Taxel
- Also called: tactile pixel, tactile element
- Category: Sensing › Contact & force
- Tags: tactile-sensing, sensor-arrays, touch, manipulation, robot-skin
A single sensing element in a tactile sensor array, the touch equivalent of a pixel; its size and spacing set the array's spatial resolution.
Definition
A taxel, short for tactile element or tactile pixel, is one sensing site in a tactile sensor array. Each taxel reports a local quantity at its location on the sensor surface: most often normal pressure, and in some designs a three-axis force, shear, vibration or proximity. Taxels can use any transduction principle, including capacitive, piezoresistive, piezoelectric, magnetic and optical sensing, and they are arranged in grids on fingertips and palms or spread more sparsely over robot skin.
The number, size and spacing of taxels, the pitch, determine what an array can resolve. Fine pitch makes it possible to locate edges and estimate contact shape, while coarse arrays mainly report where and how hard contact occurs. The elastomer layer that covers most arrays spreads each contact over several neighboring taxels, which blurs the image but also allows contact location to be interpolated more finely than the pitch. Taxels are rarely independent: in passive resistive matrices current leaks through neighboring elements and causes crosstalk, and each element has its own gain, offset, hysteresis and drift, so arrays need per-taxel calibration.
Reading many taxels quickly is an engineering problem of its own. Large arrays use row-column multiplexing, local analog-to-digital converters and serial buses to keep the wiring manageable, trading scan rate against element count. For learning, taxel readings are commonly treated as a low-resolution image when the layout is a regular grid, or, as in some recent work, as a set of points with known 3D positions on the hand, which lets one model handle sensors with different layouts. In camera-based tactile sensors there are no discrete taxels; the term is sometimes used loosely for pixels or tracked markers.
Why it matters
Taxel count, pitch and readout rate are the first numbers to check when choosing or designing a tactile sensor, and they determine what a manipulation algorithm can actually perceive. Treating taxels as calibrated, possibly coupled measurements rather than clean force readings avoids many integration problems.
Key applications
- Contact localization and slip detection on robot fingertips
- Large-area robot skin for collision detection and safe human contact
- Tactile gloves that record human grasp pressure
- Tactile input for learned dexterous manipulation policies
- Pressure mapping for prosthetic hands and medical devices
In practice
A team fits a robot hand with fingertip arrays of a few dozen capacitive taxels each. The microcontroller on each finger scans the array, subtracts a baseline captured with no contact and sends the frames over a serial bus. During calibration, each taxel's response is recorded while a probe presses known forces at known locations, giving a gain and offset per element. The grasp controller then computes the contact centroid by weighting taxel positions by their calibrated pressure, and watches for the centroid moving toward the fingertip edge as a sign of slip. For learning, the team stores each frame with the 3D position of every taxel in the hand frame, so data from fingers with different layouts can be combined in one model.
Common misconceptions
- A taxel is a whole tactile sensor — it is one element of an array; a sensor usually contains many
- Each taxel measures force directly — it measures a transduced quantity such as capacitance or resistance that must be calibrated, often nonlinearly, to estimate pressure or force
- More taxels always give better sensing — wiring, scan rate, crosstalk and the blurring effect of the cover layer limit how much a denser array helps
- Taxel readings are independent — neighboring elements are coupled mechanically through the skin and, in passive matrices, electrically through the readout
Who should know this
- Tactile sensor designers and electronics engineers
- Dexterous manipulation researchers
- Robot hand and gripper developers
- Engineers building robot skin and safety sensing
Related technologies
- Capacitive Tactile Arrays
- Row-Column Multiplexing
- Robot Skin
- Tactile Gloves
- Point Cloud Encoders
Related terms
- Tactile Sensor
- Piezoresistive Tactile Sensor
- Camera-Based Tactile Sensor
- Force-Sensitive Resistor
- Dexterous Manipulation
Related projects
- LeRobot: HuggingFace ML library for real-world robotics in PyTorch
- Isaac Lab: NVIDIA robot learning framework: one task API across Newton, PhysX and Isaac Sim
- NVIDIA Isaac GR00T: NVIDIA open foundation model for humanoid robot reasoning and skills
- OpenVLA: 7B-parameter Vision-Language-Action model for robotic manipulation
- Drake: MIT robotics toolbox for model-based dynamics, optimization, and control system design
- OpenArm: 7DOF humanoid arm for physical AI research in contact-rich environments