kovtun · hf · 2026-09-28
Elizaveta Kovtun et al. present Tactile-JEPA, a self-supervised pre-training method for distributed tactile electronic skins, whose sparse, irregularly arranged sensing elements make visual SSL methods a poor fit.
Method: a sensor connectivity graph guides spatial masking to predict embeddings of masked elements, learning topology-aware representations; dual-scale masking captures both local contact detail and global skin state.
Results: across three datasets spanning magnetic and piezoresistive sensors, different robot embodiments, and single/paired-sensor setups, Tactile-JEPA cuts force estimation error by 6.3% and in-hand orientation error by 20.8% versus the prior SOTA, with consistent gains in policy learning. The authors argue tactile sensing benefits depend critically on encoder pre-training quality. Code is open-sourced.
More from Embodied
- AUAR ships robotic microfactories in containers to build homes up to six stories — lukas_m_ziegler · 2026-09-28
- GPT-6 Astra demo turns real-room video into interactive 3D worlds for robot training — 141_1337 · 2026-09-28
- Train robots for human kitchens, or redesign kitchens for robots? — vaibhavbetter · 2026-09-28
- PACTS at IROS 2026: Joint Action-Predicate Modeling Enables Zero-Shot Robot Skill Composition — joemeno · 2026-09-28
- Robotics KOL: humanoid isn't the form factor that scales, robots need personality — vaibhavbetter · 2026-09-28
- XGBoost screens 38,052 combos to find titanium alloy with 1.73 GPa strength — bravo_abad · 2026-09-28