Latest / The Robotics Podcast with Fexingo: Autonomous Systems, Industrial Robots, and Hardware / Why Robot Arms Still Can't Open a Jar
Transcript
- Lucas: Luna, I want to start with something that sounds trivial but is actually one of the hardest unsolved problems in manipulation robotics: opening a jar. Luna: A twist-off lid? Really? I mean, I've done that since I was six. Lucas: Right, and that's exactly the point. A human child can do it, but most industrial robot arms — the ones that weld car frames or pick circuit boards — can't reliably twist a lid off a glass jar. Luna: Is it the torque? I imagine you need a lot of force to break the seal. Lucas: It's not just force — it's the combination of grip, torque, and compliance. When you twist a jar lid, your hand automatically adjusts the angle and pressure. You sense when the seal breaks and you change your motion. A robot arm programmed with a fixed torque value will either not open it or crush the jar. Luna: So it's a sensing problem as much as a strength problem. Lucas: Exactly. And it gets worse with real-world variation. Lids come in different diameters, different materials — metal, plastic — and different amounts of vacuum seal. A robot that opens one brand of pasta sauce might fail on the next. Luna: I remember reading about a 2025 study from MIT's CSAIL lab where they tested several commercial grippers on standard mason jars. The best performer still failed about 30 percent of the time. Lucas: That's the one. They used a parallel-jaw gripper with a compliant wrist and a torque sensor. The failure modes were instructive — sometimes the lid slipped because of low friction, sometimes the robot couldn't coordinate the twist with the pull, and sometimes the glass jar rotated inside the gripper. Luna: So the robot doesn't have a 'hand' that can stabilise the jar while twisting the lid. It's relying on one gripper to do both. Lucas: Right. In industry, you'd have a fixture holding the jar. But in a home environment, the jar is sitting loose on a counter. The robot has to hold it down while applying torque — that requires a second manipulator or a very clever single-arm strategy. Luna: And that's why we don't see robot butlers opening pickle jars yet. Lucas: Exactly. There's a startup called Twist Robotics — they raised about $8 million in 2024 to build a countertop robot that could open jars and bottles for people with limited hand strength. They pivoted to industrial packaging in early 2026 because the home use case was too hard. Luna: Wait, they gave up on the consumer market? That's a big deal. Assistive robotics is supposed to be a growing space. Lucas: It is growing, but the technical requirements for safe, reliable operation in an unstructured home are much higher than in a factory. At home, the jar might be sticky, the lid could be cross-threaded, the counter might be slippery. A factory can control those variables. Luna: So they basically realised that the robot was failing too often and it wasn't safe to leave it unattended. That's a trust issue. Lucas: Exactly. And that's the recurring theme in this whole series — the gap between what works in a lab or factory and what works in a messy, unpredictable home. Luna: You know, it's funny — we talk about these limitations every week, and yet the same problems keep showing up in different forms. It makes you appreciate how much our brains handle without us thinking. Lucas: Speaking of appreciating things — if today's tech conversation gave you something usable, a quick way to keep this show going is buying me a coffee dot com slash fexingo. A couple of dollars a month is genuinely what keeps these episodes ad-free and independent. Luna: Yeah, it's a small thing that makes a real difference. We don't run ads, so listener support is what keeps the lights on. Lucas: Back to the jar problem — there's also a material science angle. The friction between a rubber gripper pad and a metal lid changes with temperature and moisture. A lid that's been in the fridge behaves differently than one at room temperature. Luna: So the robot would need to adapt its grip in real time based on surface conditions. That's a control loop running at maybe a hundred hertz with vision and touch feedback. Lucas: Right. And that's computationally expensive for a small, affordable robot. Some researchers are trying to use soft robotic grippers — inflatable fingers that conform to the shape of the lid. But soft grippers have lower torque capacity. Luna: Trade-offs everywhere. What about using a tool? Like, could the robot pick up a jar opener gadget? Lucas: That's actually a promising direction. Instead of the robot's hand doing everything, you design an end effector that's specialised for jar opening — like a rotating collar with internal grippers. A few companies are exploring that for commercial kitchens. Luna: So the robot uses a dedicated 'jar opener' attachment, the way you'd use a can opener. That feels more feasible than a general-purpose hand. Lucas: Exactly. But then you need the robot to recognise the object and swap tools autonomously, which is another hard problem. Still, it's more tractable than solving general dexterous manipulation. Luna: It's like we're building a toolkit for robots, one task at a time. And jar opening is one of those tasks that reveals how far we still have to go. Lucas: I think that's the right takeaway. Next time you twist open a jar without thinking, remember — for a robot, that's a moonshot.