Global-first pulse-diagnosing robotic hand shown at WRC 2026; startup says tactile data trust is bottleneck

At the WRC 2026 booth, the hand, which is specialized for pulse diagnosis, made its global debut. Its three fingers rest on a person's wrist at the cun, guan and chi positions and perform three traditional levels of pressure. Pulse-condition analysis runs at the edge, with further diagnostics processed in the cloud. The system can currently recognize 28 pulse patterns and provide recommendations based on traditional Chinese medicine knowledge.

Youren founder and CEO Zhang Zhen told Leiphone.com in a Sept. 3 exchange that the company chose pulse diagnosis because it is the ultimate expression of human touch. "Each finger has ten high-precision probes, like ten ultrasound units. Traditional Chinese medicine is like a B-scan performed with fingertip touch," he said. Pulse sensing demands parallel processing by four classes of tactile receptors: Meissner corpuscles detect low-frequency vibration and sliding, Merkel discs sense static pressure and texture, Pacinian corpuscles catch high-frequency vibration, and Ruffini endings perceive sustained stretch and shear. Zhang contended that the roughly 25 tactile sensor companies that have entered the field in the past two years mostly rely on a single principle such as capacitance, piezoresistance, piezoelectricity, magnetic induction or visuotactile sensing, which cannot reproduce the cooperation of all four receptor types.

Zhang rejected the dominant narrative that embodied intelligence simply lacks tactile data. The real bottleneck is high-quality data supply, he said. Untrustworthy data cannot enter training loops no matter how plentiful. He cited the T-REX paper co-authored by Fei-Fei Li earlier in the year showing that adding tactile data to a vision-language-action model could degrade, rather than improve, manipulation performance. In his view, touch is the only sensor domain spanning both perception and physical control. When used for perception, a 30 Hz sampling rate is sufficient; when used in control loops, the sensor must respond at 100 to 500 Hz or faster. Making a sensor extremely sensitive usually makes it soft, which slows response, while making it fast requires hardness, which reduces its ability to adapt to surface shape. Zhang gave an example of a real manufacturing failure: during precision insertion of board-to-board connectors, a flexible piezoresistive sensor with enough sensitivity still produced 10–20 Hz oscillation in a six-axis arm, bending the pins.

Rather than push a single sensing technology to its limit, Youren designed its fingertip as a fusion of three transduction mechanisms: capacitive sensing for static pressure and distribution imaging, piezoresistive sensing for wide-range deep pressure and overload resistance, and piezoelectric sensing for vibration, slip and transient impacts. This mimics fast- and slow-adapting human receptors. A gradient in elastic modulus gives a soft surface and a stiff bottom to maintain both sensitivity and overload capacity. The company's manufacturing process, free-form printing, deposits multiple materials into integrated structures and can conformally cover curved or irregular surfaces. The company describes the result as "skin, not a patch."

The approach is encapsulated in a framework called CrediTac, in which sensor data quality is said to contribute 90 percent of the trustworthiness, algorithm design 9 percent and system implementation 0.9 percent. Zhang said the company's delivery goal is tactile data with 99.9 percent credibility. "The industry is comparing who measures more accurately; we compare whether AI dares to use it," he said.

Zhang said pulse diagnosis was initially only a technical demonstration to prove the hardest tactile challenge. But after many visitors asked about diagnostic accuracy at WRC, the team turned it into a separate business plan focused on triage in Chinese medicine hospitals, general health channels and future home-based pulse checks. Youren is collecting data at a Chinese medicine hospital and is cooperating with a Chinese medicine large-model company to improve diagnostic accuracy.

The founder's engineering background shapes the company's identity. Zhang, a Harbin Institute of Technology graduate, has 20 years of experience in semiconductor and sensor product engineering and marketing, and previously scaled a flexible sensing technology at a consumer electronics supply-chain company. The core team is drawn from major international manufacturers. Zhang contrasts his company with academic groups that stop at the model or material level: "When we talk about touch, we first talk about process, yield and reliable mass production, not parameters and models."