Latest / The Robotics Podcast with Fexingo: Autonomous Systems, Industrial Robots, and Hardware / Why Robot Arms Still Cant Handle a Wet Paper Towel
Transcript
- Lucas: So we've spent a bunch of episodes on things like raw eggs, wet soap, fitted sheets — all these objects that change shape the second you touch them. But there's one everyday item that combines almost every bad property at once, and we haven't talked about it yet. Luna: Let me guess — something that starts dry and then gets wet and floppy? Lucas: Exactly. The wet paper towel. It's the perfect storm. Low stiffness, high adhesion when wet, tears easily, and the geometry changes completely from one second to the next. A robot that can handle a dry paper towel is one thing — that's already hard. But wet? That's a whole different level. Luna: I can't think of a single commercial robot that can do this reliably. Maybe a research demo, but nothing production-ready. Lucas: Right. And the reason gets into some really interesting physics. There was a paper from MIT CSAIL last year — 2025 — that specifically studied the forces involved in pulling a wet paper towel off a roll. They used a high-speed force sensor and a camera array. What they found is that the force required to tear a sheet varies by a factor of three depending on how fast you pull and how wet the towel is. Luna: So a robot that applies a consistent grip force will either slip off or tear the towel on the next pull. There's no middle ground. Lucas: That's the core problem. A human intuitively adjusts — we feel the resistance and either increase grip or slow down. But a typical robot gripper has maybe one or two pressure settings. It can't sense the subtle change in capillary adhesion as the water spreads between the gripper pad and the towel surface. Luna: Capillary adhesion — that's when water molecules create a thin film that essentially glues the towel to the gripper, right? Lucas: Yes. And that film is incredibly strong at small scales. The force from surface tension can actually exceed the towel's own tensile strength. So the robot tries to release the towel, but the water film holds it, and the towel tears. That's why you see those research videos where the arm just flails around with a shredded piece of towel stuck to its finger. Luna: I've seen those. It's almost comical — but it's a genuine materials science challenge. What are the leading approaches to solve it? Lucas: One promising direction is hydrophobic gripper surfaces. If you coat the finger pads with a material that repels water, the capillary adhesion drops significantly. Researchers at Carnegie Mellon tested a fluoropolymer coating and got a 60 percent reduction in adhesion force. But then you lose some friction on dry surfaces, so it's a trade-off. Luna: Another approach is to use active drying — like a tiny air jet that blows the water film away before release. But that adds complexity and cycle time. Lucas: Yeah, and in a production setting, those extra seconds add up. The most elegant solution I've seen uses a granular jamming gripper — the kind that uses coffee grounds or glass beads that stiffen under vacuum. When the gripper is soft, it conforms to the towel shape. Then it vacuums to stiffen and grip. Because the particles are hydrophobic, the water doesn't cause the same adhesion problem. The gripper can release cleanly. Luna: But granular jamming has its own issues — it's slower to actuate, and the particles can get contaminated over time. Still, for a laboratory or medical setting where you need to handle wet drapes or tissues, it might be the best bet. Lucas: Exactly. And that's where the practical application lives. It's not about building a robot that can wipe your kitchen counter — though that would be nice. It's about automating tasks like handling wet biopsy samples in a hospital lab, or sorting wet laundry in an industrial setting. Those are high-value, repetitive tasks where a robot that can reliably grip a wet towel would pay for itself quickly. Luna: There's a company in Japan — I think it's called Mujin — that's working on a robot for sorting wet hospital linens. They use a combination of a soft gripper and a vision system that predicts tear points. Lucas: That's smart. Predict where the towel is weakest and apply less force there. It's a control problem as much as a hardware problem. The MIT paper I mentioned also found that the tear point is highly predictable if you have a good model of how the water distributes. They built a neural network that could predict the failure force within 5 percent just from the visual pattern of wet spots. Luna: So the solution might not be a better gripper, but a better brain — a control system that understands the material in real time. Lucas: That's the direction a lot of labs are heading. Instead of trying to make the gripper handle everything, you give the robot a richer sensory understanding. Tactile sensors that measure shear force and moisture content. Vision systems that track the towel's deformation at 200 frames per second. And then a controller that adjusts grip force and pull speed on the fly, just like a human does. Luna: It's a classic robotics trade-off: brute force versus intelligence. And for deformable objects, intelligence usually wins. Lucas: Right. And to be fair, we're making progress. The state of the art five years ago couldn't handle a dry paper towel reliably. Now we're at the point where research labs can manage wet towels about 70 percent of the time in controlled settings. That's not production-ready, but it's a proof of concept. Luna: Seventy percent — that means three out of ten attempts still fail. That's a lot of shredded towels. Lucas: It is. But compare that to something like folding a fitted sheet, which we talked about in episode 79. That's still below 20 percent success in the lab. Wet towels are actually easier than sheets because they're smaller and less irregular. Luna: So the hierarchy of difficulty goes: rigid objects first, then dry deformable, then wet deformable, then things like sheets that have multiple layers and odd geometry. Lucas: Exactly. And at the very top, probably, is something like untangling a knot of wet string. That's the nightmare scenario. Luna: Let's not go there. But I'm curious — how close are we to a commercial robot that can handle wet towels in a real hospital laundry? Five years? Ten? Lucas: I'd say five to seven years for a niche application. The key is to constrain the environment. If you know the towel size, the water content, and you can control the pull speed, you can engineer a solution. The general-purpose robot that walks into a random kitchen and wipes the counter — that's more like fifteen to twenty years, if ever. Luna: And that's okay. The value is in the constrained applications first. Hospital linens, laboratory sample handling, maybe even food processing — handling wet lettuce leaves or something. Lucas: Right. And that's where the economics start to make sense. If a robot can replace one human worker in a hospital laundry, that's a $30,000 to $40,000 annual saving. The robot might cost $100,000, so payback in under three years. That's viable. Luna: Which brings us back to the central challenge: can the robot do it reliably enough that you don't need a human standing by to fix the shredded towels? Lucas: Exactly. And that's the metric that matters. Not whether a lab demo works in a pristine environment, but whether it works for 8 hours straight in a humid laundry room with towels that are never exactly the same. Luna: So the next time you pull a wet paper towel off the roll in your kitchen and it tears, you can blame the capillary adhesion. But also appreciate that you just did something a $200,000 research robot still struggles with. Lucas: That's a humbling thought. And it keeps a lot of roboticists employed. Luna: Speaking of keeping things going — if you've enjoyed diving into the nitty-gritty of robot manipulation with us, a couple of dollars a month genuinely makes a difference. It's what keeps these episodes free and without ads. Lucas: Yeah, listener support is the only reason we can spend time on topics like wet towel adhesion instead of chasing clickbait. If you've gotten something out of the show, buy me a coffee dot com slash fexingo. Luna: All right. Back to the science. One thing I wanted to ask — are there any biological inspirations for handling wet, slippery objects? Lucas: Great question. Octopus suckers are a classic example. They can grip wet, slimy surfaces because the suction is active — they can control the pressure in each sucker independently. Some labs are working on bio-inspired suction grippers that mimic that. But they're noisy and need a vacuum pump, so not great for a quiet environment. Luna: There's also the gecko-inspired adhesive, but that relies on van der Waals forces, which are disrupted by water. So gecko tape doesn't work wet. Lucas: Right. So we're back to hydrophobic coatings and smart control. It's a tough problem, but that's why it's interesting. Luna: Alright, I think we've covered wet towels enough for one episode. Next time, maybe we'll tackle something drier. Lucas: Or maybe wetter. We haven't talked about why robot arms can't handle a sponge full of water yet. Luna: That's episode 85, I guess.