Knob

Noise covariances (Q and R)

You might say: filter tuning

Q models uncertainty in the motion or process model. R models uncertainty in a sensor measurement.

What it does

Controls how much prediction and measurement compete in the filter.

Ask

Do residuals and state errors agree with the stated covariance?

Trouble

If R is too small, noisy measurements cause jumps. If Q is too small, real motion is ignored and the estimate becomes overconfident.

Sources

1. Timothy D. Barfoot, State Estimation for Robotics, Chs. 6–7There are Jacobians embedded in the Q k and R k covariances for the noise.

Use it on your project

teardown-apply: hold "Noise covariances (Q and R)" (noise-covariances) from the teardown at https://teardownatlas.com/robotics/ against my project

Paste it into a coding agent that has the teardown skills.

An AI agent wrote this map with the teardown skill. It can contain errors: check the sources before you rely on a part. About and legal