KiriSense demonstrates core tactile sensing principle for embodied AI and robotics

By KiriSense Ltd · Updated 6 October 2026. Technology lead: Kangsheng Bretherton-Liu.

KiriSense has reached an important milestone in the development of its tactile sensing technology for embodied AI and robotics, with proof-of-concept testing at the University of Sheffield demonstrating the core physical principle behind its approach.

The tests show that deformation of the KiriSense kirigami structure produces changes in its optical response that depend on how the structure is mechanically loaded.

More importantly, normal and shear loading produce visibly different behaviour. Shear produces a substantially stronger spatial asymmetry in the optical response than normal force, consistent with the kirigami geometry acting as a directional mechanical encoder.

Put simply, different ways of touching the sensor produce distinguishable optical responses.

Why that matters

AI and computer vision have transformed what robots can see and recognise. But vision alone cannot tell a robot everything that happens once it makes physical contact with an object.

A robotic hand or gripper needs to know how hard it is gripping, whether a load is moving sideways and whether an object is beginning to move or slip.

KiriSense is developing tactile intelligence to provide that information directly at the point of contact.

Its approach combines a compliant kirigami structure with optical sensing to create a compact, camera-free tactile sensor. The architecture is being developed to resolve normal force, shear and movement while producing relatively small amounts of local sensing data.

That local contact information can then become part of the feedback loop controlling a robotic hand or gripper.

From sensing concept to physical evidence

The latest experiments provide credible evidence that the kirigami geometry produces deformation-dependent and spatially asymmetric optical signatures, and that those responses appear reversible as the structure is loaded and released.

They also provide clear qualitative evidence that shear produces greater asymmetry than normal loading.

For KiriSense, this is an important transition: from a sensing architecture based on a physical hypothesis to experimental evidence that the underlying mechanism works.

Dr Kangsheng Bretherton-Liu, CEO and Technical Lead at KiriSense, said:

“These results give us experimental evidence that the kirigami structure is behaving as we intended. Different mechanical inputs are producing distinguishable optical responses, particularly under shear. The next step is to turn those qualitative responses into controlled, quantitative sensing data.”

The work forms part of KiriSense’s proof-of-concept programme with the University of Sheffield, supported through an £82,692 Henry Royce Institute Industrial Collaboration Programme project.

University of Sheffield research collaborators Dr Shuhei Miyashita, Senior Lecturer in Robotics, and Dr Jialun (Chester) Liu, Research Assistant, are collaborating with KiriSense on experimental testing and characterisation of the company’s sensing architecture.

Giving embodied AI a sense of touch

The wider challenge is becoming increasingly important as robotics moves beyond repetitive operations in highly structured environments.

Fragile products, deformable materials, irregular shapes and mixed objects all present a different problem from conventional industrial automation. A robot cannot rely solely on knowing where an object is. It also needs feedback about what is happening as it interacts with it.

Tim Harper, Chair of KiriSense, said:

“Embodied AI ultimately has to deal with the physical world, and the physical world starts behaving differently the moment you touch it. Vision can tell a robot what an object is and where it is. Touch tells it what is happening now.

“We’ve moved beyond the design hypothesis and can see the underlying sensing mechanism working experimentally. That is an important step towards giving robotic hands and grippers useful contact intelligence at the fingertip.”

KiriSense is developing the technology as a sensing layer for robotic hands, grippers and end-effectors rather than building complete robots. Applications include industrial automation, logistics, food handling, dexterous manipulation and emerging humanoid robotic platforms.

The company has filed four UK patent applications covering aspects of its tactile sensing technology and is a member of NVIDIA Inception.

What comes next

The current results establish the core sensing principle. The next phase moves from qualitative demonstration towards quantitative characterisation.

Work now focuses on optimising the optical configuration, integrating improved electronics and measuring the response under controlled normal and shear loads. This provides the basis for calibration, repeatability testing and subsequent work on detecting movement and the onset of slip.

The commercial objective is straightforward: put KiriSense sensors onto real robotic hands, grippers and end-effectors and demonstrate where better contact information produces better robot performance.

KiriSense adds a critical missing sense to embodied AI: touch.

Back to News