How can I solve the problem of designing state observer when using model predictive control toolbox

When using the model predictive control toolbox, the matlab always says that,'Problems encountered when designing the overall state observer (Kalman filter)'. Is there anybody who can solve this problem? Thank you very much!!!!

Answer 1

[André Paradis]

For Model Predictive control purposes, your system must be observable in its state-space representation.

This being said, an observable system can become unobservable do to numerical errors. Thus, you could modify the state estimator gain using :

setestim(MPCobj,M);

Also, if your system is not defined in a state-space representation (e.g. TF), MATLAB will convert it to one. In this case, try using

MPCobj.Model.Plant=ss(model,'min');

Where your system will converted in its state-space representation with the minimal number of states.

Hope this helps

[fariba]

dear andre i have this problem with my mpc. what do you mean ,modify modify the state estimator gain? my gain in mpc is like below M =

   1.0e+03 *

    0.0013   -0.0134
    0.0000    0.0001
    0.1136   -1.1636
    0.0000    0.0000
   -0.0000   -0.0000
    0.0000    0.0000
    0.0000   -0.0000
   -0.0003    0.0134
what should i do?

Answer 2

[fariba]

dear andre i have this problem with my mpc. what do you mean ,modify modify the state estimator gain? my gain in mpc is like below M =

   1.0e+03 *

    0.0013   -0.0134
    0.0000    0.0001
    0.1136   -1.1636
    0.0000    0.0000
   -0.0000   -0.0000
    0.0000    0.0000
    0.0000   -0.0000
   -0.0003    0.0134
what should i do?