Definition (Linear time invariant control system)
We define a linear time invariant control system to be one of the form { x ˙ ( t ) = A x ( t ) + B u ( t ) x ( t 0 ) = x 0 \begin{cases}
\dot{x}(t)=Ax(t)+Bu(t) \\
x(t_{0})=x_{0}
\end{cases} { x ˙ ( t ) = A x ( t ) + B u ( t ) x ( t 0 ) = x 0 where A ∈ M n ( R ) , B ∈ M n × m ( R ) A\in M_{n}(\mathbb{R}),\ B\in M_{n\times m}(\mathbb{R}) A ∈ M n ( R ) , B ∈ M n × m ( R ) are constant matrices.
Controllability & Observability
Theorem (Rank condition for controllability)
Let A ∈ M n ( R ) , B ∈ M n × m ( R ) A\in M_{n}(\mathbb{R}),B\in M_{n\times m}(\mathbb{R}) A ∈ M n ( R ) , B ∈ M n × m ( R ) . Then, Image ( W ( t 0 , t 1 ) ) = Image ( W T ) = Image ( C A , B ) \text{Image}(W(t_{0},t_{1}))=\text{Image}(W_{T})=\text{Image}(\mathcal{C}_{A,B}) Image ( W ( t 0 , t 1 )) = Image ( W T ) = Image ( C A , B ) where C A , B = [ B , A B , A 2 B , … , A n − 1 B ] \mathcal{C}_{A,B}=[B,AB,A^{2}B, \dots,A^{n-1}B] C A , B = [ B , A B , A 2 B , … , A n − 1 B ] is called the Controllability Matrix for the pair ( A , B ) (A,B) ( A , B ) . We then say an LTIC system with the pair ( A , B ) (A,B) ( A , B ) is controllable for any pair ( t 0 , t 1 ) (t_{0},t_{1}) ( t 0 , t 1 ) if and only if rank ( C A , B ) = n
\text{rank}(\mathcal{C}_{A,B})=n rank ( C A , B ) = n or equivalently: Image ( C A , B ) = R n \text{Image}(\mathcal{C}_{A,B})=\mathbb{R}^{n} Image ( C A , B ) = R n
Theorem (Rank Condition for Observability)
Consider an LTIC system system: { x ˙ ( t ) = A x ( t ) + B u ( t ) y ( t ) = C x ( t ) + D u ( t ) \begin{cases}
\dot{x}(t)=Ax(t)+Bu(t) \\
y(t)=Cx(t)+Du(t)
\end{cases} { x ˙ ( t ) = A x ( t ) + B u ( t ) y ( t ) = C x ( t ) + D u ( t ) then we have that Image ( M ( t 0 , t 1 ) ) = Image ( M T ) Ker ( M ( t 0 , t 1 ) ) = Ker ( M T ) \begin{align*}
\text{Image}(M(t_{0},t_{1}))=\text{Image}(M_{T})\\
\text{Ker}(M(t_{0},t_{1}))=\text{Ker}(M_{T})
\end{align*} Image ( M ( t 0 , t 1 )) = Image ( M T ) Ker ( M ( t 0 , t 1 )) = Ker ( M T ) where M T = [ C C A ⋱ C A n − 1 ] ⊤ [ C C A ⋱ C A n − 1 ] M_{T}=\begin{bmatrix}C \\
CA \\
\ddots \\
CA^{n-1} \end{bmatrix}^{\top} \begin{bmatrix}C \\
CA \\
\ddots \\
CA^{n-1} \end{bmatrix} M T = C C A ⋱ C A n − 1 ⊤ C C A ⋱ C A n − 1 Therefore Image ( M ( t 0 , t 1 ) ) \text{Image}(M(t_{0},t_{1})) Image ( M ( t 0 , t 1 )) and Ker ( M ( t 0 , t 1 ) ) \text{Ker}(M(t_{0},t_{1})) Ker ( M ( t 0 , t 1 )) are independent of t 0 t_{0} t 0 and t 1 t_{1} t 1 . Moreover, Image ( M ( t 0 , t 1 ) ) = Image ( M T ) = Image ( O C A ⊤ ) \text{Image}(M(t_{0},t_{1}))=\text{Image}(M_{T})=\text{Image}(\mathcal{O}_{CA}^{\top}) Image ( M ( t 0 , t 1 )) = Image ( M T ) = Image ( O C A ⊤ ) where O C A ⊤ = [ C C A ⋱ C A n − 1 ] \mathcal{O}^{\top}_{CA}=\begin{bmatrix}C \\
CA \\
\ddots \\
CA^{n-1} \end{bmatrix} O C A ⊤ = C C A ⋱ C A n − 1 is the Observability Matrix . Finally We say the pair ( C , A ) (C,A) ( C , A ) is Observable if and only if rank ( O C A ⊤ ) = n \text{rank}(\mathcal{O}_{CA}^{\top})=n rank ( O C A ⊤ ) = n
Realization
Theorem (Minimial realization of LTIC system)
Suppose that ( A , B , C , D ) (A,B,C,D) ( A , B , C , D ) with A ∈ M n ( R ) A\in M_{n}(\mathbb{R}) A ∈ M n ( R ) is a Realization of some Weighting Pattern . Then ( A , B , C , D ) (A,B,C,D) ( A , B , C , D ) is Minimal if and only if ( A , B ) (A,B) ( A , B ) is Controllable and ( C , A ) (C,A) ( C , A ) is Observable .
Lemma (Non Uniqueness of Realization for LTIC)
Suppose ( A , B , C ) (A,B,C) ( A , B , C ) is a LTIC system Realization of T : J × J → M p × n ( R ) T:J\times J\to M_{p\times n}(\mathbb{R}) T : J × J → M p × n ( R ) . Then ( A ~ , B ~ , C ~ ) (\tilde{A},\tilde{B},\tilde{C}) ( A ~ , B ~ , C ~ ) with A ~ = P A P − 1 B ~ = P B C ~ = C P − 1 \begin{align*}
\tilde{A}&=PAP^{-1}\\
\tilde{B}&=PB\\
\tilde{C}&=CP^{-1}
\end{align*} A ~ B ~ C ~ = P A P − 1 = P B = C P − 1 where P ∈ M n × n ( R ) P\in M_{n\times n}(\mathbb{R}) P ∈ M n × n ( R ) is non singular is also a Realization .