The documentation has been updated in Release 2008a (R2008a). For previous product releases, read below:
Non-zero D models are not allowed.
Specifically:
If an MPC controller is created directly form a mod representation of state space model with non-zero D using a function like "smpccon" , an error will be thrown.
For example the system with direct feed through:
X(k+1) = 0.5x(k) + 0.5u(k)
Y(k) = x(k) + u(k)
represented in "mod" form using:
sys =ss2mod(.5,.5,1,1,1)
when used to create a default unconstrained MPC controller using:
Ks=smpccon(sys)
results in the error:
??? Error using ==> smpccon
First nu+nv=1 columns of D matrix in IMOD must be zero.
You cannot get around the problem using the step response based MPC controller synthesis functions such "mpccon" because these functions require process models in step from necessitating the use of "mod2step" to convert from state space form. However, "mod2step" nulls the first term in the impulse response , effectively killing the D term .
