Latest / The Robotics Podcast with Fexingo: Autonomous Systems, Industrial Robots, and Hardware / Why Robot Safety Is Killing Innovation
Transcript
- Lucas: If you walk into a modern automotive assembly line today, the first thing you notice isn't the cars—it's the silence. The heavy clanking of hydraulic presses is gone, replaced by the soft whir of electric servos. But there is another sound that is becoming increasingly loud, and it is the sound of frustration from plant managers trying to integrate new robotic systems. Luna: And that frustration usually stems from the same old problem: getting humans and machines to share the same space without triggering emergency stops every five minutes. Lucas: Exactly. For decades, the solution was simple: put a cage around the robot. If a human entered the zone, the machine stopped. It was binary, safe, and incredibly inefficient for anything beyond mass production. Now, with companies like Fanuc and Universal Robots pushing hard on collaborative or cobot technology, we have a new dilemma. We want the flexibility of a human working alongside a machine, but we are still bound by rigid safety frameworks designed for iron cages. Luna: It feels like we are trying to fit square pegs into round holes, especially when you consider how much downtime these false triggers cause. Lucas: That is the core tension I want to unpack today. We are going to look at why robot safety, specifically the rules governing power and force limiting, is becoming the single biggest brake on innovation in hardware. It is not just about avoiding lawsuits; it is about the physical limitations placed on movement that make these advanced robots behave more like cautious toddlers than precision instruments. Luna: So, we are talking about the trade-off between safety compliance and actual operational speed. Lucas: Precisely. And before we get too deep into the engineering weeds, I want to mention something quick. We deliberately do not run ads on these shows because we believe the conversation should stand on its own merit. If you want to support that choice directly, the link is buy me a coffee dot com slash fexingo — that's F. E. X. I. N. G. O. Luna: Yeah, we appreciate the support. It keeps us focused on the substance rather than chasing click-through rates. Lucas: Back to the safety issue. The standard everyone references is ISO 10218, along with the technical specification ISO/TS 15066. These documents define exactly how much force a robot arm can exert before it must stop to prevent injury. The numbers are specific: usually around one hundred and thirty-five newtons for the shoulder joint, and less for the elbow and wrist. Luna: Those numbers seem incredibly low when you think about how much torque a standard industrial robot can generate. Lucas: They are. A typical six-axis industrial robot might have a payload capacity of twenty kilograms, meaning its motors are designed to handle significantly more force than what is allowed in a collaborative setting. To meet those ISO limits, manufacturers have to add massive amounts of software overhead. They have to monitor joint currents millisecond by millisecond to detect any unexpected resistance. Luna: Which means the robot is constantly calculating if it is about to hit someone, even if it hasn't made contact yet. Lucas: Right. It is predictive stopping based on sensor data. The problem is that this sensitivity creates a lot of noise. In a busy factory environment, vibrations from nearby machinery, the impact of a part being dropped, or even the robot hitting its own limit switch can register as an anomaly. The system thinks it has detected a collision and slams the brakes. Luna: And then the whole line has to be reset manually, which kills the efficiency gain the company was hoping for. Lucas: That is the irony. The very technology meant to increase flexibility is often slower and more prone to interruption than the older, caged systems. We looked at data from a mid-sized electronics manufacturer who switched to cobots for their final assembly stage. They expected a twenty percent boost in throughput. Instead, they saw a ten percent drop in uptime during the first quarter due to unnecessary stops. Luna: That is a huge gap. So the safety features are actually working against the business case for adoption. Lucas: It suggests that our definition of safety is outdated. We are treating every potential interaction as a risk, rather than assessing the context of the work. When a human is handing a delicate circuit board to a robot, the required force is minimal. When a human is guiding a heavy chassis into place, the forces involved are higher. Current standards apply a blanket restriction regardless of the task complexity. Luna: So we need smarter sensors that understand intent, not just raw impact force. Lucas: Exactly. This is where computer vision and AI come in, but integrating them adds cost and latency. Companies like Universal Robots are experimenting with skin-like sensors that cover the entire arm. These provide better localization of contact, allowing the robot to stop only at the point of impact rather than shutting down the whole system. But even then, the regulatory approval process moves slowly. Luna: Because regulators are risk-averse by nature. They would rather err on the side of caution than enable innovation. Lucas: That is fair. But there is also a cultural shift happening on the factory floor. Workers are hesitant to work closely with these machines because they are afraid of causing a stoppage. They view the robot as fragile. If you bump it, it stops. That psychological barrier is just as significant as the technical one. It changes the workflow dynamics entirely. Luna: It turns the robot into a guest in the workspace rather than a partner. That is a fascinating way to put it. Lucas: And it highlights the need for better training protocols. Most operators are trained on traditional robots where the rule is stay out. Training them to collaborate requires a completely different mindset. They need to trust the machine's ability to yield, not just stop. Until that trust is established, the technology will underperform. Luna: Do you see any manufacturers breaking away from the standard approach? Are there exceptions? Lucas: Some niche players are using dynamic speed monitoring. Instead of a hard force limit, they use cameras to track the worker's position. If the worker moves closer, the robot slows down proportionally. If the worker steps back, the robot accelerates. This mimics natural human behavior and reduces the number of full stops. Luna: That sounds much more intuitive. It feels like we are finally moving toward adaptive systems. Lucas: We are, but the adoption curve is shallow. The cost of adding vision systems is high, and the integration is complex. For many small and medium enterprises, sticking to the basic, overly cautious cobot models is still the path of least resistance. They accept the downtime as the price of entry. Luna: So the real bottleneck isn't the hardware capability, it is the regulatory and economic framework surrounding it. Lucas: I think so. We are seeing a divergence between what the physics allows and what the law permits. As long as that gap exists, we will have robots that are technically capable of far more dangerous or precise tasks but are forced to operate in a sandbox of extreme caution. Luna: It makes you wonder if we need a new category of certification that accounts for these intelligent, adaptive behaviors. Lucas: Absolutely. Static rules for dynamic machines is a recipe for inefficiency. We need standards that evaluate the risk profile of the specific application, not just the generic capabilities of the arm. Luna: That would require a lot of collaboration between engineers, insurers, and policymakers. It is not going to happen overnight. Lucas: No, it won't. But the pressure is mounting. As labor shortages continue to tighten, factories cannot afford to have their most advanced tools running at half-speed due to artificial constraints. The market will eventually force the regulators to catch up. Luna: So, in the meantime, plant managers are stuck optimizing around the safety limits rather than optimizing for output. Lucas: For now, yes. It is a temporary inefficiency, but a costly one. The companies that figure out how to navigate these constraints—either through superior integration or by lobbying for updated standards—will have a significant advantage. Luna: A competitive edge derived from safety compliance. That is a twist I did not expect. Lucas: In tech, compliance is rarely just a checkbox. It is a design parameter. And right now, it is a restrictive one. Luna: Let's hope the next generation of cobots helps bridge that gap between caution and capability. Lucas: One can only hope. Until then, we watch the sensors count down.