Is the n4sid model an LTI model?

I have a prediction model derived by n4sid.

when I set the variables with setmpcsignals for MPC controller, it is warned "The first input argument of the "setmpcsignals" command must be a non-empty "lti" object".

Isn't an idss model LTI?

[LEI CAO]

My model is a 5-input-3-output state-space model which can be opened in LTI viewer. And I simulated the MPC controller in MPC tool GUI.

Now I would try with an arbitary distuabance other than step, sine or gaussian specified in MPC tool. But when I use "setmpcsignals" to define the variables, it still gives the alarm...

What could be the possible reason?

[LEI CAO]

I guess the problem is in the matrix K.

The idss model is

x(t+Ts)=Ax(t)+Bu(t)+Ke(t)

y(t)=Cx(t)+Du(t)+e(t)

The matrix D=0, K is not zero.

When I set K=0, the alarm doen't occur any more. But the model is not good enough then.

Is there any mothod to overcome this issue?

[Rajiv Singh]

What matlab release are you using? Could you post some reproduction steps? As such, an estimated idss model should be usable in MPC and you should not have to set K to zero.