Using the result from where the transformation T, we can construct a state transformation x~=Tx with x~Ė(t)y(t)ā=[A~11ā0āA~12āA~22āā]x~(t)+[B~1ā0ā]u(t)=C~x~(t)āwhere C~=CTā1 and the pair (A~11ā,B~1ā) is Controllable. We call this form the controllability normal form.
Let SāRn be an r-dimensional A-invariant subspace Rn, where AāMnā(R). Then āTāMnā(R), s.t. TATā1=[A~11ā0āA~12āA~22āā]TS=Image([Irā0ā])
Consider the LTIC system {xĖ(t)=Ax(t)+Bu(t)y(t)=Cx(t)āLet CA,Bā=[BāABāā¦āAnā1Bā] be the Controllability Matrix of pair (A,B), and let r be the rank of CA,Bā. Then, āTāMnā(R) s.t. TATā1=[A~11ā0āA~12āA~22āā]TB=[B~1ā0ā]for some A~11āāMnā(R),A~12āāMrĆ(nār)ā(R),A~22āāM(nār)Ć(nār)ā(R),B~1āāMrĆmā(R) and the pair (A~11ā,B~1ā) is Controllable. Moreover, Tā
Image(CA,Bā)=Image([Irā0ā])