GenTact Toolbox
Robot Skin for All Robots
A computational pipeline for generating custom, context-driven whole-body tactile skins that conform to any robot geometry.
Abstract
Robots lack skin — and existing sensor attachment methods are tedious, robot-specific, and hard to reproduce. We propose multi-material additive manufacturing (MMAM) as the fabrication foundation for robot skins: a single print produces a complete skin unit with embedded sensors, conductive traces, and compliant structures directly from a digital design file. We demonstrate six sensing modalities through MMAM and evaluate them across three axes: repeatability (consistent placement and signal quality across prints), scalability (from single patches to full-body humanoid coverage), and compatibility (multiple modalities in one skin on a shared microcontroller).
Authors
Caleb Escobedo*†,
Carson Kohlbrenner†,
Anna Soukhovei,
Klara Nitsche,
Artemis Shaw,
Nikolaus Correll,
Alessandro Roncone
Department of Computer Science, University of Colorado Boulder · *Corresponding author · †Equal contribution
Six Sensing Modalities
A single MMAM print can embed fundamentally different sensor physics within the same skin unit, driven by a shared microcontroller. The paper characterises six modalities:
Three Evaluation Axes
Consistent Across Prints
Self-capacitive skins achieved 139 ± 19 ADC peak response with 8.6% inter-prototype CV across three identical MMAM prints.
Any Robot, Any Size
Full-body skins fabricated for a robot arm, quadruped, and humanoid — from small sensor patches to complete body coverage.
Multiple Modalities, One Skin
Different sensing physics co-exist in a single skin unit on a shared microcontroller — calibration handled in the digital design.
The Fabrication Pipeline
- Procedural Mesh Generation — Sensor layouts are generated using the open-source GenTact Blender addon (Geometry Nodes), placing sensors, routing internal wiring, and specifying multi-material zones before any physical manufacturing occurs.
- Simulation & Task-Driven Optimisation — 1:1 import into Isaac Sim for contact simulation and sensor placement refinement against specific manipulation or HRI tasks.
- Multi-Material Fabrication — A Prusa XL printer with five independent nozzles extrudes rigid PLA (dermis), soft TPU (epidermis), conductive PLA (electrodes/traces), and support material in a single uninterrupted print.
Future Applications
The GenTact pipeline generalises wherever compliant, geometry-aware sensing is the bottleneck. Four directions with direct overlap to ongoing research: