Post by Johannes Gießibl

Someone told me what they were paying for torque sensors in robot joints. Several hundred euros per unit. At annual volumes of around 50,000 units. I did the math in my head while they were still talking. A humanoid robot needs between 20 and 40 joints with torque sensing. At several hundred euros each that is between 4,000 and 12,000 euros of torque sensing cost per robot. Before a single motor, gearbox, battery, or structural component. Before software. Before assembly. Before margin. The cost target for a humanoid robot that can actually scale is somewhere between 10,000 and 30,000 dollars total. The sensing layer alone, on current strain gauge technology, would consume between 15 and 40 percent of that entire target cost. And strain gauge based torque sensors bring three additional problems that cost cannot fix. They are not overload capable. A single impact event degrades the adhesive bond. In a robot operating in a real environment, overload events are not exceptions. They are part of the duty cycle. They drift over time systematically, not randomly, meaning the safety threshold the controller is enforcing is no longer where the system thinks it is. And their bandwidth is limited by the adhesive bond itself, which acts as a mechanical low pass filter on everything the joint actually experiences. The humanoid robotics race is real. The investment numbers are real. The technology roadmaps are serious. But nobody has solved the sensing cost structure. And you cannot build a million humanoid robots per year at 20,000 dollars each with a sensing architecture designed for 50,000 industrial robots per year at several hundred euros per joint. The physics of the application requires a completely different sensing principle. Not a cheaper version of what exists. Something fundamentally different. What does your joint sensing cost per unit today?