Latest / The Robotics Podcast with Fexingo: Autonomous Systems, Industrial Robots, and Hardware / How Collaborative Robots Are Reshaping Auto Assembly
Transcript
- Lucas: So, 45,000. That's the number of collaborative robots installed globally in automotive factories last year. Up 35 percent from 2024. And here's the thing that caught my attention: most of those cobots didn't replace a human worker. They replaced a cage. Luna: A cage? You mean the safety fencing around traditional industrial robots? Lucas: Exactly. Traditional robot cells are separated from people by physical barriers or laser curtains. You walk past them on a factory floor and they're behind these massive yellow cages because if that arm hits you, it will kill you. Cobots are designed to stop or slow down if they contact a person. So they can sit right next to a worker on an assembly line. No cage. Luna: Right, but I've heard that cobots are slower and less precise. How much are you giving up in speed or accuracy to go cage-free? Lucas: That's the trade-off question everyone asks. Let's take a concrete example. There's a tier-one automotive supplier in Ohio—they make door panels for a major U.S. automaker. Last year they put in a Universal Robots UR20 on a spot-welding station. Before that, a human worker would weld four spots per panel, cycle time about 30 seconds per panel, and they had two workers per shift. After the cobot, one worker supervises two UR20s. Each cobot cycles in about 28 seconds. So throughput stayed the same. Luna: Wait—a cobot spot welding? I thought welding was too hot and too high-force for cobots. Aren't they usually limited to pick and place or screwdriving? Lucas: Historically yes. But the payload and reach have improved. The UR20 can lift 20 kilograms, reach 1.75 meters, and it has this built-in force sensing that lets it detect if a human nudges it. In the Ohio case, they actually reduced the injury rate for that station. The human workers were getting repetitive strain from holding the welding gun. Now the cobot holds the gun, the worker guides the panel. It's a true collaboration. Luna: So the cobot is doing the heavy, repetitive motion while the human handles the part variability. That makes sense. But what about precision? I remember a story from a German engine plant—they tried a cobot for cylinder boring, and it just couldn't hold the tolerance. They had to go back to a traditional CNC machine. Lucas: You're right. Cobots generally have a repeatability of about plus or minus 0.1 millimeters. For cylinder boring, you need micron-level—like five to ten micrometers. That's a factor of ten to twenty times tighter. So cobots are not replacing high-precision machine tools. But for tasks like assembly, adhesive dispensing, inspection, light welding—they're competitive. Luna: And the cost must be a factor too. Traditional industrial robots with the cage, the controller, the installation—what are we talking, six figures? Lucas: Easily. A Fanuc or ABB robot arm might cost $50,000 to $80,000 for the arm alone, but then you add the safety system, the fencing, the interlocked doors, the light curtains—total installation can hit $150,000 or more. A cobot like the UR20 lists around $45,000. No cage, no extra safety hardware. And the programming is dramatically simpler. You can hand-guide the arm to teach it a path. No coding required. Luna: Hand-guiding? That's where you physically grab the arm and move it through the motions, and it records them? Lucas: Exactly. You set the gripper position, press a button, move the arm to the next point, press again. That's it. The Ohio supplier told me their changeover time for a new door panel variant went from forty minutes to twenty-five minutes. That's a nearly 40 percent reduction. And because the cobots are mobile—they're on a wheeled cart—they can literally roll them to a different line and retrain them in an afternoon. Luna: That flexibility is huge for automotive with all the model variants. But doesn't that also mean cobots are eating into the traditional robot market? I saw that Fanuc and ABB now both sell cobots too. Lucas: They do. Fanuc's CRX series, ABB's GoFa—they're all collaborative. The market is bifurcating. You have high-speed, high-precision traditional robots for tasks like painting or heavy lifting where speed and repeatability are paramount. And then you have cobots for tasks that require human dexterity or frequent changeovers. The International Federation of Robotics says cobot installations grew 30 percent globally in 2025, while traditional robot sales grew only 5 percent. Luna: So the growth is really in the collaborative segment. But I'm curious—are there any tasks where cobots have failed outright? Beyond the precision issue. Lucas: Yes, there's a great example from a logistics company in the UK. They tried cobots for palletizing mixed-size boxes—think Amazon warehouse. The cobot struggled because it couldn't adapt to the varying box shapes and weights quickly enough. The gripper was too rigid. They ended up going with a hybrid solution: a traditional robot with a vision system and a custom vacuum gripper. So cobots are not a silver bullet. Luna: That makes sense. The gripper technology is still a bottleneck. There's a reason our earlier episode on robot hands talked about picking up an apple. Lucas: Exactly. But for the tasks they are good at, the economics are compelling. A typical payback period for a cobot in automotive assembly is 12 to 18 months. Compare that to a traditional robot cell which might take two to three years. That shorter payback is driving adoption, especially among small and midsize suppliers who couldn't justify a full robotic cell before. Luna: So cobots are democratizing automation. What about safety? How does a cobot actually 'know' it's touching a human? Lucas: Great question. There are two main approaches. One is torque sensing in each joint. The robot monitors the current draw of the motors. If the torque exceeds a threshold—say, you bump into the arm—it registers an unexpected force and stops within milliseconds. The other approach is skin sensors—capacitive or pressure-sensitive coverings that detect contact. Some cobots combine both. The ISO standard for cobots, ISO 10218, actually specifies that power and force must be limited to avoid injury. Luna: So if I walk into a cobot, it might leave a bruise but not break a bone? Lucas: That's the theory. In practice, it depends on the speed and the tooling. If the cobot is holding a sharp cutting tool, even low force can be dangerous. So when cobots are used with hazardous end-effectors, they often still need a light curtain or a speed limit. Safety is not automatic. Luna: Good to keep that nuance. So where do you see cobots going in the next few years? Are we going to see them in every factory? Lucas: I think we'll see them in more factories, but not every factory. The biggest barrier right now is not technology—it's organizational. Many plant managers don't know how to integrate a cobot into their workflow. They need to retrain workers, redesign workstations, and rethink quality control. But the trend is clear. By 2027, I expect sensor fusion—combining vision, force, and proximity sensors—will close the precision gap. Cobots might start doing tasks like engine assembly or even light machining within 50-micron tolerances. Luna: So the cage is disappearing, but the human is still in the loop. That's a different vision of automation than the fully lights-out factory. Lucas: Exactly. And it's probably the more realistic one for the next decade. The term 'collaborative robot' might be mostly marketing—some engineers argue that true collaboration requires the robot to anticipate human intent, not just react to contact. But even as a marketing term, it's pointing to a real shift: robots are coming out of the cages and working alongside us. The question is how we design that relationship. Luna: And whether we call it collaboration or just safer coexistence, the numbers suggest it's working. Forty-five thousand units in automotive alone last year—that's a vote of confidence. Lucas: It is. And the interesting thing is that the next wave might not be in automotive at all. Food processing, pharmaceuticals, even construction are starting to adopt cobots. But that's a conversation for another episode. Luna: Agreed. For now, I think we've given listeners one concrete number to remember: 45,000. And a new perspective on what a factory floor can look like. Lucas: No cages, just collaboration. Or at least, that's the goal.