KiriSense technical insights
Tactile Sensing for Dexterous Robot Hands and Humanoids
Robot hands need contact feedback to coordinate grasping, lifting and in-hand movement. Fingertip sensing is one part of that system.
By KiriSense Ltd · Published 2026-06-02 · Updated 6 October 2026. Technology lead: Kangsheng Bretherton-Liu.
A robot hand has to manage several contacts
Dexterous manipulation changes an object’s position or orientation through coordinated finger motion and contact. A multi-finger hand must establish a grasp, share loading between contacts and respond when the object moves. A humanoid robot adds whole-body movement and tasks involving objects designed for human use.
Vision provides scene and object context. Fingertip sensing can provide local evidence about force and deformation inside a grasp, including interfaces hidden from an external camera. A tactile sensor does not by itself make a hand dexterous: mechanics, actuation, planning and control must work together.
What a fingertip measurement can tell a controller
Normal force describes loading approximately perpendicular to a contact surface. Shear force describes tangential loading. Their relationship changes as a hand lifts, turns or repositions an object. Directional shear measurements can help a controller understand how a contact is being loaded, provided the sensor has been calibrated for that purpose.
Slip is movement at the interface. Incipient slip may begin locally before the whole contact slides. Appropriate tactile measurements can help detect those changes, but shear sensing and validated slip detection are separate capabilities. A robust controller must also decide what action to take after detecting them. Read the distinction between force and slip sensing.
What tactile sensors are used in dexterous hands?
Published research includes image-based optical fingertip sensors, force and pressure arrays, and magnetic or hybrid sensors. DIGIT is a compact camera-based tactile sensor demonstrated in multi-finger in-hand manipulation. GTac combines force-sensing approaches to provide heterogeneous tactile feedback. These examples illustrate several viable ways to gather robotic touch information.
Choosing a sensor means choosing the information needed by the task. Contact geometry, force direction and slip cues have different requirements. Fingertip size, wiring, replaceability, calibration, durability and control-loop timing all matter. A compact sensor can still require substantial integration work.
Why data and timing matter for embodied AI
Embodied AI systems learn or execute actions in the physical world. Touch can connect an action with its contact consequences. Multiple fingertips also create multiple streams of data. Sampling rate, timestamps, transport and interpretation affect how quickly the controller receives useful information.
A low-data architecture could make integration simpler if it retains the signals required for the task. Detailed images may instead be valuable when local geometry and motion must be recovered. Neither data volume nor a fast sensor alone establishes end-to-end control latency; the full readout, communication and control system must be measured.
Grippers and humanoids share a practical need
Tactile sensing for grippers can support grasp adjustment during handling without requiring a human-like hand. Humanoid and dexterous robots may need more varied contacts and tasks. Both require evidence that sensing remains reliable across relevant materials, contact locations and motions.
KiriSense targets robot fingertips, dexterous hands and grippers as its technology progresses towards integration. These markets motivate the compact-signal architecture and its integration programme. The commercial context for humanoid dexterity explains why manipulation matters to the investor story.
KiriSense’s approach to robotic touch
KiriSense Ltd develops geometry-encoded robotic tactile sensing for robot fingertips and dexterous manipulation. Physical geometry encodes tactile information before downstream computation. The architecture is designed to sense normal, shear and slip information through compact signals rather than a conventional high-bandwidth tactile image stream, targeting low-cost, low-data and low-latency feedback.
Explore KiriSense’s geometry-encoded tactile sensing architecture. Read the latest proof-of-concept results for the dated evidence from the company’s research collaboration with the University of Sheffield. Technology origin and company-owned IP and technology lead Kangsheng Bretherton-Liu explain who is behind it.
Primary research
Dong et al. (2018), Maintaining Grasps within Slipping Bound by Monitoring Incipient Slip
The references support the general science and examples discussed above. KiriSense-specific claims are described in its linked proof-of-concept report.