Robots Are Learning to Feel: Nanoscale Sensors Could Finally Crack the Tactile Sensing Problem

Walk into any advanced robotics lab today and you'll find machines that can navigate, reason, and manipulate objects with increasing confidence. What you won't find is a robot that truly feels what it's holding. That gap — the absence of meaningful tactile feedback — remains one of the most underappreciated barriers between today's capable automatons and the adaptable, general-purpose robots the industry is promising. A German startup called Digid thinks nanoscale sensors might be the answer.

In an interview published this week by Robotics & Automation News, Digid co-founders Nils Konne and Christian Kreil outlined their approach: sensors small and sensitive enough to give robotic hands and surfaces something resembling the tactile resolution humans take for granted. It's a problem that's been called robotics' "final frontier" for good reason — and the Digid team is betting that going down to the nanoscale is where the breakthrough lies.

Why Touch Is Hard (and Why It Matters More Than Ever)

Humans have roughly 17,000 mechanoreceptors in each hand, giving us the ability to detect pressure, vibration, texture, and temperature with extraordinary precision. We can feel the difference between a grape and a glass marble, adjust grip force in real time, and detect a slipping object before we consciously register it.

Current robotic grippers mostly can't do this. They rely on cameras, force-torque sensors at the wrist level, or basic pressure sensors that provide coarse, low-resolution feedback. The result is robots that handle objects cautiously, slowly, and with wide safety margins — none of which is compatible with the speed and delicacy required in real-world manufacturing, surgery, or household tasks.

This isn't just an academic complaint. As humanoid robots move from demos toward actual deployment — in factories for Schaeffler, warehouses for Amazon, and soon (according to Tesla's ambitions) general home use — the absence of real tactile sensing becomes a hard operational ceiling. A robot that can't feel what it's gripping will always need wider error margins, slower speeds, and more structured environments. That's expensive.

The Nanoscale Approach

Digid's bet is that existing sensor technologies are simply too big, too rigid, and too insensitive to replicate the density of human touch receptors across a robot's surface. Their nanoscale sensors — details of which remain somewhat proprietary — aim to pack far greater sensing resolution into flexible, conformable materials that can coat fingertips, palms, and even robot skin.

The key insight from Konne and Kreil is that the problem isn't purely computational. Even the best AI in the world can't compensate for a sensor that simply doesn't capture the signal you need. Solving tactile sensing is fundamentally a hardware problem first, and the software benefits follow. It's the same logic that's driven breakthroughs in computer vision: better sensors (higher-resolution cameras, better optics) enabled better algorithms, not the other way around.

This framing matters for investors and observers watching the physical AI space. Much of the current excitement around companies like Figure AI, Agility Robotics, and 1X centers on software — foundation models, imitation learning, reinforcement learning from human feedback. But as these systems push into more complex manipulation tasks, the hardware floor will reassert itself. A robot learning from millions of demos still needs reliable sensory input to generalize what it's learned.

The Broader Trend: Robotic Skin Is Having a Moment

Digid isn't alone in chasing this problem. MIT, Stanford, and Carnegie Mellon have active research programs in tactile robotics. Startup GelSight (now part of a larger sensor ecosystem) developed gel-based optical tactile sensors that have been adopted in research settings. Meta AI's research arm has published work on using tactile sensors to improve manipulation policies. And NVIDIA's Halos safety framework — just released this week — implicitly acknowledges that better sensing is key to safer human-robot co-existence.

What makes this moment different from earlier tactile sensor research is commercial urgency. When humanoid robots were a research curiosity, "good enough" tactile sensing was acceptable. When they're being ordered in the thousands — Schaeffler recently announced plans for 2,000 units — the economic pressure to solve this at scale becomes very real.

What to Watch

Digid's technology is still early-stage, and the company hasn't disclosed deployment timelines or production costs. The key questions for investors and robotics enthusiasts alike are: Can nanoscale sensors be manufactured reliably at the volumes robot makers need? Can they survive the mechanical stress of real-world use? And critically, can the data they produce be integrated cleanly into the control pipelines that modern robot AI expects?

If those questions can be answered in the next 18-24 months, companies working on tactile sensing — including Digid and its competitors — could become critical suppliers to the humanoid robot wave. Think of them as the sensor foundries of the coming physical AI era.

For robotics enthusiasts who want to go deeper on the hardware side of this story, The Art of Doing Science and Engineering offers rich perspective on why sensing and feedback are the soul of intelligent systems. And for those tracking the investment angle, platforms like Robinhood or Interactive Brokers make it easy to follow publicly-traded robotics hardware companies as the tactile sensing race heats up.

The robots can see. They're getting smarter by the day. The next frontier is teaching them to feel — and the nanoscale engineers at Digid are quietly working on exactly that.

---

Source: Robotics & Automation News — Interview with Digid's Nils Konne and Christian Kreil