Are Unscented Extended Kalman filters (UEKF) preferred over Multiplicative Extended Kalman filters (MEKF) in satellite ADCS systems?
What are the trade-offs for these two filters?
I have read that UEKFs are better at estimating 2nd order noise effects because it samples the noise distribution which allows for faster convergence.
There is no UEKF. There is UKF. Yes, it performs better than EKF/MEKF. And yes it's computationally expensive for that matter. And yes, people are looking into SPUKF/ESPUKF to reduce that computational overhead.
But again, follow the golden rules: "If it ain't broke, don't fix it". Typically these filters alone are an illogical point of metrics from a mission perspective. This whole sensor fusion thingy is coupled with hardware. Good hardware + shizzle software = functions, then viola! Shizzle hardware + awesome software = still doesn't work, then "insert hello darkness my old friend here". Random puns apart, there have been precision pointing mission in the past that got away with just a KF and PD, only the control bandwidth had to be pumped up to 4Hz from 2Hz.
It's very mission/hardware/satellite specific. Check out the classic paper: Survey of Nonlinear Attitude Estimation Methods from hotshots featuring: John L. Crassidis, F. Landis Markley and Yang Cheng. Blue Canyon Technology from what I understand runs a 6 state MEKF. If it works for them, it could easily work for a rigid body mission for you.
P.S. Sorry for the informal lingo, had too much caffeine today. Have a good day :)
Ad Astra.
Edit (in progress):
I realized there is indeed a UEKF. Never saw it applied on a sat mission before. Shall dig it, in free time, and come up with a rather constructive answer. This Paper claims that UEKF performs better than UKF, and has less computational time. Interesting. This is not a paper on ADCS, but target tracking instead. Different frameworks.
