In the murky depths where cameras fail and sonar fragments, a former Ecovacs engineer named Cao Jinlong has identified a quiet but consequential gap: underwater robots can move, but they cannot truly see. This summer, his startup Xinyang Micro secured $1.6 million to bring acoustic perception — long trusted in medicine — into the world of autonomous machines, beginning with the humble swimming pool. It is a small entry point into a vast question about how machines will come to understand environments that human senses were never built for.
Underwater sensing startup Xinyang Micro raises $1.6M for autonomous pool robot vision
The robot can move, but it doesn't know what it's cleaned
Mark
Why does a pool robot need to know where it's been? Can't it just follow a pattern?
Mimi
A pattern assumes the pool is always the same—same shape, same obstacles, same debris distribution. But pools change. A leaf falls in. The pump moves. A child's toy ends up on the floor. A robot that can't perceive can't adapt. It cleans the same path twice and misses the corner entirely.
Mark
And cameras don't work underwater?
Mimi
Not in real conditions. Sediment, algae, particles—they cloud the lens within minutes. Ultrasound doesn't care. It travels through turbid water the same way it travels through clear water. It's the physics of the medium.
Mark
So this is really about autonomy becoming viable?
Mimi
Exactly. Remote-controlled robots can work with bad perception because a human is watching and correcting. But autonomous robots need to know their environment. Without that, they're just expensive toys.
Mark
Why start with pools instead of going straight to marine applications?
Mimi
Volume and simplicity. Pools are controlled environments. The geometry is predictable. The market is proven and hungry for better solutions. You build the technology, prove it works at scale, and then you have the credibility and the capital to tackle harder problems—deep water, salt water, extreme conditions.
Mark
What's the real bet here?
Mimi
That acoustic imaging becomes as standard in underwater robotics as cameras are in drones. Right now it's novel. In five years, it could be table stakes.
Le Pouls
- Pool-cleaning robots have been navigating blind for years — gyroscopes drifting, ultrasound giving fragments, cameras defeated by cloudy water — leaving coverage more like guesswork than cleaning.
- Xinyang Micro's CMUT-based acoustic module etches ultrasonic arrays directly onto silicon, producing compact 3D maps of pool floors and walls that turbidity cannot disrupt.
- A $1.6M seed round backed by Peak Ventures and Infini Capital gives the team runway to hire and harden the technology toward a claimed 95%+ cleaning coverage threshold.
- The $2.8B pool robot market — 4.9 million units shipped in 2025 alone — offers a high-volume, consumer-grade proving ground before the technology scales toward a $19.66B underwater robotics sector by 2034.
- As autonomous underwater robots expand into marine inspection and research, acoustic perception is positioning itself not as a niche feature but as foundational hardware infrastructure for the entire industry.
In the murky depths where cameras fail and sonar fragments, a former Ecovacs engineer named Cao Jinlong has identified a quiet but consequential gap: underwater robots can move, but they cannot truly see. This summer, his startup Xinyang Micro secured $1.6 million to bring acoustic perception — long trusted in medicine — into the world of autonomous machines, beginning with the humble swimming pool. It is a small entry point into a vast question about how machines will come to understand environments that human senses were never built for.
Cao Jinlong spent years at Ecovacs watching pool robots move without truly knowing where they had been. Gyroscopes drifted. Ultrasound sensors returned fragments. Vision systems collapsed the moment water turned cloudy. The robots could follow a path — but they couldn't map their own progress.
This August, Cao's startup Xinyang Micro closed an 11 million yuan round — roughly $1.6 million — backed by Peak Ventures and Infini Capital. The capital will fund hiring, but the real story is the technology: a perception module combining a CMUT array, a custom ASIC chip, and software designed to give underwater robots something close to genuine spatial awareness in murky conditions.
CMUT — Capacitive Micromachined Ultrasonic Transducer — is familiar from medical imaging, but Xinyang Micro's approach is distinct. Rather than bulky piezoelectric ceramics suited to long-range sonar, the company uses MEMS fabrication to etch micro diaphragms onto silicon wafers, arranging them into dense phased arrays. The result maps a three-dimensional acoustic image within roughly 20 meters. Sediment doesn't degrade it. Turbidity is irrelevant. The company claims this pushes cleaning coverage above 95 percent — a significant leap from the incomplete passes current robots manage.
The immediate market is concrete and large: 34 million pools worldwide, 4.9 million robot units shipped in 2025, $2.8 billion in global retail sales, and projections climbing toward 7 million units by 2030. Chinese manufacturers dominate the segment, and demand for genuine automation — not just programmed motion — is clear.
But Xinyang Micro is using pools as a beachhead. The broader underwater robotics market, spanning inspection drones to marine research platforms, sits at $5.82 billion today and is projected to reach $19.66 billion by 2034. As those systems shift from remote operation to true autonomy, perception becomes the defining bottleneck. Acoustic imaging, proven first in consumer pools, could become standard hardware across the entire category — a quiet technology finding its depth.
Cao Jinlong spent years at Ecovacs watching pool-cleaning robots bump around in water, and he noticed something that bothered him: they didn't actually know what they were doing. The machines could move. They could follow a programmed path. But ask them which corners of the pool they'd already cleaned and which ones they'd missed, and the answer was essentially a guess. Gyroscopes drifted. Single-point ultrasound sensors gave only fragments of information. Vision systems, the obvious choice, failed the moment sediment clouded the water.
This summer, Cao's startup, Xinyang Micro, closed an 11 million yuan funding round—about $1.6 million—to solve that problem. Peak Ventures and Infini Capital backed the deal, announced in August 2026. The money will go toward hiring and building out the team, but the real asset is the technology: a perception module that combines a CMUT array, a custom ASIC chip, and software to give underwater robots something like actual vision in murky conditions.
CMUT stands for Capacitive Micromachined Ultrasonic Transducer. It's not new—medical ultrasound has used it for years—but applying it to underwater robotics required a different approach. Traditional piezoelectric ceramic transducers work well for long-range sonar and deliver high power, but they're bulky and hard to pack densely. Xinyang Micro uses MEMS semiconductor technology instead, etching micro diaphragms directly onto silicon wafers and arranging them into phased arrays. The result is a compact, high-density sensor that can map a three-dimensional acoustic image of the pool floor and walls within about 20 meters. Water turbidity doesn't matter. Particles don't degrade the signal. Ultrasound just works.
The company claims its module pushes cleaning coverage above 95 percent—a meaningful jump from the scattered, incomplete passes that current robots make. Cao told 36Kr that this gap between motion and awareness was the core problem: robots could move, but they couldn't reliably know where they'd been or where they needed to go. Acoustic imaging closes that gap.
The market is substantial. There are roughly 34 million swimming pools worldwide as of 2025, and the pool-cleaning robot segment shipped about 4.9 million units that year, with projections to reach 7 million by 2030. Global retail sales hit $2.8 billion in 2025, and the segment is dominated by Chinese manufacturers. It's a proven, growing market with clear demand for better automation.
But Xinyang Micro is thinking bigger. The broader underwater robotics market—everything from inspection drones to marine research platforms—was valued at $5.82 billion in 2025 and is expected to reach $19.66 billion by 2034. As those robots shift from remote operation to autonomous navigation, perception becomes the critical bottleneck. Acoustic imaging could become standard hardware across the entire category. Pool cleaning is the beachhead: a consumer application with high volume, clear ROI, and room for innovation. Master that, and the path to marine applications—deeper, darker, more complex—becomes visible.
Citations marquantes
The robot can move, but it is difficult to accurately know which areas have been cleaned and which have not.— Cao Jinlong, founder of Xinyang Micro, speaking to 36Kr
Envie de l'histoire complète ?
Lire l'original sur app.dealroom.co ↗
Technology