•
|
The LQROutput command calculates the LQR state feedback gain for a system with output weighting.
|
•
|
The system sys is a continuous or discrete-time linear system object created using the DynamicSystems package. The system object must be in state-space (SS) form.
|
•
|
In continuous-time domain, the optimal state feedback gain, , is calculated such that the quadratic cost-function
|
is minimized by the feedback law u = -Kx subject to the following state-space
•
|
In discrete-time domain, the optimal state feedback gain, , is calculated such that the quadratic cost-function
|
is minimized by the feedback law u[n] = -Kx[n] subject to the following state-space
•
|
Q and R are expected to be symmetric. If the input Q and/or R are not symmetric, their symmetric part will be considered since their antisymmetric (skew-symmetric) part has no role in the quadratic cost function.
|
•
|
In LQR with output weighting, the weighting matrix Q acts on the outputs rather than the states as in the standard LQR. Also, the weighing matrix N acts on the inner product of the outputs and inputs (creating a bilinear form) rather than the inner product of the states and inputs.
|
•
|
In addition to the state feedback gain, depending on the corresponding option values, the command returns the closed-loop eigenvalues and the solution of the associated Riccati equation.
|
•
|
The output weighting LQR problem is equivalent to the standard LQR problem with the following weighting matrices:
|
•
|
The direct gain Kr is computed as follows:
|
(discrete time)
•
|
If sys contains structured uncontrollable or unobservable states, they are removed using ReduceSystem before computing the LQR state feedback. The resulting gain is then filled with zeros at positions corresponding to the removed states; however, the other outputs are not filled and, consequently, they may have lower dimensions as expected.
|