This should hopefully let us have more accurate odometry, especially if we have a drivetrain like Mecanum that has low quality odometry.
Though it should also be noted that we may want to just dedicate wheels specifically to odometry instead, as this one CD user points out:
Using separate odometer wheels in X and Y, like this example from 1986, is very common in FTC where most teams use mecanums. Nearly all the top teams use them to achieve their autonomous movements. It is inherently accurate as it is measuring the result, not the input, of wheel rotations. No math, no derivation, no slippage, just a precise measure of how far the robot has moved in X and Y.
If we could have some sort of ball in each corner that has 2 rotary encoders each, also, we could probably detect if our wheels leave the ground (1 or more balls' velocity drops suddenly compared to the others) and temporarily drop confidence in our odometry in the pose estimator
This should hopefully let us have more accurate odometry, especially if we have a drivetrain like Mecanum that has low quality odometry.
Though it should also be noted that we may want to just dedicate wheels specifically to odometry instead, as this one CD user points out:
Source: https://www.chiefdelphi.com/t/mecanum-odometry-off-but-sensors-seem-ok/392401/13
If we could have some sort of ball in each corner that has 2 rotary encoders each, also, we could probably detect if our wheels leave the ground (1 or more balls' velocity drops suddenly compared to the others) and temporarily drop confidence in our odometry in the pose estimator