How should I modify my real-time IMU data for usage in insfilterA​sync/insfi​lterMARG object(s)?

I'm trying to use the example imu-gps fusion code from the MATLAB example (site adress is at the end of the passage) for my real-time ins-gps applicaiton. I sense that there is a discrepency in the imu sensor frames between my data and MATLAB's imu sensor. I can see that accleration in Z axis is 9.81 in the simulated data whereas in realtime data it is -9.81 (Both sensor body and local navigation frames are NED.) Although I have modified the acceleration, angular velocity and magnetic field vectors many times, I can't find the true modification for the filter to work. So how should i modify my accelerometer, gyroscope and magnetometer data in order to use in insFilterAsync/insfilterMARG specially in functions fuseaccel, fusegyro and fusemag?

Example code: https://www.mathworks.com/help/fusion/ug/pose-estimation-from-asynchronous-sensors.html

[zipeng li]

I meet the same problem as you. My imu data is x-right, y-front, z-up. I transform them to x-front y-right z- down to send into insfilter of matlab (which maybe x-front, y-right, z-down in NED). However, I found that I also need to negative the measurement of accelerometer, and do not need to negative the measurement of gyroscope. I am very confused.

[Ilkay Ataol]

I had taken both of the acceleration and angular velocity measurements negative and adjusted the measurement noise parameters so i was wondering that the measurement noise parameters made the filter the work somehow. Since i wasn't sure, I wanted to ask :) I will try the filter without taken negative of gyroscope following your comment.

Answer 1

[Brian Fanous]

This example may help in understanding how to do this:

https://www.mathworks.com/help/nav/ug/logged-sensor-data-alignment-for-orientation-estimation.html

You may need to swap and/or invert axes, but, essentially, you want to map your recorded IMU data to the same convention as imuSensor outputs.

[Ilkay Ataol]

This example has helped me a lot and i also agree with zipeng li on taking negative of accelaration and leaving angular velocity as it is. Hovewer, when I tried the alignment of the logged orientation and imufilter orientation, I had unexpected results so I didn't take the conjugate of the logged orientation for alignment purposes and the result was as I expected. So I was wondering why the orientation was conjugated for alignment in this example. Can you elaborate on this issue?