Question 1.
In state estimation using Extended Kalman filter, Is there a way to determine noise covariance matrices (Q & R) other than the try and error method? as I only have access to a simulation of the model, and the parameters optimized in the simulation may not work as expected in real-life applications.
I was thinking that maybe there is a way to initialize Q and R with some random values and with some technique we let them converge.
Question 2.
Regarding linearization of the motion model and measurement model, should any additive noise be added? if so, what for? and how to determine noise values?
To determine the qualitatively of the noise created, you can display the noise and the real image, and to display the quantitative noise image, you can use both the noise image diagram and the color values in the image. to add noise to the filter, see the following: Simulate Film(Grain In An Image With Photoshop) , (https://support.apple.com/guide/motion/add-noise-filter-motn169fc8d2/mac#:~:text=Table%20of%20Contents-,Add%20Noise%20filter%20in%20Motion,-Adds%20an%20overlay)
- Badges
- Science topic
- Similar topics
- Cognitive Science
- Thinking