Lower-Limb Calibration Uses Foot Contact

Single-source brief: arXiv reports a self-contained calibration method, but independent confirmation was not supplied.

What Changed

A new method uses onboard sensors during static double-support stances.

It does not require motion-capture systems or fiducial targets, according to arXiv.

The framework uses joint encoders and an IMU mounted on the pelvis.

Both feet must stay fixed on the ground. The method checks whether the calculated foot-to-foot relationship stays constant.

It then solves for 12 joint-offset values with nonlinear least squares.

Why It Matters

Joint encoder offsets affect whether a humanoid’s lower-leg model matches its real pose.

Bad offsets can cause errors in foot height and leg odometry.

The reported tests cut foot-height RMS residuals from 4.26 to 2.20 mm on A3.

On A2, residuals fell from 8.03 to 1.43 mm.

The researchers also used a LiDAR-inertial reference to check one pitch-linked error channel.

Deployment Reality

This is a calibration approach, not a deployed humanoid product.

The reported evidence includes simulation and real-machine injection tests.

The method needs varied leg poses to identify several offsets well.

Some pitch-chain offsets remain dependent on a standing-posture prior.

That limit matters for operators. A system can improve contact consistency without fully separating every joint error.