Hi Everyone,
I have been working on BMX160 for a couple of weeks. The target is to identify the absoulte orientation Yaw angles of several carts which are horizontally placed on the floor. The carts could be heavy and is supposed to be moved manually. So the measurement does not require high dynamics. And the yaw orientation info is needed only when the carts are in static state.
The problem I got is the unexpected magnetic distortion. Since it is indoor application, it suffers a lot of uncertainties of magnetic field variations in different locations.
As the measurement procedure, the mag sensors were calibrated after installation in the cart. Because only the absolute yaw was needed, I only calibrated the X-axis and Y-axis mag values with ellipse-fitting to get the hard-iron offset and soft-iron correction 2x2 matrix.
The absolute Yaw-angle could be extracted by solving atan2(magY, magX) after calibration, assume that magnetic environment were consistent over the room. However, as I mentioned above, the magnetic field could be distorted at different locations. I got really poor results when identifying the relative Yaw angle of 2 carts (YawCart1 - YawCart2). The error could be more than 30 degrees comparing with the results of optical measurement system (which I trust to give me the groud truth relative angle).
I have 2 doubts if I could really use this Absolute Orientation IMU to indicate Yaw-angle in my application:
1. Is there any possible way of compensating the unexpected magnetic distortion?
2. May sensor fusion algorithm help to improve accuracy in my application? As I understand, the fusion algorithm still needs the trustable magnetic information to do the correction. What if the carts were initially placed in different locations in the room when the sensor systems first start and the locations suffered different magnetic distortions?
I kind of get stuck here...Could anyone give me some suggestions? Thank you in advance!
Best regards
Disen
