Here are some results for several joints, with both datasets collected while the robot was walking:
Velocity [rad/s]
Joint
With the gains proposed
The value currently saved in robots-configuration
torso roll
l hip roll
Acceleration [rad/s^2]
Joint
With the gains proposed
The value currently saved in robots-configuration
torso roll
l hip roll
[!IMPORTANT]
According to the plots, the velocity/acceleration of the torso roll appears smoother. This is likely because the original configuration did not enable the Kalman filter for the torso roll.
This PR tunes the gains of the Kalman filter for the joint velocities and acceleration in the EMS.
The filter has been tuned by an automatic procedure that is explained in https://github.com/ami-iit/component_bipedal-locomotion/issues/157
Here are some results for several joints, with both datasets collected while the robot was walking:
Velocity [rad/s]
Acceleration [rad/s^2]
cc @isorrentino @DanielePucci