Latest / The Robotics Podcast with Fexingo: Autonomous Systems, Industrial Robots, and Hardware / Why Robot Safety Protocols Are Holding Back Automation
Transcript
- Lucas: Here is a number that keeps me up at night in the best possible way: the average time from robot purchase order to first cycle is currently about nine months. That is nearly double what it was five years ago. We have machines that can weld, paint, and assemble with micron-level precision, yet they sit in crates or on pallets for half a year before they touch a single product. Luna: Nine months feels like an eternity when you are trying to scale production. Is it mostly procurement friction, or are we hitting walls on the floor? Lucas: It is almost entirely the latter. The hardware is ready. The software is ready. The problem is the paperwork and the physical reconfiguration required to make those machines legal to operate next to human workers. We are seeing companies buy state of the art cobots only to wrap them in three-foot steel cages because the insurance underwriters won't sign off otherwise. It creates this absurd paradox where the most advanced tech is treated like a hazardous waste container. Luna: So the constraint isn't capability anymore, it's liability. And who decides what counts as safe? Is it the engineers or the lawyers? Lucas: It is a messy intersection of both. On one side, you have ISO standards that were written for industrial arms moving at high speeds inside fenced zones. Those rules assume a static environment. But today, you are putting a robot arm that weighs fifty pounds into a space where humans walk, bend, and reach. The standard requires either total separation or extremely expensive force-limiting sensors. If you choose separation, you lose the flexibility you bought the robot for. If you choose sensors, you are looking at adding ten thousand dollars in hardware just to get a basic light curtain system approved. Luna: Ten thousand dollars just for the safety gear seems steep. Does that cost actually pay off in reduced downtime, or is it just dead weight? Lucas: It pays off if you calculate the alternative. A single pinch-point incident can shut down a line for weeks during investigation. So the safety spend is technically insurance. But here is the catch: the inspection process itself is slow. You cannot just turn the key and go. Every change in the work envelope requires a new risk assessment. If you move the robot two feet to the left to accommodate a new conveyor, you might need to redo the entire safety documentation. That administrative lag is what eats those nine months. Luna: That sounds incredibly rigid. How do factories manage that level of bureaucracy when their actual workflows are changing weekly? Lucas: They don't. They stall. I spoke with a plant manager in Michigan last week who told me they had three robots sitting in the loading dock for four months because they couldn't agree on the safe operating distance for a new packaging station. The engineering team wanted to test it dynamically. The safety committee wanted a fixed fence. The result was zero output from equipment that could have doubled their capacity. It is a classic case of optimizing for the worst-case scenario instead of the actual use case. Luna: But isn't worst-case optimization necessary when dealing with heavy machinery? You can't gamble with employee safety. Lucas: I am not suggesting we gamble. I am suggesting we are using blunt instruments for nuanced problems. Modern cobots have torque sensors that detect resistance instantly. If a human walks into the path, the robot doesn't need to slam to a halt; it can decelerate smoothly based on proximity. But the regulations often demand that binary stop. So we install massive fences to guarantee that binary stop, even though the machine itself is capable of much safer, more fluid operation. We are building walls because the rules don't understand gradients. Luna: So the technology has evolved past the rules. That is a dangerous gap to have. Who gets to update these standards? It usually takes years. Lucas: Exactly. The standard-setting bodies move at the speed of consensus, which is glacial compared to silicon and servo motors. Meanwhile, vendors are shipping products that claim 'safe collaboration' out of the box. But until a factory can prove that claim to an auditor, it is just marketing. And auditors are risk-averse by profession. They would rather reject a safe setup than approve a risky one and face scrutiny later. So the default answer is always 'build a bigger cage.' Luna: This feels like a trust issue as much as a technical one. If the data shows the robot is safe, why does the paper trail matter so much? Lucas: Because paper trails assign blame. If something goes wrong, the company needs to show they followed the prescribed protocol. The protocol says 'install guard.' If you skip the guard and rely on software limits, and then a power surge causes a glitch, you are exposed. The guard is a physical proof of due diligence. It is not efficient, but it is defensible. In the current legal landscape, defensibility often trumps efficiency. Luna: Defensibility over efficiency. That is a brutal trade-off for margins. Does this mean small manufacturers are just locked out of this tech? Lucas: Precisely. Large corporations have dedicated safety departments and legal teams who can navigate the gray areas. They can negotiate with insurers. They can afford the upfront cost of custom safety engineering. A shop with fifty employees cannot. They see the robot brochure, they see the price tag, and then they realize the compliance cost doubles it. So they stick with manual labor. They stay competitive on volume, but they cannot compete on innovation speed. Luna: That widens the gap between big players and everyone else. It feels like we are stifling a revolution because we are afraid of the paperwork. Lucas: We are. And the irony is that these robots are often safer than humans. Humans get tired. Humans drop tools. Humans make judgment errors under pressure. A properly tuned cobot follows its parameters exactly. But we treat it like a wild animal that needs to be chained up. It is a cultural hangover from the early days of automation, where machines were dumb, loud, and genuinely dangerous. Luna: If the machines are objectively safer, why is the perception stuck in the past? Is it just inertia? Lucas: It is inertia, yes. But it is also financial structure. Insurance premiums are tied to historical loss ratios. If you introduce new technology, your risk profile changes. Insurers don't like unknown variables. So they charge higher premiums for unproven setups. To lower those premiums, you have to provide more evidence, which brings us back to the guards and the documentation. It is a closed loop. You pay more because you are new. You pay more because you lack history. You lack history because you can't deploy fast enough to build it. Luna: There has to be a middle ground. Some kind of provisional certification or a sandbox environment where these systems can be tested and validated faster. Lucas: There are pilots starting in Europe, specifically in Germany where the automotive industry is pushing hard. They are creating 'regulatory sandboxes' where companies can operate with modified safety protocols under strict monitoring. If no incidents occur over six months, the protocol gets codified. It is a data-driven approach to regulation. Instead of assuming danger, you prove safety through continuous telemetry. It is brilliant in theory, but it requires regulators to trust digital logs over physical barriers. Luna: Trusting digital logs is a huge leap. One hack, one glitch, and you are blaming the code. Physical barriers don't hack. Lucas: True. Cybersecurity becomes the new safety concern. If your robot is connected to the network for remote monitoring, a breach could override those soft stops. So now you have dual threats: physical injury and digital sabotage. Managing both simultaneously adds another layer of complexity to deployment. You aren't just securing the perimeter; you are securing the protocol stack. Luna: So we are solving one set of problems only to invite another. This is getting complicated very quickly. Lucas: It is. And that complexity is the real cost driver. It isn't just the sensor hardware. It is the integration labor. You need engineers who understand mechanics, software, law, and cybersecurity all at once. Those people are rare and expensive. So you end up paying premium wages to bridge the gaps that the regulations created. Luna: Let's pause for a second. This deep dive into robotics compliance wouldn't be possible without our listeners. If today's tech conversation gave you something usable, consider supporting the show directly. We are completely ad-free and listener-funded. A small group already chips in monthly through buy me a coffee dot com slash fexingo, and that is literally what funds making this many of these episodes possible. It keeps the research sharp and the hosts independent. If you value that, check it out. Thanks for keeping the lights on. Lucas: Back to the gridlock. We mentioned the German sandbox model. The US is slower to adopt that. OSHA is notoriously cautious. They tend to look at precedents. Since there are few precedents for fully unguarded collaborative cells, they default to the old rules. Until a major court case forces a reinterpretation, or a large coalition of manufacturers lobbies for change, the status quo holds. Luna: So we are waiting for a legal earthquake or a political shift to unlock this tech. That seems like a fragile strategy for global competitiveness. Lucas: It is. And look at the opportunity cost. If deployment time drops from nine months to three months, the return on investment for these machines changes dramatically. Companies start buying smaller batches, testing more scenarios, iterating faster. The market becomes more agile. Right now, it is frozen. We have the tools to transform manufacturing, but we are handcuffing ourselves with red tape. Luna: Handcuffed by our own caution. It makes you wonder if other industries are facing similar bottlenecks. Autonomous vehicles come to mind. Lucas: Absolutely. Same pattern. The tech works. The safety data is improving. But the liability framework hasn't caught up. Everyone is waiting for someone else to take the first step legally. It is a collective action problem. If every factory waited for perfect clarity, we would never automate anything beyond simple pick and place tasks. Luna: Is there any movement on the vendor side? Can robot makers help solve this by simplifying the compliance burden? Lucas: Some are trying. There are platforms emerging that auto-generate the safety documentation based on the robot's configuration. You plug in the specs, the software runs the risk assessment algorithms, and spits out a draft report for the auditor. It cuts the admin time in half. It is a good start, but it doesn't change the underlying rulebook. It just makes the inefficiency slightly less painful. Luna: A digital assistant for red tape. I suppose that is better than nothing. But it still feels like we are treating the symptom, not the disease. Lucas: We are. The disease is the mismatch between dynamic technology and static regulation. Until we find a way to make regulation adaptive—where rules evolve as fast as the machines—we will keep seeing these deployment delays. Nine months is the new normal, and it is dragging down productivity across the board. Luna: So the question for the next decade is whether we can build a regulatory system that moves at the speed of silicon. Lucas: That is the million-dollar question. And until we answer it, those robots will keep sitting in crates, waiting for permission to work. Luna: Waiting for permission, but held back by fear. It is a reminder that sometimes the hardest part of innovation isn't building the future, it's convincing the present to let it in. Lucas: Exactly. The physics are solved. The economics are clear. The only thing missing is the trust. And trust is the hardest variable to engineer.