kriswiner
commented
5 years ago The Madgwick (or Mahony) filter is an absolute orientation filter. When properly used the quaternions are 1, 0, 0, 0 when the board is oriented toward North parallel to the Earths surface.
You can select which axis (accel, say) is defined to be North. The Madwick algorithm requires a consistent sensor data input depending on the convention you select. I use NED but ENU can also work, so that the entry has to be (aN. aE, aD, gN, gE, gD, mN, mE, mD) or (aE, aN, aU, gE, gN, gU, mE, mN, mU). As long as the data entry is consistent for one of these conventions, the yaw will always read 0 when the device points North.
On Mon, Nov 5, 2018 at 10:20 AM amarsh notifications@github.com wrote:
I have a question that is really very basic but I am struggling with. Consider a scenario where you control the board containing the IMU, but you do not have control over where the board gets installed. In this case the filter is going to return the orientation of the board, not the orientation of the thing you really care about. What is the correct way to transform the filter output into the desired reference frame?
I assume you do something like the following. Collect some amount of data while the vehicle is in a steady state from the desired reference frame and then from the installation location. For each position average the acquired data and zero the gyro values (since we are at steady state). Repeatedly feed the computed values into the filter until the output converges. This should give you a quaternion representing the orientation at each point of interest.
At this point I am not sure what to do. Can you simply compute roll pitch and yaw for each location and then compute a constant to remove from measurements at the installed location? Or do you need to compute and perform a quaternion rotation prior to converting to roll, pitch, and yaw?
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/kriswiner/MPU9250/issues/322, or mute the thread https://github.com/notifications/unsubscribe-auth/AGY1qrQMUoDqEQ_q1irfwv4r-uqtS-kSks5usIF_gaJpZM4YO_kQ .
amarsh
I have a question that is really very basic but I am struggling with. Consider a scenario where you control the board containing the IMU, but you do not have control over where the board gets installed. In this case the filter is going to return the orientation of the board, not the orientation of the thing you really care about. What is the correct way to transform the filter output into the desired reference frame?
I assume you do something like the following. Collect some amount of data while the vehicle is in a steady state from the desired reference frame and then from the installation location. For each position average the acquired data and zero the gyro values (since we are at steady state). Repeatedly feed the computed values into the filter until the output converges. This should give you a quaternion representing the orientation at each point of interest.
At this point I am not sure what to do. Can you simply compute roll pitch and yaw for each location and then compute a constant to remove from measurements at the installed location? Or do you need to compute and perform a quaternion rotation prior to converting to roll, pitch, and yaw?