Tactile sensing is the ability of a robotic or prosthetic system to detect and interpret physical contact information — including contact force magnitude, direction, distribution, texture, slip, and temperature — through sensors embedded in or on the surface of an end effector or robotic skin. It is the mechanical analogue of the human sense of touch and provides information that visual sensing alone cannot supply, such as the internal stress distribution of a grasped object or the onset of slippage. Tactile sensing is fundamental to dexterous manipulation, safe human-robot interaction, and feedback-controlled assembly, and has become a defining capability differentiating next-generation robotic systems from conventional industrial manipulators.
Content
- Interest in robotic tactile sensing dates to the 1970s when researchers recognised that the grasping failures of early industrial robots stemmed from their inability to sense contact forces. Initial sensors were simple binary contact detectors or single-axis force gauges at the wrist. The development of multi-axis force/torque sensors with six degrees of freedom in the 1980s enabled more sophisticated compliant manipulation, but these remained expensive and fragile. Research groups at MIT, Carnegie Mellon, and Stanford drove early advances in tactile array sensors using resistive, piezoelectric, and optical transduction principles.
- Modern tactile sensors fall into several families. Resistive and capacitive arrays detect contact location and normal force through changes in electrical properties under mechanical deformation. Piezoresistive sensors — including flexible MEMS arrays — provide high spatial resolution and sensitivity. Optical tactile sensors such as the GelSight family image the deformation of a compliant gel surface illuminated from inside, recovering rich three-dimensional contact geometry from a standard camera sensor. Magnetic tactile sensors embed ferromagnetic markers in an elastomer and track their displacement with magnetic field sensors. Visuotactile sensors combine optical imaging with deformable contact surfaces to provide contact geometry, force distribution, and texture information at camera frame rates.
- Tactile sensing is significant because dexterous manipulation — picking up a strawberry without crushing it, threading a needle, assembling small electronic components — requires closed-loop contact force regulation that vision-based systems cannot provide at the required bandwidth and spatial resolution. As robots move from structured industrial cages into unstructured environments shared with humans — healthcare, domestic service, flexible manufacturing — tactile sensing becomes essential for safety and capability. The difference between a robot that can manipulate only hard, rigidly positioned objects and one that handles soft, deformable, and variably positioned objects is substantially determined by tactile sensing quality.
- By 2024–2025, tactile sensing is experiencing rapid progress driven by soft robotics, neuromorphic sensing, and data-driven approaches. Soft robotic fingers instrumented with distributed tactile arrays are enabling human-level grasping performance across diverse object categories. Neuromorphic event-driven tactile sensors — inspired by the spike-based signalling of biological mechanoreceptors — offer microsecond temporal resolution and ultra-low power consumption. Deep learning models trained on large tactile datasets are learning to classify materials, estimate object properties, and predict grasp stability from raw sensor readings. Integration of tactile sensing with large visual-language models is enabling robots to combine verbal instructions with haptic feedback for adaptive manipulation in open-ended tasks.