Hello!
I have tested the sensor fusion library 4.2 and 5.0 all. My sensor is ADIS16445,a 6-axis with gyro + acc. I have tuned parameters according to my sensor noise in fusion.h.
1.Bu all this two libraries don't performance well in continuous shaking condition in hand.For example ,i shake sensors in a flat desk continuously,the pitch and roll will increase slowly.
I have used some AHRS products like Xsens-MTI300 or SBG-IG500,they all performance well in shaking condition in hand,their pitch and roll don't increase ,but only a little oscillation according to the shake.I have tried many methods,like low pass filtering the accelerometers,but can't performance well. I don't know how can they do this?
2.When computing the incremental fast (typically 200Hz) rotation vector rvec (deg),i tried to not minus the gyro bias,this seems performance well.I think the algorithm can't estimate the gyro bias accurately under maneuvering,so the inaccuracy estimated bias will cause the attitude error.I average the gyro datas in 3-5 sec,and use this as the inital bias to be deducted.
3.the linear acceleration time constant,linear acceleration decay factor ! How should I set this parameter?? What is the meaning of this parameter?
Sorry, I can't comment on performance based on the ADIS16445. We have not seen a problem with shake when using Freescale sensors with the V5 library. One of Mark Pedley's standard tests is to shake the IMU has fast as he can to see how fast it recovers. V5 recovers very quickly with our sensors. That said, the library SHOULD be compatible with any manufacturer's sensors. Mark has successfully ported to a number of different manufacturers with similar results. But please don't ask - Freescale can't offer support on competitive devices (the good will of my masters only goes so far :smileywink:). You'll need to rely on other community members or ADI for that.
Regards,
Mike