By Da Cheung
For years, modern imaging and robotics have been giving machines an ever-clearer view of the physical world. However, endowing them with a sense of touch — specifically, the ability to gauge physical resistance and force in confined spaces — has remained a stubborn bottleneck.
Now, a new generation of micro-sensors is emerging from Chinese universities, utilizing novel mediums like light and heat to replace bulky electronic components. These breakthroughs could provide a much-needed “tactile interface” for humanoid robots and minimally invasive surgical tools, transforming how machines interact with fragile human tissues and complex physical tasks.
At the heart of this shift is the six-axis force sensor, a device designed to measure physical movement, orientation, and force in three-dimensional space. While traditional versions — which account for more than 80% of the market — rely heavily on metallic strain gauges and complex wiring, their reliance on electronics has prevented them from being miniaturized for millimeter-scale environments. Yet the financial incentive to shrink these sensors is immense. The global market for six-axis force sensors is projected to surge from $356 million in 2025 to $1.9 billion by 2032, driven by a blistering 58% global annual growth rate in the humanoid robotics sector, according to market research firm LP Information.
Bypassing electronics with light
To address the space constraint, a research team at Shanghai Jiao Tong University (SJTU), led by Associate Professor Yang Jianlong, opted to remove electronic components entirely. In an article published by the academic journal Optica in May 2026, the team unveiled an all-optical sensor with an outer diameter of just 1.7 millimeters. Tech media DeepTech called it the world’s smallest six-axis force sensor.
Rather than relying on electronic wires, the SJTU sensor uses a soft silicone tip combined with a coherent fiber bundle. When the sensor presses against an object, the tip undergoes microscopic deformations that alter the internal light distribution. These shifting light patterns are transmitted back to a camera, where an AI algorithm acts as a translator, decoding the light images into precise multidimensional force readings.
Yang’s inspiration stemmed from his previous work with optical coherence tomography — a high-resolution optical imaging technique — in neonatal intensive care units. He realized that modern surgical tools and robots can see clearly but often “touch blindly,” struggling to gauge how much force they are applying. To demonstrate this optical sensor’s potential, the researchers conducted simulated tumor palpation experiments. By scanning the sensor over a gelatin model, the system successfully detected and located a hidden hard sphere by sensing shifts in physical resistance — showcasing its promising application in minimally invasive robotics, flexible instruments, and biomedical operations.
Feeling with heat and flexibility
In parallel, researchers at Tsinghua University, led by Professor Zhu Rong, have approached the tactile challenge using heat. According to research published in Nature Communications, Zhu’s team developed an ultralight, flexible sensor that weighs only 0.3 grams and easily fits on a fingertip.
The Tsinghua sensor relies on a layered design of thin-film thermistor, a type of sensor whose electrical resistance changes in response to temperature, to detect changes in an elastic piezo-thermic material — a specialized substance that alters its thermal properties when subjected to physical strain. This simple structure avoids complicated manufacturing while maintaining a high degree of sensitivity to normal force, lateral force, and rotational torque.
The applications of this flexible sensor span from robotic gripping to human assistance. When attached to a robotic hand, the sensor allows delicate operations like opening a childproof medicine bottle. Alternatively, disabled or elderly individuals could wear the sensors on their fingertips to easily teleoperate assistive robots or manage smart home devices with slight touches. The research team noted that the device passed 100,000 cycles of repeated-use testing without suffering interference from changes in ambient temperature and humidity.
Navigating the path to real-world applications
Despite promising laboratory results, making these sensors robust enough for real-world application introduces significant challenges.
For optical sensors relying on AI, “hallucination” is a major risk. To calibrate the SJTU sensor, researchers had to use generative AI diffusion models — algorithms that create realistic synthetic data — to generate around 100,000 artificial images to represent vast multidimensional forces. However, Yang notes that in a clinical setting, an AI system must recognize its own boundaries. He argues that a medical AI should not be an unconstrained self-learning black box, but rather a supervised system that knows when its readings are no longer reliable.
Furthermore, the properties of materials change over time. The silicone used in optical sensors can slowly relax under continuous pressure, causing slight drifts in data. Both research teams also face hurdles in standardizing the manufacturing process. Small deviations during fabrication can alter the baseline properties of each sensor, currently necessitating individual calibration for every device.
For decades, the heavy lifting of industrial automation relied on rigid, high-capacity sensors. But as automation moves from the factory floor into hospitals, homes, and care facilities, the requirements have fundamentally changed. While mass-production costs and exact lifespans remain unclear, these early micro-sensors highlight a critical transition. By converting mechanical contact into optical or thermal signals, they aim to fill a critical void, granting machines the nuance needed to handle a delicate world.
Sources
