AI Gains the Ability to Feel Surfaces Using Quantum Technology

ronaldfrenn

While AI has advanced in areas such as vision, conversation, and computation, it has struggled with a fundamental mechanical function: the ability to measure or "feel" surfaces. However, researchers at Stevens Institute of Technology's Center for Quantum Science and Engineering (CQSE) have developed a method that enables AI to sense surfaces, a breakthrough in tactile recognition.

According to Stevens physics professor Yong Meng Sua, "AI has acquired the sense of sight through computer vision and object recognition, but it has yet to develop a human-like sense of touch, such as distinguishing between rough newspaper paper and smooth magazine paper."

In a recent breakthrough, Sua, along with CQSE Director Yuping Huang and doctoral candidates Daniel Tafone and Luke McEvoy, demonstrated a system that combines quantum technology with AI. Their setup uses a photon-emitting scanning laser and AI models trained to analyze surface textures captured by the laser.

"This is a marriage of AI and quantum," Tafone explains.

Their system, outlined in the journal Applied Optics, works by pulsing a specially created beam of light at a surface to "feel" it. The photons bounce off the surface, carrying back speckle noise—random imperfections in the image. While speckle noise is typically seen as a problem for clear imaging, this team uses it to their advantage, processing it with AI models trained to interpret it as valuable data. This approach allows the system to accurately discern the surface’s topography.

The team tested their system using 31 industrial sandpapers with varying roughness, ranging from 1 to 100 microns thick. (For reference, a human hair is about 100 microns thick.) The laser pulses passed through transceivers, bounced off the sandpapers, and returned for analysis by the AI model.

Initial tests showed a root-mean-square error (RMSE) of about 8 microns. However, after refining the system with multiple samples, the accuracy improved to within 4 microns, comparable to the most advanced industrial profilometer devices.

Interestingly, the system worked best on finer-grained surfaces, such as diamond lapping film and aluminum oxide.

This new method has broad potential applications. For instance, in skin cancer detection, the system could identify minute surface roughness differences that are too small for the human eye but are significant enough to distinguish harmless conditions from dangerous melanomas.

Huang elaborates, "Tiny differences in mole roughness, undetectable by human examiners, could be measured using our quantum system, aiding early detection."

In manufacturing, the technology could improve quality control by identifying tiny surface defects that could lead to mechanical failures, ensuring that parts meet strict specifications.