PanoTac: A Dense Three-Axis Tactile Sensor for Robust Touch-Centric Manipulation

Binglin Wang, Xiaole Jiang, Zicheng Teng, Xinqi Liu, Yunzhu Li, Lerrel Pinto, Zichen Jeff Cui, Binghao Huang

November 2026

Four autonomous rollouts with the live three-axis tactile field.

TL;DR A sub-$50 gripper fingertip that fuses a dense piezoresistive array with magnets and magnetometers into a 12x32 map of normal and shear force. With it, visuo-tactile diffusion policies reach 91.7% success on test-tube prep, grain scooping and salt pouring, against 45% for vision alone.

What they do

Piezoresistive skins give dense contact maps but no shear. Magnetic skins give shear but only at a few points. PanoTac stacks the two in one compliant fingertip for a parallel-jaw gripper, producing a dense three-axis tactile image at 60 Hz.

How it works

Hardware. A 12x32 piezoresistive array (2 mm pitch) sits on TPU tiles over two groups of neodymium magnets, with ten magnetometers at the base. Each group is a 2x2 checkerboard of alternating polarity, whose far field decays as r⁻⁵ instead of r⁻³, which cuts cross-talk between the two fingertips. Costs under $50, about 3 h of printing and 30 min of assembly.
Representation. The normal channel is the contact map scaled by the magnetic z response. The two shear channels are magnetometer x/y readings interpolated onto the same grid, giving 12x32x3 per finger.
Learning. Demonstrations come from VR teleoperation with tactile heat-map and audio feedback. A U-Net diffusion policy takes a wrist camera image, proprioception and both tactile fields through a small CNN.

Results

Franka with a Robotiq 2F-85 on three tasks: test-tube preparation (find a tube’s orientation in hand and detect stopper release), grain scooping (infer spoon pose and resistance) and salt pouring (tell full from empty by weight).
Overall success 91.7% vs 45.0% vision-only. Removing either the spatial contact map or the shear channels lowers success.

Limitations

The magnetometers are sensitive to ferromagnetic objects and external fields. Force resolution is limited by elastomer compliance and magnetometer count. The fingertip has not yet been tried on a dexterous hand.

Videos

Test-tube preparation, autonomous, 2x speed.

Sensor response to pressing and shearing.

Figures

Fig. 1 : We propose PanoTac , a low-cost, compliant tactile fingertip for dense three-axis tactile sensing. We showca... — Fig. 1 : We propose PanoTac , a low-cost, compliant tactile fingertip for dense three-axis tactile sensing. We showcase PanoTac with three tasks (i) test-tube preparation, where in-hand state helps infer tube orientation and shear feedback helps detect stopper release; (ii) grain scooping, where contact patterns help infer spoon orientation and guide robot–environment interaction; and (iii) salt pouring, where weight cues distinguish filled from empty bottles and inform pouring decisions.

Fig. 2 : PanoTac Tactile Sensing Platform. Part (a) shows the assembled sensor and an exploded view of the multilayer... — Fig. 2 : PanoTac Tactile Sensing Platform. Part (a) shows the assembled sensor and an exploded view of the multilayer design. Part (b) shows the indentation setup for sensor characterization. Part (c) shows the consistency of (i) the normal response z and (ii) the RMS shear (x,y) magnitude across three sensor under normal loads of 0.4 – 12 N. Shaded bands indicate \pm 1 standard deviation. Part (d) illustrates how we construct the 12\times 32\times 3 dense three-axis tactile representation.

Fig. 4 : Qualitative Results. Representative stages of (a) test-tube preparation, (b) grain scooping, and (c) salt po... — Fig. 4 : Qualitative Results. Representative stages of (a) test-tube preparation, (b) grain scooping, and (c) salt pouring. Paired tactile visualizations show responses from both fingertips, and red arrows indicate motion directions. Blue bars mark stages shown for both task branches. Detailed task descriptions are provided in Sec. V .

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.

Summaries are written automatically from each paper's text. Figures and videos come from the paper and its project page; all credit goes to the authors.

Matt Strong