Error using Kalman function - cannot compute stabilizing Riccati solution
22 views (last 30 days)
Show older comments
Please see below a snippet of my code:
sys=ss(A_ae_control,B_ae_control,C_ae_control,0);
Qn=0.0001;
Rn=0.01*eye(size(C_ae_control,1));
[kest,L,P]=kalman(sys,Qn,Rn);
I can't seem to figure out why this won't run. I keep getting the following error:
Cannot compute the stabilizing Riccati solution P for the Kalman filter.
This could be because:
* RN is singular,
* [QN NN;NN' RN] needs to be positive definite,
* The E matrix in the state equation is singular.
Any insight would be appreciated. I have attached my A, B and C matrices.
Also, is it possible to somehow create a Kalman estimator without any measurement noise (only a disturbance). I tried setting Rn to 0 but not sure why this does not work.
4 Comments
Emir Acevedo
on 13 Sep 2022
I had designing a LQR controller and after an Kalman Filter. For mi satellite control, I used the same Q and R matrix in Ricatti to solve Kest. May that can fix your problem.
Regards.
Answers (1)
Ivo Houtzager
on 17 Apr 2023
Your state-space model is probably not a minimal realization, in other words not fully obervable or not fully controllable. Perform the function minreal to remove unobservable or uncontrollable states.
sys=ss(A_ae_control,B_ae_control,C_ae_control,0);
sys=minreal(sys);
Qn=0.0001;
Rn=0.01*eye(size(C_ae_control,1));
[kest,L,P]=kalman(sys,Qn,Rn);
2 Comments
rajesh r
on 17 Apr 2023
My model is:
A = [1 0.01; 0 1];
B = [0 ; 0];
C= [1 0];
D = 0;
after minreal,
the complete A matrix is zero
Ivo Houtzager
on 17 Apr 2023
You have no inputs defined, because both B and D are zero. Thus your model is fully uncontrollable. So you have review your model inputs for correctness.
See Also
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!