-
Notifications
You must be signed in to change notification settings - Fork 9
Description
while aligning to climb, when the drivetrain is supposed to be holding still at its goal, we noticed noise of about 5 mm amplitude.
hiding the nearby apriltags removed the noise, so presumably this is coming from the tag sightings, or something about the logic that processes tag sightings.
we should reduce this noise to 1 mm or less.
for example, we could adjust the uncertainties computed in Uncertainty.java (increase the vision uncertainty or reduce the state uncertainty)
or we could make the top-level drivetrain controller less responsive: this is configured in ControllerFactorySE2.java. a good candidate for investigation would be the velocity gain: there's probably more noise in the velocity signal than the position signal (look at it before twisting the gain knob).
or we could add some other kind of filter for the pose estimate, to simply smooth out high-frequency input.
or something else.