The following question is from a System Theory test without answers or solutions:

Consider the continuous-time state-space representation

$\frac{d}{dt}x(t)=Ax(t)+Bu(t), \quad y(t)=Cx(t), \quad t\in \mathbb{R}^{+}$

With the marices given by

$A=\begin{bmatrix}0&-1&0&0\\-2&0&0&0\\1&0&-1&0\\0&0&0&-2 \end{bmatrix}, \quad B=\begin{bmatrix}0\\0\\-1\\1 \end{bmatrix} \quad \text{and} \quad C=\begin{bmatrix}1&0&0&-1 \end {bmatrix}$

Observe that the controllabilty and observabiltiy matrices are given by

$\mathcal{C}=\begin{bmatrix}0&0&0&0\\0&0&0&0\\-1&1&-1&1\\1&-2&4&-8 \end{bmatrix} \quad \text{and} \quad \mathcal{O}=\begin{bmatrix}1&0&0&-1\\0&-1&0&2\\2&0&0&-4\\0&-2&0&8 \end{bmatrix}$

Moreover, the reduced row-echelon forms of the controllability and observability matrices are given by

$\text{rref}(\mathcal{C})=\begin{bmatrix}1&0&-2&6\\0&1&-3&7\\0&0&0&0\\0&0&0&0 \end{bmatrix} \quad \text{and} \quad \text{rref}(\mathcal{O})=\begin{bmatrix}1&0&0&0\\0&1&0&0\\0&0&0&1\\0&0&0&0 \end{bmatrix}$

A Kalman decomposition of this state space realization with the matrices in the form

$\frac{d}{dt}z(t)=\bar{A}z(t)+\bar{B}u(t), \quad y=\bar{C}z(t), \quad t\in \mathbb{R}^{+}$

with $z(t)=M^{-1}x(t)$ can be obtained using a similarity transformation with the transformation matrix $M$ given by:


$A) \quad M= \begin{bmatrix}0&0&1&0\\0&0&0&1\\1&0&0&0\\0&1&0&0 \end{bmatrix}$

$B) \quad M= \begin{bmatrix}0&0&1&1\\0&1&0&0\\-1&0&0&0\\1&1&1&1 \end{bmatrix}$

$C) \quad M= \begin{bmatrix}0&-1&1&2\\-2&0&1&-2\\1&0&1&2\\0&0&1&-2 \end{bmatrix}$

$D) \quad M= \begin{bmatrix}1&0&2&1\\0&0&-2&1\\2&1&2&1\\1&1&-2&1 \end{bmatrix}$

$E) \quad \text{None of the above}$

I am uncertain which answer is correct.

As far as I know the transformation matrix $M$ can be constructed as follows $M =\begin{bmatrix} V_{r\bar{o}}&V_{ro}&V_{\bar{r}\bar{o}}&V_{\bar{r}o} \end{bmatrix}$

in which the $V$ vectors are determined using the reachable space $\mathcal{R}$ and the unobservable space $\mathcal{UO}$ as follows:

$V_{r\bar{o}} \in \mathcal{R}$ and $V_{r\bar{o}} \in \mathcal{UO}$

$V_{ro} \in \mathcal{R}$ and $V_{ro} \notin \mathcal{UO}$

$V_{\bar{r}\bar{o}} \notin \mathcal{R}$ and $V_{\bar{r}\bar{o}} \in \mathcal{UO} $

$V_{\bar{r}o} \notin \mathcal{R}$ and $V_{\bar{r}\bar{o}} \notin \mathcal{UO}$

in which: $\mathcal{UO} =$ ker$(\mathcal{O}) = \begin{bmatrix}0\\0\\1\\0 \end{bmatrix} \quad$ and $\quad \mathcal{R} = \text{im}(\mathcal{C})= \begin{bmatrix}0&0\\0&0\\-1&1\\1&-2 \end{bmatrix}$

For answers $A$ and $B$ the third column ($V_{\bar{r}\bar{o}}$) is not a member of the set $\mathcal{UO}$. Which leaves only options $C$ and $D$. But as far as I know it is impossible to create the third columns of $C$ and $D$ using only $\mathcal{UO}$ and therefore $C$ and $D$ are also not correct. Thus the correct answer would be $E) \quad \text{None of the above}$.

Is there anyone with knowledge on the Kalman decompositon that can verify this approach?

Thanks in advance.


You have to keep in mind that some of $V_{r\bar{o}}$, $V_{ro}$, $V_{\bar{r}\bar{o}}$ and $V_{\bar{r}o}$ can be empty. The only thing you know for sure is that the combined number of columns need to add up to the state dimension. For example a completely controllable and observable state space model would have $V_{ro}$ span the entire state space.

You do know that the span of $V_{r\bar{o}}$ and $V_{ro}$ should match the span of $\mathcal{R}$, which is of dimension 2. So the first two columns of $M$ should have the same span as $\mathcal{R}$. The only column of $\mathcal{UO}$ lies in the span of $\mathcal{R}$, so $V_{r\bar{o}}$ should have the same span as $\mathcal{UO}$. Since $\mathcal{UO}$ lied completely inside $\mathcal{R}$ means that $V_{\bar{r}\bar{o}}$ has to be empty. This would also imply that $V_{\bar{r}o}$ has two columns.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.