Latest / The Robotics Podcast with Fexingo: Autonomous Systems, Industrial Robots, and Hardware / Why Robot Factory Moves Are So Painfully Slow
Transcript
- Lucas: So here's a number that stuck with me: 47 days. That's how long ZF Friedrichshafen, the German automotive supplier, needed to move 112 robots from an old plant in Schweinfurt to a new facility twelve miles down the road. Luna: Wait — twelve miles and it took a month and a half? For something that's essentially unplug, truck, plug in again? Lucas: That's exactly the reaction most people have. And it gets to the heart of a really underappreciated problem in industrial robotics: deploying or moving a robot isn't like moving a CNC machine or a conveyor belt. It's more like transplanting a nervous system. Luna: Okay, so what's the bottleneck? Is it the physical rigging, the wiring, the safety certification — what specifically chewed up those 47 days? Lucas: All of the above, but the killer is software calibration. Each robot has what's called a kinematic model — a mathematical description of every joint, link, and axis. That model lives inside the controller. When you unplug the robot and move it, you don't get to keep that model. You have to rebuild it on site because even a millimeter shift in the mounting plate changes the relationship between the robot's base and the world around it. Luna: So every bolt hole, every floor anchor, every conveyor interface has to be re-measured and re-programmed. That sounds like the kind of thing someone would have automated by now. Lucas: You'd think. And there are tools — laser trackers, photogrammetry systems, even collaborative measurement arms. But the problem is that most factory robots run on proprietary control platforms. KUKA's controller doesn't speak the same language as Fanuc's or ABB's, and even between generations from the same vendor, the settings aren't forward-compatible. ZF's Schweinfurt plant had a mix of KUKA KR 210s and some older KR 150s. The new facility was supposed to standardize on the KR 220 series. That alone added about two weeks of conversion work. Luna: Let me guess — they had to manually re-teach every single program point because the reach and dynamics of the new arm are different. Lucas: Exactly. And it's not just the motion paths. You also have to redefine the tool center point — the TCP — which is the exact point in space where the robot's end effector does its work. If the TCP is off by half a millimeter on a welding robot, you get cold welds that fail quality inspection. ZF's quality team ran 3,000 test welds over three days just to verify the new TCPs on the first twenty robots. Luna: So the 47 days isn't just about moving hardware. It's about rebuilding the digital brain for every single unit. That must cost a fortune in downtime. Lucas: It does. ZF estimated that the Schweinfurt move cost them roughly 1.2 million euros in lost production — and that's conservative, because they scheduled it over a plant shutdown period. If they had tried to do it while the old line was still running, the number would be multiples higher. This is actually a huge competitive issue for manufacturing in high-wage countries. The more expensive it is to reconfigure a plant, the less willing companies are to adapt to changing product demand. Luna: Right, because if it takes six weeks and a million-plus euros to move a line, you're not going to do it on a whim. You lock in a production plan and stick with it, which is the opposite of the flexible manufacturing everyone talks about. Lucas: Precisely. And that rigidity is what a startup called FlexMove is trying to solve. They've built a middleware layer that sits between the robot controller and the factory execution system. It captures the kinematic model, the TCP data, the safety zones, and the program logic in a vendor-agnostic format. When you move a robot, you plug it into FlexMove's calibration box, and it automatically adjusts the model to the new floor position using a set of fiducial markers you place on the mounting plate. Luna: I've heard of them. They're based out of Munich, I think? Early stage, maybe thirty people? Lucas: Yeah, thirty-two as of last month. They closed a 4.5 million euro seed round in February. Their demo at Automatica in 2025 showed a KUKA robot that they moved across the show floor and re-calibrated in under four hours — including a full re-validation of ten pick and place programs. That's versus the typical two-to-three-day re-teach for a single program. Luna: Four hours versus three days is the kind of math that gets a plant manager's attention. But adoption has to be slow, right? Because you can't retrofit every legacy controller easily. Lucas: That's the challenge. FlexMove currently supports KUKA, Fanuc, and ABB — which covers maybe 60 percent of the installed base. But the older controllers, like the Fanuc R-30iA or the KUKA KR C2, don't have the API hooks to allow external calibration injection. So for those, you still need the manual laser-tracker process. Still, even if FlexMove only helps with the newer generation, it could cut the average move time by maybe 40 percent. Luna: If it cuts even 20 percent, that's millions in savings across an industry that moves hundreds of robots a year. I wonder how many of those moves are actually happening. Like, is this a growth market or a niche? Lucas: It's surprisingly large. A report from the International Federation of Robotics — the IFR — estimates that roughly 8 percent of the global installed robot fleet, about 300,000 units, is relocated within a plant or between plants each year. That's a lot of calibration work. And the trend is toward more moves, because product lifecycles are shortening and manufacturers want to share robots across lines. So the pain is real and growing. Luna: I'm also thinking about the safety re-certification process. When you move a robot, you have to re-validate the safety-rated speed and force limits, the emergency stop circuits, the zone guarding. That's a separate track that software calibration doesn't touch. Lucas: Great point. Safety re-certification added about ten days to ZF's 47-day timeline. Each robot had to pass a static torque test, a dynamic stopping-distance test, and a full risk assessment per ISO 10218. That's not something FlexMove's middleware can speed up, because it's physical verification. But some companies are starting to use simulation-based safety validation, where you model the robot and its environment in a digital twin and run the safety tests virtually. That can cut the physical testing time by half, but it's still early. Luna: So the broader picture is: moving a robot is a systems integration problem that involves hardware, software, and certification, and the industry is only beginning to build the tools to make it fast. It's almost like the robotics industry forgot that robots themselves need to be mobile. Lucas: That's exactly it. The robots are designed to sit on a concrete pad for ten years. The industry built really good stationary tools, and now the market is demanding that those tools become reconfigurable. I think we'll see more startups like FlexMove, and maybe the big vendors will eventually bake this into their own controllers. But for now, it's a messy, expensive, and largely hidden cost of automation. Luna: Honestly, this conversation gave me a new appreciation for how much invisible work happens under a factory floor. I mean, we look at a robot and think 'oh, it's just a machine', but there's this whole digital scaffolding around it. Lucas: Yeah. And I think that's worth a coffee. Luna: Oh, absolutely. If today's tech conversation gave you something usable — maybe a new way to think about factory flexibility, or just a good number to cite — that's the link. It's buy me a coffee dot com slash fexingo. Keeps this show ad-free and independent. Lucas: Lowest bar in the world. A buck, five bucks, whatever the episode was worth to you. No pressure, no perks, just support for a show you like. And we get right back to the tech. Luna: So speaking of hidden costs — let's talk about the actual floor preparation. ZF's new facility in Schweinfurt required reinforced concrete pads for the heavier KR 220s, and the floor flatness tolerance was something like 0.2 millimeters per meter. That alone took two weeks of civil engineering before the first robot even arrived. Lucas: Right. And that's the kind of thing FlexMove can't fix. The robot's base has to be physically level within tight specs, or the kinematic model will be wrong regardless of software. So the 47 days also includes the time to pour and cure the concrete, install the anchor bolts, and let everything settle. That's just physics. Luna: So the realistic timeline for a fast move — best case, with FlexMove and pre-validated safety simulations — what does that look like? Lucas: I think you could get it down to about 12 to 15 days for a similar scale: 100-plus robots. That's assuming the floor is ready, the controllers are from the supported generation, and you have a digital twin of the new layout already built. That's a huge improvement, but it's still two weeks of downtime. The goal for the industry is probably to get under a week, but that's going to require standardization that nobody has really achieved yet. Luna: And standardization would mean the robot vendors agreeing on a common calibration format, which is basically asking competitors to cooperate on something that could be a differentiator. That's a tough sell. Lucas: It is. But the customers — the big automotive OEMs, the electronics manufacturers — they are starting to push for it. I know that BMW and Volkswagen have both floated the idea of a 'plug and play robot' standard in their supplier requirements. If they make it a condition of purchase, the vendors will find a way. Luna: And that could change the entire economics of factory reconfiguration. If you can move a robot in three days instead of three weeks, you can start thinking about seasonal production lines, short-run customization, maybe even robots that follow products through the factory instead of staying in one cell. Lucas: Exactly. That's the vision. But we're not there yet. Right now, we're still in the world where a twelve-mile move takes 47 days. And that's a useful reality check for anyone who thinks industrial robots are just big, fast toys. They are precise, powerful, and deeply fragile in their relationship to the physical building they sit in. Luna: Fragile is a great word for it. One misplaced bolt, one bad floor pour, and the robot might as well be on the moon. I love that this episode drilled into the concrete and bolts side of robotics instead of the AI hype. Lucas: That's the stuff that actually matters on a factory floor. Thanks for digging into it with me.