Latest / The Robotics Podcast with Fexingo: Autonomous Systems, Industrial Robots, and Hardware / Robot Batteries Are the Hidden Safety Problem
Transcript
- Lucas: So last month, Locus Robotics had to recall around 40,000 of their warehouse bots. Not a software glitch. Not a navigation issue. The batteries were catching fire. Luna: Forty thousand bots. That's practically their entire deployed fleet, isn't it? Lucas: It's a huge chunk, yeah. Locus is one of the bigger players in warehouse automation — they power a lot of the e-commerce fulfillment you've probably benefited from. And the recall wasn't optional. The Consumer Product Safety Commission got involved because these lithium-ion packs were overheating during charging. Luna: This feels like the kind of thing everyone in robotics has been quietly worried about for years. We talk so much about sensors and AI, but the battery is literally the thing that can set the whole robot on fire. Lucas: Exactly. And the Locus recall is just the most visible example. There've been incidents with delivery robots in college towns, with AGVs in factories, even with some humanoid prototypes in testing. The fundamental problem is that we're asking these robots to run for longer shifts, charge faster, and pack more power into smaller form factors. That pushes the lithium-ion chemistry into a zone where thermal runaway becomes a real risk. Luna: Can you unpack thermal runaway for someone who isn't a battery chemist? I've heard the term, but I'm fuzzy on what actually happens. Lucas: Sure. Inside a lithium-ion cell, you've got a positive electrode, a negative electrode, and a liquid electrolyte that carries ions between them. Normally it's all stable. But if a cell gets punctured, or overcharged, or just develops an internal short — maybe from a manufacturing defect — the temperature starts climbing. At around 80 degrees Celsius, the separator membrane that keeps the electrodes apart starts to melt. That causes more shorting, which generates more heat. It cascades. Within seconds you can hit several hundred degrees, and the cell vents flaming gas. If it catches neighboring cells, the whole pack goes up. Luna: And that's what happened with the Locus bots? A charging station malfunction somehow triggered it? Lucas: According to the recall notice, the bots' batteries could overheat even during normal charging cycles. The root cause seems to be a combination of the charger firmware and the battery management system — the BMS — not communicating properly. So the BMS thought the cells were at a safe temperature when they weren't. Luna: So it's a software problem as much as a chemistry problem. Lucas: It always is, at this level. The BMS is the brains of the battery pack. It monitors voltage, current, temperature, and it's supposed to shut things down if anything goes out of spec. But if the firmware has a bug, or if the temperature sensors are placed in a spot that doesn't detect a hot cell in the middle of the pack, you get blind spots. Locus scrambled to push a firmware fix and offer replacement batteries, but the trust hit is real. Warehouses are high-stakes environments — one fire can shut down a facility for days. Luna: I remember reading about a fire at an Amazon warehouse in 2021 that was linked to a lithium-ion battery in a robot. Amazon never confirmed the exact cause, but it forced them to reevaluate their charging infrastructure. Lucas: Yeah, Amazon's had their share of incidents too. They've moved to dedicated charging rooms with fire-rated walls and automated sprinklers in some facilities. But that's a workaround, not a solution. The real fix has to come from battery chemistry and better thermal management. Luna: What are the alternatives? Are there chemistries that are inherently safer? Lucas: A few are starting to show up in commercial robots. Lithium iron phosphate, or LFP, is one. It's less energy dense — so you need a bigger pack for the same runtime — but it's much more thermally stable. LFP cells can be punctured without catching fire. Some warehouse bots are switching to LFP for that reason. Another approach is solid-state batteries, where the liquid electrolyte is replaced with a ceramic or polymer. They're virtually non-flammable. The problem is cost and manufacturing scale. Solid-state is still expensive and hard to produce in the shapes robotics needs. Luna: How about the cooling systems themselves? Could better liquid cooling or phase-change materials reduce the risk? Lucas: Absolutely. Some of the high-end humanoid robots — think the ones from Tesla Optimus or Figure — use liquid cooling loops that run through the battery pack and even the motors. That keeps everything at a steady temperature even under heavy load. But liquid cooling adds weight, complexity, and potential leak points. For a small delivery robot that's supposed to cost a few thousand dollars, you can't justify a whole cooling loop. So they rely on passive cooling and careful cell spacing. Luna: There's also a regulatory piece here. The UL certification for robot batteries — I think UL 9540 and 1973 — those are evolving, but they're still behind the curve. Lucas: You're right. UL 9540 covers energy storage systems, and UL 1973 covers batteries for stationary applications. But robots are mobile. They vibrate, they bump into things, they operate in dusty environments. The existing standards don't fully cover the abuse a robot battery experiences over thousands of hours of operation. The industry is pushing for a specific UL standard for robot batteries, but it hasn't been finalized yet. Luna: So we're in this awkward phase where the technology is scaling fast — thousands of bots in the field — but the safety standards are still catching up. Lucas: That's the tension. Investors are pouring money into robotics because the unit economics look great — a robot costs $30,000 but replaces two warehouse workers at $40,000 a year each. But if you have to add a fire-suppression system to every charging station, or replace batteries every 18 months instead of 36, the math changes. Luna: And of course there's the human cost. I don't think people realize how dangerous a robot fire can be in a warehouse full of cardboard and packing materials. It's not like a car fire in a parking lot. Lucas: Exactly. Warehouses are basically fuel for fires. And the lithium-ion gas vented during thermal runaway is toxic — hydrogen fluoride, carbon monoxide, the works. So the fire department needs special training and equipment. Some facilities are now installing dedicated battery fire extinguishers — they use a vermiculite-based agent that smothers the pack thermally. Luna: I've heard of those — they look like big blankets you drop over the robot. Lucas: Yeah, they're called fire blankets or battery fire containment covers. They're effective but they're a last resort. The better approach is to prevent the fire from happening at all. And that means better BMS design, better cell quality, and maybe a shift to safer chemistries even if it means shorter runtime. Luna: One thing that gives me hope is that the larger automakers have been dealing with lithium-ion safety for a decade. Their electric cars have sophisticated battery management and cooling. Some of that technology is starting to trickle down into the robotics space. Lucas: It is, but slowly. An EV battery pack costs thousands of dollars and the BMS is a significant fraction of that cost. In a robot, the battery might be 10 to 15 percent of the total cost. There's less margin to put in a high-end BMS with redundant temperature sensors and cell-balancing circuits. But as the incidents pile up, I think buyers will start demanding better safety specs. Luna: And that's where regulation could push the market. If the Occupational Safety and Health Administration or the fire code bodies start requiring certain battery certifications for robots in warehouses, everyone will adopt them overnight. Lucas: That's probably the fastest route to change. I know some large logistics companies are already writing battery safety requirements into their procurement contracts. They're saying 'if your robot doesn't meet UL 9540A or have a thermal runaway test report, we won't buy it.' That kind of market pressure forces suppliers to fix the problem. Luna: So where does that leave the humanoid robot projects? They're aiming for a 10-kilowatt-hour battery pack to run for a full shift. That's a massive amount of energy in a human-shaped package. If a humanoid goes into thermal runaway in an office or a home, it's a disaster. Lucas: It's a huge challenge. And it's one reason I think we're still a few years away from humanoids working alongside people in uncontrolled environments. The safety case isn't there yet. You need multiple layers of protection — fire-resistant housing, gas venting paths, a BMS that can survive a single-point failure. Most current humanoid prototypes don't have that level of engineering because they're still R&D machines. The ones that do, like some of the more mature designs from Agility Robotics or Boston Dynamics, they've put serious thought into battery safety. But they're also not running for 8-hour shifts yet in dense spaces. Luna: It feels like battery safety is one of those unglamorous problems that will determine how quickly robotics scales. Not the AI, not the dexterity — just the power source. Lucas: Absolutely. And it's a solvable problem. We just need the industry to treat it as a priority, not an afterthought. The Locus recall was a wake-up call. Let's hope the industry listens. Luna: Quick honest thing — we do this show ad-free and listener-funded. A handful of people chip in monthly through buy me a coffee dot com slash fexingo, and that's what keeps the episodes coming. If you're getting something out of these conversations, that's the way to keep them going. Lucas: Yeah, it's a small group and it makes a real difference. No pressure, but if it's on your mind, that's where you can help. Luna: Alright, back to the battery question. What's one concrete thing a robotics engineer can do today to make their next prototype safer? Lucas: I'd say put the BMS on a separate circuit with its own power supply, so even if the main robot brain crashes, the BMS stays alive and can disconnect the battery. And don't skimp on temperature sensors — put at least one per every group of three cells, and make sure they're placed at the center of the pack where heat concentrates. Those two things alone would prevent a lot of fires. Luna: Good advice. And maybe test it by overcharging a cell in a lab, just to see what happens. Lucas: Right — don't just simulate it. Actually run the abuse tests. The data from a real thermal runaway is worth more than a thousand simulations. Luna: Thanks, Lucas. That's a solid note to end on.