This would happen for BNO055. Because the situation on the vehicle is more complicated than on hand. If I rotate the BNO055 randomly, and then goes back to the original position, the yaw angle would differs from the beginning for more than 10 deg.
And the rotation rate on hand should be different from that on vehicle. For example, it may takes 1 second to rotate the BNO055 360 degree on hand, but it may takes around 10 seconds for the vehicle to rotate 360 degrees. And for gyro, long time means large bias integration. So the yaw angle bias on the vehicle is larger.
And the vibration on the vehicle should also be larger that on hand. That would be another main factor.
We still recommend to use NDOF mode for BNO055 to get yaw angle.
Thank you for your reply. I would want to use NDOF mode, but with this mode I have checked that calibration of magnetometer is lost quickly and value of the yaw angle changes abruptly when sensor doesn't find the magnetic north. Also, there is a lot of electrical noise around the sensor and I suppose that it affects the magnetometer.
Also to the accelerometer and gyroscope measurements? If I don't use the magnetometer in IMUPLUS mode, magnetic interference will affect the accuracy?
If you use IMU mode only, magnetic interference will not affect the accuracy, but slow movement will affect the accuracy.