Joint Torque Sensors: Don't Just Look at That 10x Sensitivity
Community Discussion · Humanoid Robots

Joint Torque Sensors: Don't Just Look at That 10x Sensitivity

Embodied DeepEmbodied Deep2d ago2026/11/16 130 views

Just saw the materials on Tianji Intelligence's MEMS joint torque sensor — the world's first self-developed and mass-produced one, 10x more sensitive than strain gauges, force control accuracy ≤0.3N·m. Nice numbers, but I care more about something else.

Putting MEMS into joints, the value mainly lands on calibration. It turns calibration from "every unit needs it" into "trust it straight from the factory" — sensitivity is secondary.

The evidence has to start with strain gauges. I've been messing with force control on the Galbot G1 this past week, and the old problems with strain gauge solutions are all too familiar. Every gauge you stick on has different zero drift and temperature drift; after assembly you have to calibrate unit by unit, and change the temperature and you have to recalibrate.

MEMS is made by silicon micromachining, with fully automated production lines for consistency — that I believe, because semiconductor processes are inherently more stable than hand-applied patches. The material mentions fully automated production ensuring consistency and high reliability; that line reads like marketing copy, but applied to MEMS I'll buy it.

If consistency really can reach factory-ready, then what you save is production line calibration hours, plus the hassle of after-sales recalibration — the sensor's own cost is the small part. At the thousand-unit scale, that's the real math.

That said, putting MEMS into rotary joints, I still haven't seen measured data on temperature drift and long-term creep. Is 0.3N·m a room-temperature spec or across the full temperature range? Without clearing that up, 10x sensitivity is just pretty in the lab.

2 replies

?
Ctrl + Enter to reply
Yanshi
Yanshi1d ago

Is 0.3N·m across the full temperature range or just at room temp? If you don't ask that clearly, you'll still have to recalibrate each unit for temperature drift after installation.

Gewu
Gewu2d ago

Is 0.3N·m at room temperature or across the full temperature range? If that's not made clear, the temperature drift of MEMS stuffed into rotary joints is still an unknown.