Latest / The Robotics Podcast with Fexingo: Autonomous Systems, Industrial Robots, and Hardware / Why Robot Arms Still Cant Handle a Wet Bar of Soap
Transcript
- Lucas: So, here's a question that might sound like a joke but is genuinely stumping some of the best robotics labs in the world: why can't a robot arm pick up a wet bar of soap? Luna: Wait, is that actually a problem they're working on? I figured after a hundred episodes of things robots can't do, soap would be low on the list. Lucas: It's surprisingly high on the list. And it's not just a shower curiosity — understanding slipperiness is critical for everything from surgical robots handling tissues to home-care robots helping someone bathe. A wet bar of soap has basically zero friction when you try to grab it. Luna: Right, because our fingers have oil and texture, plus we can adjust grip force in real time. What do robots have? Lucas: Most industrial grippers rely on high-friction materials like rubber or silicone, and they use vision to plan a grasp. But the moment the soap is wet, the coefficient of friction drops dramatically — sometimes by a factor of ten or more. The gripper applies its standard force, and the soap just squeezes out. Lucas: Researchers at MIT's CSAIL recently ran a study where they tried to get a KUKA arm with a standard parallel-jaw gripper to pick up a bar of soap from a wet dish. Out of a hundred attempts, it succeeded exactly zero times. Even with tactile sensors on the fingertips. Luna: Zero? That's brutal. And tactile sensors are supposed to give feedback about pressure distribution — what went wrong? Lucas: The sensors detect contact, but they can't predict slip before it happens. The soap has a changing property: as you squeeze, water gets displaced, so the friction changes second by second. The control loop can't keep up. Humans instinctively micro-adjust grip, but robots are still playing catch-up. Luna: So it's not just about better sensors — it's about a fundamentally different control algorithm that anticipates slip. Lucas: Exactly. Some teams are working on what they call 'slip anticipation' using high-speed cameras and force torque sensors. The idea is to detect the moment when the object starts to move relative to the gripper, and react in milliseconds. But water complicates that because it acts as a lubricant, so the slip onset is much faster. Lucas: One approach from the University of Tokyo involves using a gripper with micro-spikes that dig into the soap, but that leaves marks and isn't practical for delicate tasks. Another uses electrostatic adhesion — you apply a voltage to create an attractive force. That works on dry surfaces but fails when there's a water layer. Luna: So basically, the humble bar of soap exposes a fundamental gap in robotics: handling objects with variable surface properties. And this isn't just a lab curiosity — it has real-world implications. Lucas: Absolutely. Think about surgical robots: they need to grasp slippery tissues like the liver or intestines without damaging them. Or think about a robot designed to help an elderly person bathe — it has to handle a wet washcloth, a slippery arm, and yes, a bar of soap. If it can't manage soap, it can't bathe someone safely. Luna: And in industrial settings, there are plenty of slippery parts — oily engine components, wet food items, frozen goods. So solving the soap problem is actually a multi billion dollar challenge. Lucas: Right. There's a company called RightHand Robotics that makes picking systems for warehouse fulfillment. They handle millions of items a day, but they explicitly avoid anything wet or slippery unless it's in a rigid container. So the soap problem is a known limiter for e-commerce automation as well. Lucas: Let's get into the physics a bit more. The key variable is the coefficient of friction, or COF. Dry soap on dry rubber has a COF around 0.8 — pretty grippy. Wet soap on wet rubber drops to about 0.1. That's a reduction of 87 percent. The robot's control system is tuned for the higher number, so when it encounters the lower one, it either squeezes too hard and ejects the soap, or doesn't squeeze hard enough and drops it. Luna: And because the COF changes dynamically as water is squeezed out, the robot can't just learn a single 'soap force.' It needs to adapt continuously. Lucas: Precisely. Some researchers are using reinforcement learning in simulation, where the robot tries millions of grasps on virtual soap bars with randomized wetness. But transferring that to the real world is tough — the simulation has to be incredibly accurate about fluid dynamics, which is computationally expensive. Luna: I've heard about soft robotics being a possible solution. Soft grippers that conform to the object's shape might handle slipperiness better. Lucas: They help, but they're not a silver bullet. Soft grippers, like the ones from Soft Robotics Inc., use pneumatic chambers that inflate around an object. They can conform to soap, but the water still creates a lubricating layer. The gripper can envelop the soap, but it may still slip out if the grip isn't actively adjusted. Lucas: There's also work on biomimetic adhesives, like gecko-inspired dry adhesives that use van der Waals forces. Those work on smooth surfaces but fail when there's a water film, because the water prevents close contact between the adhesive and the object. Luna: So it seems like every promising technology hits a wall when water is added. What's the most creative solution you've seen? Lucas: Maybe the most unusual is from a group at Carnegie Mellon — they built a gripper that uses a tiny ultrasonic vibrator to create a thin air layer between the gripper and the soap, a bit like how a hovercraft works. The air layer reduces friction, but it also prevents the water from creating suction. They got a 60 percent success rate in lab tests, which is the best I've seen for wet soap. Luna: Sixty percent is still not reliable enough for a production robot, but it's progress. And it shows that sometimes you need to think outside the box — literally using air to solve a water problem. Lucas: Exactly. And that kind of thinking is what we try to highlight on this show. Speaking of which, if today's tech conversation gave you something usable — maybe a new way to think about friction or an insight into why your robot vacuum still struggles — we do have a small way you can support the show, and it's completely optional. Luna: Yeah, and it's really low-key. If today felt worth the price of a coffee to you, you can find us at buy me a coffee dot com slash fexingo. No pressure, no perks, just knowing that listener support keeps this podcast ad-free and focused on the stuff that matters. Lucas: Right. And we'll keep digging into the weird and wonderful challenges in robotics. So, back to soap — what's next? The ultrasonic gripper is promising, but scaling it is hard. The vibration mechanism is bulky and noisy. So the search continues. Luna: Do you think we'll see a robot that can reliably pick up a wet bar of soap within the next five years? Lucas: I think we'll see specialized robots that can do it in controlled environments, like a surgical robot handling slippery tissue, within two or three years. But a general-purpose robot that can walk into any bathroom and grab any bar of soap? That's probably a decade out. The problem touches on perception, control, materials science, and fluid dynamics all at once. Luna: So it's one of those 'simple' tasks that actually requires solving multiple hard problems simultaneously. Lucas: Exactly. And that's what makes it fascinating. It's not just about soap — it's about building robots that can handle the messy, variable, slippery real world. And until we crack that, robots will remain in factories and labs, not in our homes helping with the dishes or the bath. Luna: Well, here's hoping the next breakthrough comes from an unexpected place, like a grad student who drops the soap one too many times in the lab shower. Lucas: I wouldn't be surprised. Some of the best robotics ideas start with a human failure. Thanks for listening, and we'll be back next week with another thing robots still can't do.