Latest / The Robotics Podcast with Fexingo: Autonomous Systems, Industrial Robots, and Hardware / Why Robot Arms Still Cant Handle Raw Chicken
Transcript
- Lucas: So I was reading about a new gripping system out of a lab in the Netherlands, and it's designed to pick up raw chicken fillets from a conveyor belt. And I thought, okay, another gripper demo. But then I saw the video, and it's genuinely remarkable. It's the kind of thing that makes you realize how much we take for granted about handling slippery, irregular objects. Luna: Raw chicken? That's oddly specific. What's so hard about picking up a piece of meat? Lucas: Everything, apparently. The skin is low-friction, the shape is never the same twice, the weight varies, and it's wet. For a robot arm, that's a perfect storm. I mean, we've done episodes on locks, bottle caps, hammers, shoelaces, wiring harnesses, buttons, lightbulbs, pancakes, scissors, mirror finishes, screws, potatoes, pens, needles, green beans, towels, eggs, cable threading, and screwdrivers. But raw chicken is almost like the boss level. Luna: The boss level of what? What makes it worse than, say, a slippery egg or a floppy towel? Lucas: With an egg, at least you have a predictable shape. A towel has some structure, even if it's flexible. Raw chicken is amorphous, slippery, and it deforms under pressure. If you grip it too hard, you squish it and the juices run out. If you grip it too lightly, it slides right through your fingers. And you have to do this thousands of times an hour. Luna: So what's the typical approach in a poultry plant right now? I'm guessing it's not robots. Lucas: You're right. It's mostly human hands. I read that in the U.S., the poultry processing industry employs over 250,000 workers, and a big chunk of that is doing tasks like this—picking up fillets, trimming fat, sorting pieces. It's repetitive, cold, and dirty work. And it's a huge labor cost. The average hourly wage in poultry processing is around fifteen dollars, and with turnover rates over forty percent, it's a constant headache for plant managers. Luna: So the incentive to automate is huge. Why hasn't it happened yet? Lucas: Because the problem is genuinely hard. The classic rigid gripper—two metal fingers that close like a pincer—just doesn't work. It'll either crush the meat or let it slip. Vacuum suction doesn't work because the skin is porous and wet. Even the best vision systems struggle to segment a fillet from a pile on a moving belt, because there are no clear edges. It's a mess. Luna: So what's the Dutch lab doing differently? Are they using some kind of soft robotic hand? Lucas: Exactly. The gripper is made of a soft silicone material with multiple fingers that can individually conform to the shape of the chicken. And they've added a sensor that measures the contact pressure in real time, so the grip adjusts dynamically. It's like a robot hand that can feel the difference between a firm grip and a crush. Luna: That sounds like something we talked about in the warehouse episode—the hand that picks anything. But how does that translate to something as slippery as this? Lucas: Great question. The warehouse hand relies on a combination of vision and a gripper that can conform to random objects. But the key difference here is the slipperiness. They've done extensive tests with chicken skin, and they found that the friction coefficient is incredibly low—like zero point two or even less. That means any sideways force causes the object to slide. So the gripper has to be extremely gentle on the first contact, then gradually increase pressure. Luna: So it's a delicate dance. How fast can it actually go? Lucas: In the lab, they're hitting about fifteen picks per minute. That's not blazing fast, but it's comparable to a slow human. And they're working on getting it faster. The real breakthrough is that they've proven it's possible. The gripper can pick up a raw fillet, place it on a scale, and do it without tearing the meat or dropping it. That's a huge step. Luna: Fifteen picks per minute doesn't sound like it's ready for a full production line. What's the realistic timeline for this to actually replace humans? Lucas: That's the million-dollar question. The lab's founder has said they're aiming for a commercial system within three years. But I'm skeptical. The challenge isn't just the gripper—it's the whole system. You need robust vision, the arm has to be hygienic and easy to clean, and you need to handle the variability of real-world conditions. Think about a bird that's been stunned, bled, plucked, and cut—every single one is different. Luna: Right, and in a food processing plant, there are strict sanitation rules. You can't just have a robot arm dripping with grease and bacteria. Lucas: Exactly. The Dutch lab is working on a washdown version that can withstand high-pressure cleaning. But that adds cost and complexity. I saw a quote from the founder saying, 'We're not trying to replace all workers, but we can take over the most repetitive, ergonomically punishing tasks.' That seems like the right framing. Luna: So what's the economic case? If it's fifteen picks per minute, and a human can do maybe twenty or thirty, how do the numbers work out? Lucas: That's where it gets interesting. A robot can work twenty-four hours a day, doesn't need breaks, doesn't get sick, and doesn't quit. If you look at total cost of ownership, a robot system might cost a hundred and fifty thousand dollars upfront, plus maintenance. But over a few years, it can actually be cheaper than a human worker when you factor in benefits, turnover, and training. A study from the Food and Agriculture Organization estimated that automation could reduce processing costs by twenty to thirty percent in the next decade. Luna: But you mentioned the turnover rate is forty percent. That's huge. So even if the robot is slightly slower, it might be more reliable. Lucas: Exactly. And that's a pattern we see across many industries. The reliability and consistency of a robot can outweigh raw speed. For example, in the bakery industry, robots have been used for years to move bread loaves, but it's only recently that they've gotten good at handling the deformable, sticky dough. Now they're taking over tasks that humans used to do, not because they're faster, but because they're more consistent. Luna: It's funny, we've spent over a hundred episodes on all the things robots can't do, and now we're finally seeing the tide turn. What's the next frontier after raw chicken? Lucas: Well, if they crack raw chicken, the next big test is probably handling raw fish. Think about sushi-grade salmon—it's even more delicate and slippery. And then there's the challenge of deboning, which requires a lot of dexterity. But let's not get ahead of ourselves. Even getting to the point where a robot can reliably pick up a chicken fillet is a big deal. Luna: I know we're in the middle of this, but I wanted to mention something that's been on my mind. This show exists because of the people who listen and support us. We don't have ads, and that's a conscious choice. But it means we rely on a small group of listeners who chip in through buy me a coffee dot com slash fexingo. It's not a lot of people, but it's enough to keep the lights on and keep us making deep dives like this. Lucas: Yeah, it really is listener supported. And it's one of those things that if you get value from a deep dive on something like why raw chicken is so hard for robots, that support is what makes it possible. No pressure, but it's nice to know the people who care are the ones who keep it going. Luna: Anyway, back to the chicken. You mentioned the sensor that measures pressure. How does that work exactly? Is it like a force-sensitive resistor? Lucas: It's more sophisticated. They're using something called capacitive sensing. The gripper has multiple layers of silicone with conductive patterns, and when the fingers touch an object, the capacitance changes. That gives them a pressure map across the contact area. So they can see exactly where the pressure is high and where it's low, and adjust the grip in real time. It's like having a sense of touch. Luna: So it's not just about the hardware, it's also about the software that interprets the sensor data. Lucas: Precisely. The lab uses machine learning to train the gripper on hundreds of pick attempts. They record the pressure maps and the outcomes, and the algorithm learns what a successful grip feels like. That's a big shift from traditional robot programming, where you hard-code the motion. Now it's more like teaching by demonstration. Luna: That's fascinating. So the robot is basically learning from experience, which is something we've seen more and more in other applications. Do you think we'll see this approach spread to other food items, like maybe avocados or tomatoes? Lucas: Absolutely. In fact, the same lab is already testing the gripper on avocados, which are notoriously delicate. And I've seen similar prototypes for handling soft fruits in packing facilities. The key is that the gripper adapts to the object, not the other way around. That's the fundamental shift. Luna: So what would it take for a poultry plant to actually install one of these? Is it just a matter of cost, or are there other barriers? Lucas: Cost is a big factor, but also integration. You need to retrofit the robot into an existing line, which can be disruptive. And you need skilled technicians to maintain it. There's also a cultural barrier—some plant managers are skeptical that a robot can keep up with a human crew. But I think as the technology improves and the cost comes down, we'll see more adopters. There are already companies like JBS and Tyson Foods investing in automation research. Luna: It's interesting, because we've talked about the labor shortage in many industries, and poultry processing is no exception. So the pressure to automate is only going to grow. Lucas: Exactly. And that's why I think this breakthrough is so timely. It's not just about replacing workers; it's about making the work safer and less physically demanding. The tasks that robots are taking over are the ones that cause repetitive strain injuries and high turnover. If we can automate those, we might actually improve job quality for the humans who stay. Luna: That's a nice way to put it. So, to sum up, raw chicken is a nightmare for robots because of slipperiness, irregularity, and deformation. But a Dutch lab has developed a soft gripper with capacitive sensing that can handle it, and it might be ready for commercial use in a few years. Is that about right? Lucas: That's the gist. And it's a perfect example of how robotics is slowly chipping away at the last mile of automation. It's not going to happen overnight, but every time we crack a problem like this, it opens the door to a dozen new applications. I'm excited to see what comes next. Luna: Me too. And I guess we'll have to do an episode in a few years to see if they've actually made it to the factory floor. Lucas: Absolutely. We'll put that on the calendar. For now, I think we've given everyone a new appreciation for the humble chicken fillet. Luna: And maybe a little more respect for the people who process it by hand every day. Lucas: Definitely. That's a good note to end on.