Tuning of GNSS aided INS and PARS aided INS

[Trym Synnevåg]

Hi, i am working on a dataset with IMU, GNSS and PARS data produced from a real experiment of a fixed wing UAV.

I am wondering if there is any document for tuning of the Preintegration factor, Bias factor and GNSS factor for FGO in GTSAM.

In my current FGO system the Preintegration factor and GNSS factor base its covariances on the uncertainty of the experimental setup and sensors. Noise in GNSS, accelerometer and gyroscope measurements. These tuned covariances are also used in an MEKF which estimate on the same dataset and yield a good estimation. So based on this i assume these should be accurate.

However, there are additional tunable variables in the FGO not present in the MEKF.  In the preintegration factor the 'integrationCovariance' and 'biasAccOmegaInt' should be tuned. And in the bias factor the bias noise for accelerometer and gyroscope should be tuned.

Do you have any tips and tricks for tuning these and also some more information about their actual impact on the system. Based on my experience of tuning the Bias factor has a considerable impact on the estimations at least.

Lastly, what is the difference between the Bias factor covariance and the Preintegration bias covariance. Should these be equal? Or are they tuned separately.

Best regards,

Trym Synnevåg