PanoTac fingertip sensor improves robot touch for tricky tasks
PanoTac: A Dense Three-Axis Tactile Sensor for Robust Touch-Centric Manipulation
RoboticsArtificial IntelligenceMachine Learning
Summary
Robots need to feel what they touch to handle delicate or messy tasks well. The authors built PanoTac, a cheap fingertip sensor that senses pressure and sliding forces in three directions with many tiny sensing points. This sensor helps robots do better at tasks like scooping grains or pouring salt by giving them more detailed touch information than vision alone. The authors showed that without either touch location details or sideways force sensing, robots perform less well.
What this means in practice
- •For robotics engineers: Equip robots with detailed touch sensing to improve handling of delicate and granular objects in diverse manipulation tasks.
- •For industrial automation teams: Implement enhanced tactile sensing in robot grippers to increase reliability in complex assembly or packaging that involves varied contact forces.
Authors
Binglin, Wang, Xiaole, Jiang, Zicheng Teng, Xinqi Liu, Yunzhu Li, Lerrel Pinto, Zichen Jeff Cui, Binghao Huang
Abstract
Contact-rich manipulation benefits from tactile feedback that captures both detailed contact geometry and interaction forces. When scooping grains, contact patterns help infer spoon orientation, while resistance transmitted through the grasp guides motion. To provide these complementary sensing capabilities, we present PanoTac, a low-cost, compliant tactile fingertip designed for dense three-axis tactile sensing. Each fingertip provides a $12 \times 32 \times 3$ tactile representation, with one normal ($z$) and two shear ($x$, $y$) response components at each taxel. The sensor operates at 60 Hz and costs less than $50 per fingertip. Sensor characterization quantifies response repeatability and consistency across sensor instances and contact locations. We integrate PanoTac into a robotic system with a visuo-tactile pipeline. Across test-tube preparation, grain scooping, and salt pouring, the resulting policies outperform vision-only baselines by an average of 46.7 percentage points. Ablations show that removing either spatial contact information or shear feedback reduces overall performance across the task suite. Our project page is available at https://panotac.github.io.