I know that Kalman filter is optimal filter under some assumption like process and measurement noise are Gaussian. But if the process and measurement noise is non-Gaussian, the estimation of the Kalman filter will start to be inaccurate, especially with noisy measurements. Can you suggest adaptive Kalman filter algorithm for non-Gaussian noise condition?

  • $\begingroup$ Anything other than Gaussian... has a wide variation. Can you add some detail regarding what specific non-Gaussian characteristics your noise regime has? $\endgroup$
    – Peter K.
    Commented Mar 10, 2023 at 19:36
  • 1
    $\begingroup$ First, "adaptive filter" narrows the list of possibilities down too far -- you want to ask for non-linear approaches in general. Second, either do as @PeterK. asks and edit your question to describe your specific problem, or to make it clear that you're looking for a survey answer that talks about all the possible approaches. Dealing with nonlinear systems or non-Gaussian noise, or cost functions that aren't quadratic isn't a nice deductive process like the Kalman filter -- you end up applying a bag of tricks, and different situations require different tricks. $\endgroup$
    – TimWescott
    Commented Mar 10, 2023 at 22:34
  • $\begingroup$ Example of non-gaussian noise is shot noise or time varying noise. $\endgroup$
    – guidolard
    Commented Mar 15, 2023 at 20:02


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.