I have a nonlinear system, and I need to use the extended kalman filter to estimate it. I know I need the jacobian, but once I get that, is everything else the same as the normal kalman filter?

I currently have the state equations updating in a loop because everything is discrete, so for example

x(0) = 5
for k 2:100
x(k) = x(k-1) + 2

I am unsure how to get the jacobian from constants like this, I was considering by hand but some of the equations are a little complicated and I figured there would be an easier MATLAB solution I dont know. And once I get the jacobian, I am unsure where to go. I have some working code for the regular linear case kalman filter from an older project, can I somehow make that work?

Thanks for any help

  • $\begingroup$ your state equation is linear and deterministic with a constant input. what does your measurement equation look like? $\endgroup$ – user28715 Dec 13 '18 at 8:53
  • $\begingroup$ Hi! It' will be much appreciated if you supply mathematical details of your nonlinear system: it's process equation and its measurement equation. $\endgroup$ – Fat32 Dec 13 '18 at 10:52

It's true that once the nonlinear system is linearized, things will be similar to a linear Kalman filter, except the fact that all the state dependent matrices will also be updated (according to previous / current state estimates) in the loop as well as the state estimates themselves.

You should better provide your systems's equations for a better answer.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.