Dynamic Kalman Filters / Optimization

[varun-un]

There is some optimization that can be done on our KFs, such as our parameter and coefficient specifications for things like sensor noise, so there is a potential to implement automated/learning optimization models to minimize our variances and errors to find better coefficients, if they exist.

This task is to research and potentially develop some of these methods, which may be anything from simple gradient descent optimization on our LKF to full on Adaptive KFs. This is ambitious and may be unnecessary, so if we are to implement something like this, there needs to be justification/proof of benefit

[varun-un]

Useful paper: https://www.sciencedirect.com/science/article/pii/S1367578823000731?fr=RR-2&ref=pdf_download&rr=8bb6aa6d6f3a3934

[varun-un]

Look at Bayesian Optimization: https://docs.google.com/presentation/d/1CxUI5u8NG1F5PyMZDT-WGybDiwL74pYkN6WoqNeT1R8/edit?usp=sharing

Slide 28-29 in PDR