Latest / The Robotics Podcast with Fexingo: Autonomous Systems, Industrial Robots, and Hardware / The Robot Arm That Finally Drives a Screw Straight
Transcript
- Lucas: So there's this thing that happens on every assembly line in the world, thousands of times a day: a robot arm drives a screw into a hole. And it's supposed to be simple. The arm positions the screw, spins it, done. But anyone who's actually watched one knows it's not simple at all. The screw drops, the threads strip, the bit slips. And for decades, the fix has been brute force: bigger arms, stiffer frames, tighter tolerances. Luna: And yet, even with all that, you still see those little plastic bins of rejected screws next to every station, right? Lucas: Exactly. And that's the puzzle. Why can a robot, which can move a thousand-pound payload with millimeter precision, still mess up something as basic as putting a screw into a hole? It's because screwing isn't about positioning, it's about feel. And feel is exactly what robots have lacked. Luna: So what's changed? I read something about a startup in Germany that claims they've cracked it with a flexible wrist. Is that real or just hype? Lucas: It's real, and it's a company called FlexiDrive. They're not a household name, but they've been working on this for six years. And their approach is almost counterintuitive. Instead of making the arm stiffer, they add a deliberately compliant wrist. Think of it like a human wrist: it's not rigid, it gives a little, and that compliance lets you feel the screw engage the threads. Luna: So the robot is literally learning to be a bit floppy? That seems backwards. Lucas: It seems backwards, but it's actually how every skilled human does it. When you screw in a lightbulb, you don't just jam it in. You gently twist, feel the threads catch, then apply more torque. The flexibility in your wrist gives you sensory feedback. FlexiDrive has built that into a robot joint. The wrist has a series of springs and sensors that measure deflection and force in real time, and the control algorithm uses that to adjust the angle and pressure as it drives. Luna: So instead of programming the exact path, the robot is essentially feeling its way in. How much of a difference does that actually make on the line? Lucas: In their demo, which they run continuously at their lab in Munich, they drive two thousand screws in a row with zero failures. That includes screws going into plastic, aluminum, and even wood with varying density. And they're not cherry-picking perfect materials. They deliberately warp some of the test pieces, off-angle holes, slight burrs, the kind of real-world chaos that makes traditional robots strip a thread. Luna: Two thousand in a row, no failures. That's got to be a game-changer for someone like an automotive plant where any stoppage costs thousands per minute. Lucas: Automotive is exactly where this starts to matter. A typical car has about three thousand threaded fasteners. And on a traditional line, you might have a rejection rate of one to two percent, which sounds low, but when you're doing fifty cars an hour, that's a screw job every minute. With FlexiDrive's wrist, they claim rejection rates are down to nearly zero. Luna: But wait, isn't the automotive industry already using robots for this? They've been doing it for decades. Why haven't they solved this already? Lucas: They've solved it by spending a fortune on precision. The traditional approach is to use a rigid arm, a very accurate vision system, and a fixture that holds the part in exactly the right place. That works, but it costs. You need custom tooling, high-end cameras, and constant recalibration. And it breaks down when the part isn't perfect. So the big automakers have accepted the trade-off: invest heavily in the environment to make the robot's job easier. Luna: So FlexiDrive is saying: let's make the robot adapt to the part, not the other way around. That would be huge for smaller manufacturers who can't afford those massive setups. Lucas: And that's really the exciting part. The cost of a FlexiDrive wrist is around forty thousand dollars, which is a fraction of what a traditional precision cell costs. And they're not just a component. They're selling an entire screwdriving unit that can be bolted onto any existing industrial arm. So a mid-sized factory that already has a robot can upgrade without tearing out the whole line. Luna: Forty thousand dollars is still a chunk of change. How quickly does that pay off for a typical manufacturer? Lucas: They've done the math with their early customers. One customer, a maker of electric motors, was seeing about five percent of screws being driven incorrectly, requiring manual rework. They calculated that rework was costing them around twenty thousand dollars a month. With the FlexiDrive unit, they eliminated that rework entirely, so the payback period was about two months. Luna: Two months. That's a no-brainer for any plant manager. So why isn't every robot arm manufacturer already doing this? Lucas: Because it's hard. The control software is the real secret. It's not just a springed wrist. You need to process force and torque data at hundreds of hertz and make micro-adjustments in real time. That's a control engineering problem, not a hardware problem. And the big robot makers like KUKA and ABB have been focused on raw speed and payload, not this kind of finesse. Luna: So there's a moat there. FlexiDrive isn't just selling a sensor, they're selling years of algorithms. Lucas: Exactly. And they've filed a bunch of patents on the control method. So even if a big player wanted to copy them, they'd have to work around that. But the bigger picture is that this is part of a broader shift in robotics: moving away from rigid, pre-programmed automation toward systems that can handle variability. That's what will unlock the next wave of automation in industries like electronics assembly, where parts are delicate and tolerances are tight. Luna: Speaking of electronics, I'd imagine that's a perfect use case. Circuit boards are full of tiny screws that are easy to strip. Lucas: And that's actually where FlexiDrive is seeing the most interest. Their next big pilot is with a laptop manufacturer in Taiwan. The challenge there is that the screws are so small, they can barely be seen, and the torque limits are tiny. If you apply too much, you crack the plastic standoffs. Too little, and the screw vibrates loose. The compliant wrist can feel that sweet spot every single time. Luna: It's almost like the robot is developing a sense of touch. And that's something we've been exploring on this show, how tactile feedback is the next frontier. Lucas: We have, and this is a perfect example. But there's a question I keep coming back to: is this just an incremental improvement, or is it a fundamental shift? Because on one hand, it's still just a screwdriving tool. On the other, it's proof that robots can be taught to feel, not just to see. Luna: I think it's more than incremental. It's the difference between a robot that needs everything to be perfect and a robot that can work in the messy, imperfect real world. And that's a big deal. Lucas: That's a good way to put it. And it's why I think this technology could spread beyond screwing. If you have a wrist that can sense and adapt, you could apply it to other tasks like inserting connectors, pressing bearings, even polishing. The underlying principle is the same: compliant control with real-time feedback. Luna: So what's the catch? There's always a catch. Is it just speed? Is it slower than a traditional rigid arm? Lucas: It's a bit slower, maybe ten percent, because the wrist is constantly making micro-adjustments. But in most operations, the cycle time is dominated by positioning, not the actual driving. And when you factor in the reduced rework and downtime, the effective throughput is actually higher. The math works out. Luna: And that's the kind of thing that gets a CFO's attention. It's not just a technical toy; it's a real business case. Lucas: Right. And speaking of the business case, if you found this episode useful, you might be wondering how we keep making these deep dives without a paywall or ads. The answer is listener support. We stay ad-free thanks to people who chip in a few bucks. If today's tech conversation gave you something usable, you can keep it going at buy me a coffee dot com slash fexingo. That's buy me a coffee dot com slash fexingo. No pressure, but it makes a real difference. Luna: Yeah, and if you're a regular listener, you know we always dig into the specifics. That kind of research takes time, and we appreciate every coffee. Lucas: Now, back to the technology. One thing I want to clarify is how FlexiDrive's wrist differs from the force-torque sensors that have been around for years. Those sensors measure force at the end of the arm, but they don't provide compliance. They're like a rigid arm with a strain gauge. FlexiDrive's wrist is spring-loaded, so it actually absorbs energy and returns it. That's what allows the gentle back and forth that mimics a human hand. Luna: So it's not just sensing, it's actuation. The wrist is doing something, not just measuring. Lucas: Exactly. And that's what makes it a novel piece of engineering. They've essentially built a variable-stiffness actuator, which is a hot topic in robotics research. But they've made it work in a package that's rugged enough for a factory floor. Luna: Okay, so what's next for FlexiDrive? Are they going to stay niche, or are they aiming for the big leagues? Lucas: They're already in talks with two of the top three industrial robot makers about integrating their wrist as an optional accessory. That would be a massive validation. And they've just closed a series B round of fifty million dollars to scale up production. So they're clearly aiming big. Luna: Fifty million in a funding round, that says a lot about investor confidence. There must be real demand. Lucas: And I think the demand is only going to grow as more manufacturers look to automate tasks that are currently done by hand. The cost of robotics has been dropping, but the bottleneck has always been the flexibility. Technologies like this are what unlock the next wave. Luna: So if I'm a small business owner listening to this, should I be thinking about upgrading my line soon, or is this still early adopter territory? Lucas: I'd say it's early adopter, but not by much. The payback is so fast that it's worth doing a pilot. FlexiDrive is offering a trial program where you can send them your actual parts and they'll run them on their system. That's a low-risk way to see if it works for your application. Luna: That's smart. Test before you invest. I like that. Lucas: And that's the kind of practical approach that's going to drive adoption. The technology is impressive, but the real proof is in the ROI. And for a lot of manufacturers, the math is going to be hard to ignore. Luna: Well, it sounds like the humble screw is finally meeting its match. I'm curious to see if this trickles down to consumer products like phones and appliances. Lucas: I think it will, especially as the cost comes down. And that's the exciting part: we're at the beginning of a shift where robots stop being rigid machines and start being adaptable tools. This episode, we saw one practical example, but the principle could change a lot of things. Luna: Thanks for breaking that down, Lucas. It's one of those topics where the details really matter. Lucas: Thanks, Luna. And if you want to see the demo for yourself, we'll put a link to FlexiDrive's site in the show notes. Until next time.