Closed sbrodeur closed 7 years ago
Madgwick filter in post processing seems to give incorrect values. For example, while the z axis should always stay straight up, it tend to move a drift a lot. Will need to investigate.
Turns out I was probably at fault for the previous Madgwick issues. However, we have to address the gain that is set to 0.1 in live mode and 0 in offline mode. A higher gain increases convergence rate but adds noise. 0 gain is smooth and 0.1 jitters just a bit. We will have to decide if faster convergence is worth the slight amount of jitters induced. For more details about the settings see : http://wiki.ros.org/imu_filter_madgwick
The gain should be set to 0.1 in offline mode. At 0.0, if I look at the equations, there is no filtering. This is why it is noisier in offline mode compared to real-time mode.
Gain increased to 0.1 in commit d0f07f1484c710d7ac7f4bd2c296cd2a40336ef5 .
Use these commands to visualize each sensor: