Touch is not simply pressure. It is a composite signal rich in multidimensional perceptual information. TacTip optoelectronic tactile sensors turn contact geometry, shear, slip, texture and subtle dynamic changes into information robots can use.
Overview
Inspired by biological touch, TacTip uses an internal optoelectronic sensor to observe deformation of embedded features inside a compliant skin. The result is a tactile stream that is visual at its core, yet deeply physical in what it reveals.
Ward-Cherrier et al., "The TacTip Family: Soft Optical Tactile Sensors with 3D-Printed Biomimetic Morphologies", Soft Robotics, 2018·Lepora, "Soft Biomimetic Optical Tactile Sensing with the TacTip: A Review", IEEE Sensors Journal, 2021
Get the highlights
Respond before the moment passes
World-leading 120 Hz optoelectronic tactile reporting captures rapid changes across contact. Faster feedback helps robots respond during grasping, adjustment and delicate manipulation, turning each interaction into a continuous stream of information for more responsive control.
Sense the slightest disturbance
Even the disturbance of a gentle breeze becomes a tactile cue. TacTip captures subtle vibrations and dynamic changes, helping robots notice light contact and changing conditions earlier, with a finer touch for soft objects, precision components and fragile targets.
More signals. More possibilities.
Multimodal output gives models richer ways to interpret contact, from contact shape to changing mechanics. Flexible data interfaces connect touch with perception, learning and control, allowing the same interaction to support different tasks and expand the possibilities for embodied intelligence.
Read slip before the grip is lost
As one of the few optoelectronic tactile sensors capable of sensing slip, TacTip turns relative motion at contact into actionable cues. Robots can recognize grasp conditions and adjust their strategies, maintaining better control through movement, rotation and coordinated finger actions.
Less calibration. More coordination.
Individual sensors need no separate calibration, addressing inconsistent outputs across multiple units. A shared perception and control approach can connect more fingertips with less repeated tuning, helping teams move from adapting each sensor to coordinating an integrated multi-finger system.
Built for repeated contact
Long-term laboratory use has demonstrated a lifetime of hundreds of thousands of contacts. Through repeated grasps, continuous experiments and frequent interactions, TacTip's durability keeps teams focused on their tasks and makes touch a lasting capability for research and robotic applications.
Locate the details that matter
A distinctive sensing principle combined with computation delivers micrometre-scale touch localization. Small shifts in contact become useful spatial information, helping robots locate contact during precision alignment, touch-guided movement and detailed exploration, with a clearer basis for finer manipulation.
Every contact adds understanding
Leading multi-point touch sensing captures changes across several contact locations at once. Models gain a more complete picture of complex interactions, helping robots understand how objects are supported, constrained and manipulated during multi-finger grasping and other demanding contact tasks.
Our product family
Optoelectronic tactile fingertips, configurable research kits and tactile dexterous hands connect contact sensing, experimentation and robot integration.
Applications
A decade of published research has taken TacTip from benchtop characterisation to closed-loop control on real robot systems.
Changes in tangential contact can indicate incipient slip, giving a gripper feedback for grasp adjustment.
Closed-loop tactile servoing traces unknown edges and surfaces — a foundation for inspection and assembly.
Dynamic pin signatures discriminate fine surface textures beyond what vision alone resolves.
Tactile feedback enables rolling, sliding and re-orienting objects within the hand — without looking.
Fingertip form factors bring high-resolution touch to multi-fingered hands and humanoid platforms.
Through gentle contact, optoelectronic tactile sensing captures small, rhythmic skin deformations for pulse-waveform research and wearable prototypes.
James & Lepora, "Slip Detection for Grasp Stabilization with a Multifingered Tactile Robot Hand", IEEE Transactions on Robotics, 2021·Lepora & Lloyd, IEEE RAM, 2021·Lin et al., "Bi-Touch", IEEE RA-L, 2023
Research story
TacTip provides a bridge between tactile mechanics and computational perception. Marker motion can be tracked, transformed into tactile descriptors and integrated into downstream models for contact interpretation, manipulation planning and adaptive control.
Capture dense local deformation patterns for richer contact interpretation than sparse binary touch signals.
The biomimetic morphology supports safer physical interaction and improved adaptation to object surfaces.
Designed for classical vision pipelines and modern learning workflows, from tracking-based inference to foundation-model-assisted robotics.
A compact fingertip form factor makes TacTip suitable for dexterous manipulation research and human–robot interaction studies.
Signal pathway
Soft contact deforms compliant skin
Internal markers shift under load
Embedded optoelectronic sensor captures deformation field
Vision pipeline extracts tactile features
Robot policy uses touch for action
Lin et al., "Tactile Gym 2.0", IEEE Robotics and Automation Letters, 2022·Lepora & Lloyd, "Pose-Based Tactile Servoing", IEEE Robotics & Automation Magazine, 2021
Simulation · TactileLab
TactileLab is an end-to-end, GPU-accelerated tactile simulation and learning framework built on NVIDIA Isaac Lab. Rather than slowly simulating elastomer deformation, it represents contact with two complementary signals — normal contact depth and a dense tangential (shear) displacement field — a compact, transferable tactile observation that runs across thousands of parallel environments for large-scale reinforcement learning. A multi-modal real-to-sim pipeline then carries policies from simulation onto real multi-finger robot hands.
Shear-sensitive tactile rendering
A multi-finger Allegro hand manipulates an object in simulation while each fingertip’s tactile signal is rendered live — contact depth plus an in-plane tangential field that captures shear, slip and micro-motion. This shear-sensitive representation gives in-hand manipulation policies far more than depth alone, and GPU parallelism lets thousands of these environments train at once.
Tactile Lab · NVIDIA Isaac Lab
Tactile Lab — our new simulator built on NVIDIA Isaac Lab — trains tactile policies across thousands of GPU-parallel environments for dramatically faster learning. The resulting policy drives a real robot to follow an object’s trajectory using touch alone (shown at 6× speed), carrying control straight from large-scale simulation to physical hardware.
"TactileLab: Efficient Shear-Sensitive Tactile Simulation for Sim-to-Real Multi-Finger Dexterous Manipulation", 2026 (under double-blind review)·Mittal et al., "Isaac Lab: A GPU-Accelerated Simulation Framework for Robot Learning", NVIDIA, 2024·Church et al., "Tactile Sim-to-Real Policy Transfer via Real-to-Sim Image Translation", CoRL, 2022
Application demo
The demo illustrates how TacTip can capture subtle skin-surface deformation during fingertip contact. In a pulse-waveform research setup, the soft dome conforms gently to the wrist or fingertip while the internal optoelectronic sensor tracks micro-motion patterns, turning a weak mechanical signal into a rich tactile waveform for waveform research and wearable prototyping.
Lepora, "Soft Biomimetic Optical Tactile Sensing with the TacTip: A Review", IEEE Sensors Journal, 2021
Heritage
2009
The first TacTip is developed at Bristol Robotics Laboratory, inspired by the dermal papillae of human skin.
2018
Soft Robotics publishes the family paper, exploring 3D-printed biomimetic tactile sensors in dome, gripper, manipulation and cylinder morphologies.
2021
High-fidelity tactile simulation and sim-to-real deep RL bring TacTip into the learning era.
2022
TacTip sensing merges with the DIGIT platform; one sim-to-real pipeline spans three sensor types.
2023
Bimanual tactile manipulation policies trained in simulation deploy directly to dual-arm hardware.
2026
A GPU-accelerated, shear-sensitive tactile simulator built on NVIDIA Isaac Lab — massively parallel training and real-to-sim transfer for multi-finger dexterous manipulation.
2026
TacTip Robotics brings the lineage to industry: sensors, the TactileGym simulation platform, and supply-chain customisation.
Chorley et al., ICAR, 2009·Ward-Cherrier et al., Soft Robotics, 2018·Lin et al., IEEE RA-L, 2022–2023
Our team
Research, product engineering, software, manufacturing and customer applications come together in one team. We build optoelectronic tactile sensing around real contact tasks.
Partners
Our partner network is still growing. We welcome universities, labs, industry teams and investors to join the TacTip ecosystem.
Hover to reveal names
...
More partners joining
