Multivariable Control Systems Ali Karimpour Assistant Professor Ferdowsi University of Mashhad Chapter 5 Chapter 5 Controllability, Observability and Realization Topics to be covered include: • Controllability of Linear Dynamical Equations • Observability of Linear Dynamical Equations • Canonical Decomposition of a Linear Time-invariant Dynamical Equation • Realization of Proper Rational Transfer Function Matrices • Irreducible Realizations Irreducible realization of proper rational transfer functions Irreducible Realization of Proper Rational Transfer Function Vectors Irreducible Realization of Proper Rational Matrices 2 Ali Karimpour Sep 2009 Chapter 5 Controllability and Observability of Linear Dynamical Equations Definition 5-1 Definition 5-2 3 Ali Karimpour Sep 2009 Chapter 5 Controllability and Observability of Linear Dynamical Equations Theorem 5-1 4 Ali Karimpour Sep 2009 Chapter 5 Controllability and Observability of Linear Dynamical Equations Theorem 5-2 5 Ali Karimpour Sep 2009 Chapter 5 Controllability and Observability of Linear Dynamical Equations Theorem 5-2(continue) 6 Ali Karimpour Sep 2009 Chapter 5 Canonical Decomposition of a Linear Time-invariant Dynamical Equation x Ax Bu y Cx Eu x PAP 1 x PBu A x B u x Px y CP 1 x Eu C x E u Theorem 5-3 The controllability and observability of a linear time-invariant dynamical equation are invariant under any equivalence transformation. Proof: Let we first consider controllability S B AB PB A 2 B ..... A n1 B PB PAP 1 PB PA2 P 1 PB ..... PAn1 P 1 PB AB A 2 B ..... A n1 B PS Similarly we can consider observability 7 Ali Karimpour Sep 2009 Chapter 5 Canonical Decomposition of a Linear Time-invariant Dynamical Equation Theorem 5-4 x Ax Bu Consider the n-dimensional linear time –invariant dynamical equation y Cx Eu If the controllability matrix of the dynamical equation has rank n1 (where n1<n ), then there exists an equivalence transformation x Px which transform the dynamical equation to xc Ac xc 0 A12 xc Bc u Ac xc 0 y Cc xc Cc Eu xc and the n1-dimensional sub-equation x c Ac xc Bc u y Cc xc Eu 8 is controllable and has the same transfer function matrix as the first system. Ali Karimpour Sep 2009 Chapter 5 Canonical Decomposition of a Linear Time-invariant Dynamical Equation Theorem 5-4 (Continue) Furthermore P=[q1 q2 … qn1 … qn]-1 where q1, q2, …, qn1 be any n1 linearly independent column of S (controllability matrix) and the last n-n1 column of P are entirely arbitrary so long as the matrix [q1 q2 … qn1 … qn] is nonsingular. Proof: See “Linear system theory and design” Chi-Tsong Chen x Ax Bu x Px xc Ac xc Bcu y Cx Eu y Cc xc Eu n dimensiona l n1 n dimensiona l G(s) Hence, we derive the reduced order controllable equation. 9 Ali Karimpour Sep 2009 Chapter 5 Canonical Decomposition of a Linear Time-invariant Dynamical Equation Theorem 5-5 x Ax Bu Consider the n-dimensional linear time –invariant dynamical equation y Cx Eu If the observability matrix of the dynamical equation has rank n2 (where n2<n ), then there exists an equivalence transformation x Px which transform the dynamical equation to xo A0 xo A21 0 xo Bo u Ao xo Bo y C0 xo 0 Eu xo and the n2-dimensional sub-equation x o Ao xo Bo u y Co xo Eu 10 is observable and has the same transfer function matrix as the first system.Ali Karimpour Sep 2009 Chapter 5 Canonical Decomposition of a Linear Time-invariant Dynamical Equation Theorem 5-5 (Continue) Furthermore the first n2 row of P are any n2 linearly independent rows of V (observability matrix) and the last and the last n-n2 row of P is entirely arbitrary so long as the matrix P is nonsingular. Proof: See “Linear system theory and design” Chi-Tsong Chen x Ax Bu x Px y Cx Eu xo Ao xo Bou y Co xo Eu n dimensiona l n2 n dimensiona l G(s) Hence, we derive the reduced order observable equation. 11 Ali Karimpour Sep 2009 Chapter 5 Canonical Decomposition of a Linear Time-invariant Dynamical Equation x Ax Bu Theorem 5-6 (Canonical decomposition theorem) Consider the n-dimensional linear time –invariant dynamical equation y Cx Eu There exists an equivalence transformation x Px which transform the dynamical equation to xco Aco x 0 co xc 0 A12 Aco 0 A13 xco Bco A23 xco Bco u Ac xc 0 y 0 Cco xco Cc xco Eu xc and the reduced dimensional sub-equation x co Aco xco Bco u y Cco xco Eu is observable and controllable and has the same transfer function matrix 12 as the first system. Ali Karimpour Sep 2009 Chapter 5 Canonical Decomposition of a Linear Time-invariant Dynamical Equation Definition 5-3 A linear time-invariant dynamical equation is said to be reducible if and only if there exist a linear time-invariant dynamical equation of lesser dimension that has the same transfer function matrix. Otherwise, the equation is irreducible. Theorem 5-7 A linear time invariant dynamical equation is irreducible if and only if it is controllable and observable. Theorem 5-8 13 Ali Karimpour Sep 2009 Chapter 5 Realization of Proper Rational Transfer Function Matrices Dynamical equation (state-space) description This transformation x Ax Bu The input-output description (transfer function matrix) is unique G(s) C (sI A) 1 B E y Cx Eu The input-output description (transfer function matrix) 1 G(s) C (sI A) B E Realization This transformation is not unique Dynamical equation (state-space) description x Ax Bu y Cx Eu 1. Is it possible at all to obtain the state-space description from the transfer function matrix of a system? 2. If yes, how do we obtain the state space description from the transfer function matrix? 14 Ali Karimpour Sep 2009 Chapter 5 Realization of Proper Rational Transfer Function Matrices Theorem 5-9 A transfer function matrix G(s) is realizable by a finite dimensional linear time invariant dynamical equation if and only if G(s) is a proper rational matrix. Proof: See “Linear system theory and design” Chi-Tsong Chen 15 Ali Karimpour Sep 2009 Chapter 5 Definition 5-4 Irreducible realizations Theorem 5-10 16 Ali Karimpour Sep 2009 Chapter 5 Irreducible realizations Before considering the general case (irreducible realization of proper rational matrices) we start the following parts: 1. Irreducible realization of Proper Rational Transfer Functions 2. Irreducible Realization of Proper Rational Transfer Function Vectors 3. Irreducible Realization of Proper Rational Matrices 17 Ali Karimpour Sep 2009 Chapter 5 Irreducible realization of proper rational transfer functions ˆ0 s n ˆ1 s n 1 ...... ˆ n g (s) , 0 0 ˆ 0 s n ˆ1 s n 1 ...... ˆ n 1 s n 1 2 s n 2 ...... n ˆ0 g ( s) ˆ 0 s n 1 s n 1 ...... n 1s n 1 2 s n 2 ...... n ˆ0 ˆ0 y ( s) u ( s) u ( s) yˆ ( s ) u ( s ) gˆ ( s)u ( s) eu ( s) n n 1 s 1s ...... n ˆ 0 ˆ 0 gˆ ( s ) gˆ ( s) yˆ ( s) gˆ ( s) u ( s) g ( s) y( s) u ( s) yˆ ( s) u ( s) x Ax bu yˆ cx 0u x Ax bu ˆ0 y yˆ u cx eu ˆ 0 18 Ali Karimpour Sep 2009 Chapter 5 Irreducible realization of proper rational transfer functions 1 s n 1 2 s n 2 ...... n ˆ0 g ( s) n n 1 ˆ 0 s 1 s ...... n yˆ ( s) 1s n 1 2 s n2 ...... n gˆ ( s) u ( s) s n 1s n1 ...... n There are different forms of realization Observable canonical form realization Controllable canonical form realization Realization from the Hankel matrix 19 Ali Karimpour Sep 2009 Chapter 5 Observable canonical form realization of proper rational transfer functions 1 s n 1 2 s n 2 ...... n ˆ0 g ( s) ˆ 0 s n 1 s n 1 ...... n 1s n 1 2 s n 2 ...... n gˆ ( s) s n 1s n 1 ...... n y ( n ) 1 y ( n 1) ...... n y 1u ( n 1) 2 u ( n 2) ...... n u xn (t ) y(t ) (1) xn1 (t ) y (1) (t ) 1 y(t ) 1u(t ) xn (t ) 1 xn (t ) 1u(t ) (1) xn2 (t ) y ( 2) (t ) 1 y (1) (t ) 1u (1) (t ) 2 y(t ) 2u(t ) xn1 (t ) 2 xn (t ) 2u(t ) ................................. x1 (t ) y ( n 1) (t ) 1 y ( n 2) (t ) 1u ( n 2 ) (t ) ... n 1 y (t ) n 1u (t ) x2 (t ) n 1 xn (t ) n 1u (t ) (1) ( n) ( n1) (1) ( n 1) x (t ) y (t ) 1 y (t ) 1u (t ) ... n1 y (t ) n1u (1) (t ) n y(t ) nu (t ) n xn (t ) nu (t ) (1) 1 20 Ali Karimpour Sep 2009 Chapter 5 Observable canonical form realization of proper rational transfer functions x n (t ) y (t ) (1) x n 1 (t ) y (1) (t ) 1 y (t ) 1u (t ) x n (t ) 1 x n (t ) 1u (t ) (1) x n 2 (t ) y ( 2 ) (t ) 1 y (1) (t ) 1u (1) (t ) 2 y (t ) 2 u (t ) x n 1 (t ) 2 x n (t ) 2 u (t ) ................................. x1 (t ) y ( n 1) (t ) 1 y ( n 2 ) (t ) 1u ( n 2 ) (t ) ... n 1 y (t ) n 1u (t ) x 2 (t ) n 1 x n (t ) n 1u (t ) (1) (1) x1 (t ) y ( n ) (t ) 1 y ( n1) (t ) 1u ( n1) (t ) ... n1 y (1) (t ) n1u (1) (t ) n y(t ) nu (t ) n xn (t ) nu (t ) x1 0 x 1 2 x3 0 . . xn 0 n x1 n 0 ... 0 n 1 x2 n 1 1 ... 0 n 2 x3 n 2 u . ... . . . . 0 ... 1 1 xn 1 0 ... 0 x1 x 2 yˆ 0 0 ... 0 1 x3 . xn 21 Ali Karimpour Sep 2009 Chapter 5 Observable canonical form realization of proper rational transfer functions 1s n 1 2 s n 2 ...... n gˆ ( s) s n 1s n 1 ...... n x1 0 x 1 2 x3 0 . . xn 0 n x1 n 0 ... 0 n 1 x2 n 1 1 ... 0 n 2 x3 n 2 u . ... . . . . 0 ... 1 1 xn 1 0 ... 0 x1 x 2 yˆ 0 0 ... 0 1 x3 . xn 1s n 1 2 s n 2 ...... n ˆ0 g (s) n n 1 s 1s ...... n ˆ 0 x1 0 x 1 2 x3 0 . . x n 0 n x1 n 0 ... 0 n 1 x2 n 1 1 ... 0 n 2 x3 n 2 u . ... . . . . 0 ... 1 1 xn 1 0 ... 0 x1 x 2 ˆ yˆ 0 0 ... 0 1 x3 0 u ˆ 0 . xn22 Ali Karimpour Sep 2009 Chapter 5 Observable canonical form realization of proper rational transfer functions 1s n 1 2 s n 2 ...... n ˆ0 g (s) s n 1s n 1 ...... n ˆ 0 x1 0 x 1 2 x3 0 . . x n 0 n x1 n 0 ... 0 n 1 x2 n 1 1 ... 0 n 2 x3 n 2 u . ... . . . . 0 ... 1 1 xn 1 0 ... 0 The derived dynamical equation is observable. 1s n 1 2 s n 2 ...... n gˆ ( s) s n 1s n 1 ...... n x1 x 2 ˆ yˆ 0 0 ... 0 1 x3 0 u ˆ 0 . xn Exersise 1: Why? The derived dynamical equation controllable as well if numerator and denominator of g(s) are coprime. Exersise 2: Why? 23 Ali Karimpour Sep 2009 Chapter 5 Controllable canonical form realization of proper rational transfer functions 1 s n 1 2 s n 2 ...... n ˆ0 g ( s) n n 1 ˆ 0 s 1 s ...... n Let us introduce a new variable N ( s) 1 s n 1 2 s n 2 ...... n g ( s) D( s ) s n 1 s n 1 ...... n D ( s )v ( s ) u ( s ) y ( s ) N ( s )v ( s ) We may define the state variable as: Clearly x1 (t ) v(t ) x (t ) (1) v ( t ) 2 x(t ) .. .. .. .. xn (t ) v ( n 1) (t ) x1 x2 , x 2 x3 , ..... , x n1 xn xn v ( n ) (t ) u (t ) n v(t ) n1v (1) (t ) ... 1v ( n1) (t ) u(t ) n x1 (t ) n1 x2 (t ) ... 1 xn (t ) 24 Ali Karimpour Sep 2009 Chapter 5 Controllable canonical form realization of proper rational transfer functions g ( s) 1 s 2 s ...... n 0 n n 1 0 s 1 s ...... n n 1 n2 N ( s) 1 s n 1 2 s n 2 ...... n g ( s) D( s ) s n 1 s n 1 ...... n x1 x2 , x 2 x3 , ..... , x n1 xn xn v ( n ) (t ) u (t ) n v(t ) n1v (1) (t ) ... 1v ( n1) (t ) u(t ) n x1 (t ) n1 x2 (t ) ... 1 xn (t ) x1 0 x 0 2 x3 . . 0 xn n 1 0 ... 0 1 ... . . ... 0 0 ... n 1 n 2 ... 0 x1 0 0 x2 0 . x3 0u 1 . . 1 xn 1 y n n 1 n 2 x1 x 2 ... 1 x3 . xn 25 Ali Karimpour Sep 2009 Chapter 5 Controllable canonical form realization of proper rational transfer functions N ( s) 1 s n 1 2 s n 2 ...... n g ( s) D( s ) s n 1 s n 1 ...... n x1 0 x 0 2 x3 . . 0 xn n 1 0 ... 0 1 ... . . ... 0 0 ... n 1 n 2 ... 0 x1 0 0 x2 0 . x3 0u 1 . . 1 xn 1 y n n 1 n 2 x1 x 2 ... 1 x3 . xn 1s n 1 2 s n 2 ...... n ˆ0 g (s) s n 1s n 1 ...... n ˆ 0 x1 0 x 0 2 x3 . . 0 xn n 1 0 ... 0 1 ... . . ... 0 0 ... n 1 n 2 ... 0 x1 0 0 x2 0 . x3 0u 1 . . 1 xn 1 y n n 1 n 2 x1 x 2 ˆ ... 1 x3 0 u ˆ 0 . xn 26 Ali Karimpour Sep 2009 Chapter 5 Controllable canonical form realization of proper rational transfer functions 1s n 1 2 s n 2 ...... n ˆ0 g (s) n n 1 s 1s ...... n ˆ 0 x1 0 x 0 2 x3 . . 0 xn n 1 0 ... 0 1 ... . . ... 0 0 ... n 1 n 2 ... 0 x1 0 0 x2 0 . x3 0u 1 . . 1 xn 1 The derived dynamical equation is controllable . N ( s) 1 s n 1 2 s n 2 ...... n g ( s) D( s ) s n 1 s n 1 ...... n y n n 1 n 2 x1 x 2 ˆ ... 1 x3 0 u ˆ 0 . xn Exersise 3: Why? The derived dynamical equation observable as well if numerator and denominator of g(s) are coprime. Exersise 4: Why? 27 Ali Karimpour Sep 2009 Chapter 5 Controllable and observable canonical form realization of proper rational transfer functions Example 5-2 Derive controllable and observable canonical realization for following system. 2s 3 18s 2 48s 32 g ( s) s 3 6s 2 11s 6 2s 3 18s 2 48s 32 6s 2 26s 20 g ( s) 3 2 3 2 2 s 6s 11s 6 s 6s 11s 6 Observable canonical form realization is: x1 0 0 6 x1 20 x 1 0 11 x 26u 2 2 x 3 0 1 6 x3 6 x1 y 0 0 1 x 2 2u x3 It is not controllable. Why? Controllable canonical form realization is: 1 0 x1 0 x1 0 x 0 x 0 u 0 1 2 2 x3 6 11 6 x3 1 x1 y 20 26 6 x2 2u x3 It is not observable. 28 Why? Ali Karimpour Sep 2009 Chapter 5 Irreducible realization of proper rational transfer functions Example 5-3 Derive irreducible realization for following transfer function. 2s 3 18s 2 48s 32 g ( s) s 3 6s 2 11s 6 6 s 20 2s 3 18s 2 48s 32 6s 2 26s 20 2 g ( s) 2 2 3 2 3 2 s 5s 6 s 6s 11s 6 s 6s 11s 6 Observable canonical form realization is: x1 0 6 x1 20 x 1 5 x 6 u 2 2 x1 y 0 1 2u x2 It is controllable too. Why? Controllable canonical form realization is: 1 x1 0 x1 0 x 6 5 x 1 u 2 2 x1 y 20 6 2u x2 It is observable too. Why? 29 Ali Karimpour Sep 2009 Chapter 5 Irreducible realization of proper rational transfer functions Realization from the Hankel matrix 0 s n 1s n 1 ...... n g ( s) n s 1s n 1 ...... n g (s) h(0) h(1)s 1 h(2)s 2 h(3) s 3 ...... The coefficients h(i) will be called Markov parameters. h(2) h(3) h(1) h(2) h(3) h(4) h(3) h(4) h(5) H ( , ) . . . . . . h( ) h( 1) h( 2) ... ... ... ... ... ... h( ) h( 1) h( 2) . . h( 1) 30 Ali Karimpour Sep 2009 Chapter 5 Irreducible realization of proper rational transfer functions Realization from the Hankel matrix Theorem 5-11 Consider the proper transfer function g(s) as 0 s n 1 s n 1 2 s n 2 ...... n g ( s) s n 1 s n 1 ...... n then g(s) has degree m if and only if H (m, m) H (m k , m l ) for every k , l 1, 2 , 3, .... 31 Ali Karimpour Sep 2009 Chapter 5 Irreducible realization of proper rational transfer functions Realization from the Hankel matrix Now consider the dynamical equation g (s) c(sI A) 1 b e s 1c( I s 1 A) 1 b e x Ax bu y cx eu e cbs 1 cAbs 2 cA2bs 3 ..... h(i) cAi 1b h(2) h(1) h(2) . . H (n 1, n) . . h( n) h(n 1) h(n 1) h(n 2) ..... ..... ..... ..... ..... ..... i 1, 2 , 3 , ...... h( n) h(n 1) . . h(2n 1) h(2n) Let the first σ rows be linearly independent and the (σ+1) th row of H(n+1,n) be linearly dependent on its previous rows. So [a1 a2 ..... a 1 0 ..... 0] H (n 1, n) 0 32 Ali Karimpour Sep 2009 Chapter 5 Irreducible realization of proper rational transfer functions Realization from the Hankel matrix [a1 a2 1 0 ..... 0] H (n 1, n) 0 ..... a We claim that the σ-dimensional dynamical equation 1 0 0 0 0 1 0 0 0 x . . . . . . . 0 0 0 a a a 2 3 1 y 1 0 0 ..... 0 ..... 0 ..... 0 ..... 0 ..... . ..... . ..... 0 ..... a 1 0x h(0)u 0 h(1) 0 h(2) 0 h(3) . x . u . . 1 h( 1) a h( ) (I) is a controllable and observable (irreducible realization). Exercise 5: Show that (I) is a controllable and observable (irreducible realization) of x Ax bu y cx eu 33 Ali Karimpour Sep 2009 Chapter 5 Irreducible realization of proper rational transfer functions Example 5-4 Derive irreducible realization for following transfer function. 2s 3 18s 2 48s 32 g ( s) s 3 6s 2 11s 6 g (s) 2 6s 1 10s 2 14s 3 10s 4 34s 5 230s 6 ..... 10 14 6 10 14 10 H (4,3) 14 10 34 10 34 230 We can show that the rank of H(4,3) is 2. So 6 5 1 0H (4,3) 0 Hence an irreducible realization of g(s) is 1 0 6 x x 10u 6 5 y 1 0x 2u 34 Ali Karimpour Sep 2009 Chapter 5 Realization of Proper Rational Transfer Function Vectors Consider the rational function vector g1 ( s ) g (s) 2 G (s) . . g q ( s) e g 1 1 (s) e g ( s ) 2 2 G ( s) . . . . e q g q ( s ) e1 e 2 1 . G(s) n s 1 s n 1 ..... n . e q 11 s n 1 12 s n 2 .... 1n n 1 n2 s s .... 22 2n 21 . . s n 1 s n 2 .... q2 qn q1 35 Ali Karimpour Sep 2009 Chapter 5 Realization of Proper Rational Transfer Function Vectors e1 e 2 1 G(s) . n s 1 s n 1 ..... n . e q 0 0 . x . 0 n 1 0 ... 0 1 ... . . ... . . ... 0 0 ... n 1 n 2 ... 0 0 0 0 . . x u . . 1 0 1 1 11 s n 1 12 s n 2 .... 1n n2 n 1 .... s s 2n 22 21 . . 2 n 1 n s s .... qn q2 q1 y1 1n y 2 2n . . . . yq qn 1( n 1) 2( n 1) 1( n 2) ... 11 2( n 2) ... 21 . . . . q ( n 1) q ( n2) e1 e 2 ... . x . u ... . . ... q1 eq This is a controllable form realization of G(s). We see that the transfer function from u to yi is equal to i1 s n 1 i 2 s n 2 .... in ei 36 n n 1 s 1 s ..... AlinKarimpour Sep 2009 Chapter 5 Realization of Proper Rational Transfer Function Vectors Example 5-5 Derive a realization for following transfer function vector. s3 G ( s ) ( s 1)( s 2) s4 s3 s3 ( s 1)( s 2) 0 ( s 3) 2 1 G (s) s 4 1 ( s 1)( s 2)( s 3) ( s 1)( s 2) s3 s 2 6 s 9 0 1 3 2 2 1 s 6 s 11s 6 s 3s 2 Hence a minimal dimensional realization of G(s) is given by 1 0 0 0 x 0 0 1 x 0u 6 11 6 1 9 6 1 0 y x u 2 3 1 1 37 Ali Karimpour Sep 2009 Chapter 5 Realization of Proper Rational Matrices There are many approaches to find irreducible realizations for proper rational matrices. 1. One approach is to first find a reducible realization and then apply the reduction procedure to reduce it to an irreducible one. Method I, Method II, Method III and Method IV 2. In the second approach irreducible realization will yield directly. 38 Ali Karimpour Sep 2009 Chapter 5 Realization of Proper Rational Matrices Method I: Given a proper rational matrix G(s), if we first find xij Aij xij biju j an irreducible realization for every element gij(s) of G(s) as yij cij xi j eiju j y1 y11 y12 y2 y22 y22 x11 A11 x 0 12 x 21 0 x 22 0 y1 c11 y 0 2 0 x11 b11 0 A12 0 0 x12 0 b12 u1 b21 0 u 2 0 A21 0 x 21 0 0 A22 x 22 0 b22 x11 c12 0 0 x12 e11 e12 u1 0 c 21 c 22 x 21 e21 e22 u 2 x 22 0 0 Clearly this equation is generally not controllable and not observable. To reduce this realization to irreducible one requires the application of the reduction procedure twice (theorems 5-4 and 5-5). 39 Ali Karimpour Sep 2009 Chapter 5 Realization of Proper Rational Matrices x11 A11 x 0 12 x 21 0 x 22 0 y1 c11 y 0 2 Proof: c11 c12 0 0 0 c 21 ( sI 0 c 22 A11 ) 1 0 x11 b11 0 A12 0 0 x12 0 b12 u1 b21 0 u 2 0 A21 0 x 21 0 0 A22 x 22 0 b22 x11 c12 0 0 x12 e11 e12 u1 0 c 21 c 22 x 21 e21 e22 u 2 x 22 0 0 0 0 0 ( sI A12 ) 1 0 0 0 ( sI A21 ) 1 0 0 0 b11 0 e 0 b 0 12 11 e12 b21 0 e21 e22 0 ( sI A22 ) 1 0 b22 0 c11 ( sI A11 ) 1 b11 e11 c12 ( sI A12 ) 1 b12 e12 g11 ( s) g12 ( s ) G(s) 1 1 c 21 ( sI A21 ) b21 e21 c 22 ( sI A22 ) b22 e22 g 21 ( s ) g 22 ( s) 40 Ali Karimpour Sep 2009 Chapter 5 Realization of Proper Rational Matrices Method II: Given a proper rational matrix G(s), if we find the controllable canonicalform realization for the ith column, Gi(s), of G(s) say, x i Ai xi bi u i x1 A1 x 0 2 .... . x p 0 y C1 C 2 Proof: C 1 C2 0 .... A2 .... . .... 0 .... .... C p ( sI .... C p C1 ( sI A1 ) 1 b1 e1 0 x1 b1 0 x 2 0 . .. . A p x p 0 x e1 e2 .... 0 u1 b2 ... 0 u 2 . ... . . 0 ... b p u p ep u 0 ... y i Ci xi ei u i This realization is always controllable. It is however generally not observable. b1 0 ... 0 0 b ... 0 0 ( sI A2 ) 1 ... 0 2 e e .... e 1 2 p . . ... . . . ... . 0 0 ... ( sI A p ) 1 0 0 ... b p g11 ( s ) g12 ( s ) ... g1 p ( s ) g ( s ) g ( s ) ... g ( s ) 21 22 2p 1 G ( s) .....C p ( sI A p ) b p e p . . ... . 41 g q1 ( s ) g q 2 ( s ) ... g qp ( s ) A1 ) 1 0 ... 0 Ali Karimpour Sep 2009 Chapter 5 Realization of Proper Rational Matrices Method III: Let a proper rational matrix G(s), where G ( s ) G ( s ) G ( ) consider the monic least common denominator of G(s) as ( s) s m 1 s m1 2 s m2 ... m Then we can write G(s) as G( s) 1 R1s m1 R2 s m2 ... Rm G() ( s) Then the following dynamic equation is a realization of G(s). Ip 0p ... 0p 0p 0 p 0 0 0 I ... 0 p p p p p . . ... . x .. u x . 0 0 0 ... I p p p p 0 p m I p m 1 I p m 2 I p ... 1 I p I p y Rm Rm 1 Rm 2 ... R1 x G ()u Exercise 6: Show that the above dynamical equation is a controllable realization42of G(s) Ali Karimpour Sep 2009 Chapter 5 Realization of Proper Rational Matrices Method IV: It is possible to obtain observable realization of a proper G(s). Let G(s) H (0) H (1)s 1 H (2)s 2 ........... Consider the monic least common denominator of G(s) as ( s) s m 1 s m1 2 s m2 ... m Then after deriving H(i) one can simply show H (m i) 1H (m i 1) 2 H (m i 2) ... m H (i) i 1 (I ) Exercise 7: Proof equation (I) Let {A, B, C and E} be a realization of G(s) then we have G(s) E C (sI A) 1 B E CBs 1 CABs 2 CA2 Bs 3 ........... 43 Ali Karimpour Sep 2009 Chapter 5 Realization of Proper Rational Matrices Then {A, B, C and D} be a realization of G(s) if and only if E H (0) and H (i 1) CAi B i 0 , 1, 2 ,.... Now we claim that the following dynamical equation is a realization of G(s). Iq 0q 0q 0 0q Iq q . . x . 0q 0q 0q m I q m 1 I q m 2 I q y I q 0 0 ... 0 x H (0)u ... ... ... ... ... 0q H (1) H (2) 0 q . x u .. Iq H ( m 1 ) H (m) 1 I q We can readily verify that H (2) H (3) H (i 1) H (3) H (4) H (i 2) 2 i , A B , ... , A B AB ... ... ... H (m 1) H (m 2) H (m i ) CAi B H (i 1) E H (0) 44 Ali Karimpour Sep 2009 Chapter 5 Realization of Proper Rational Matrices Now we shall discuss in the following a method which will yield directly irreducible realizations. This method is based on the Hankel matrices. q p Let G(s) be G(s) H (0) H (1)s 1 H (2)s 2 ........... Consider the monic least common denominator of G(s) as ( s) s m 1 s m1 2 s m2 ... m Define 0 p I p N 0 p . 0 p 0p ... 0p ... Ip ... . ... 0p ... mI p 0 p m 1 I p ..0 p . m 2 I p ... . Ip 1 I p 0p 0q 0 q M . 0q m I q Iq 0q ... 0q Iq ... . . ... 0q 0q ... m 1 I q m2 I q ... 0q 0 q . Iq 1 I q 45 Ali Karimpour Sep 2009 Chapter 5 Realization of Proper Rational Matrices 0 p I p N 0 p . 0 p 0p 0p Ip ( s) s m 1 s m1 2 s m2 ... m Iq 0q ... 0 p mI p 0 0q ... 0 p m 1 I p q . M . ... ..0 p . m 2 I p . ... ... . 0p ... Ip 1 I p 0q m I q 0q ... Iq ... . ... 0q 0q ... m 1 I q m2 I q ... 0q 0 q . Iq 1 I q We also define the two following Hankel matrices H (2) H (1) H (2) H (3) T . . H (m) H (m 1) H ( m) H (m 1) . H (2m 1) H (3) H (2) H (4) ~ H (3) T . . H (m 1) H (m 2) H (m 1) H (m 2) . H ( 2 m) It can be readily verified that ~ T MT TN M i T TN i i 0 , 1, 2 ,..... 46 Ali Karimpour Sep 2009 Chapter 5 Realization of Proper Rational Matrices It can be readily verified that ~ T MT TN Let I k ,l k l (l k ) M i T TN i i 0 , 1, 2 ,..... be as the form I k ,l I k 0 Note that the left-upper-corner of M iT = TN i is H(i+1) so: H (i 1) I q,qm M iTI Tp, pm I q,qmTN i I Tp, pm i 0 , 1, 2 , .... It can be readily verified that But we want Irreducible Realization of Proper Rational Matrices 47 Ali Karimpour Sep 2009 Chapter 5 Irreducible Realization of Proper Rational Matrices 48 Ali Karimpour Sep 2009 Chapter 5 Irreducible Realization of Proper Rational Matrices Example 5-6 Derive an irreducible realization for the following proper rational function. 2s 2 3s 2 2 ( s 1 ) G( s) 4s 5 s 1 Least common denominator of G(s), is 1 s 3s 5 s 1 ( s) s( s 1) 2 2 0 1 1 1 2 0 2 3 0 3 4 0 4 5 0 5 6 0 6 G( s) 1 2 s 1 2 s 1 2 s 1 2 s .1 2 s 1 2 s ..... 4 3 1 2 0 3 0 1 1 2 1 2 1 2 H (1) H (2) H (3) 2 0 3 0 4 0 T H (2) H (3) H (4) 1 2 1 2 1 2 H (3) H (4) H (5) 3 0 4 0 5 0 1 2 1 2 1 2 Non-zero singular values of T are 10.23, 5.79, 0.90 and 0.23. So, r = 4. 49 Ali Karimpour Sep 2009 Chapter 5 Irreducible Realization of Proper Rational Matrices - 0.3413 0.2545 - 0.8902 - 0.1621 - 0.2357 - 0.5238 - 0.0581 - 0.0071 0.5127 - 0.2078 - 0.1054 - 0.8264 Yr 0.2357 05238 0 . 0581 0 . 0071 - 0.6738 0.2627 0.4316 0.5392 - 0.2357 - 0.5238 0.0581 0.0071 Yˆ Yr S 1 / 2 - 0.4003 - 0.0196 0.4905 - 0.6574 0.1049 0.5872 0.6022 0.5306 0.5496 - 0.1057 - 0.0978 0.1026 Ur 0.1382 0.5432 0 . 3875 0.1888 - 0.6989 0.2311 - 0.2949 0.4522 0.5432 0.3875 0.1888 0.1382 H Uˆ S 1 / 2U r - 1.0915 0.6121 - 0.8443 - 0.0770 - 0.7539 - 1.2598 - 0.0551 - 0.0034 1.6398 - 0.4999 - 0.1000 - 0.3923 0.7539 1.2598 0.0551 0.0034 - 2.1553 0.6317 0.4093 - 0.2560 - 0.7539 - 1.2598 - 0.0551 - 0.0034 - 1.2803 0.3355 1.7579 - 0.4421 - 2.2356 - 0.0471 1.4124 - 0.2543 - 1.3066 0.5557 0.4652 - 0.5711 - 0.0927 - 0.3675 - 0.2797 - 0.3121 - 0.2519 0.0487 - 0.0896 0.2147 0.4421 1.3066 0.3675 0.0896 50 Ali Karimpour Sep 2009 Chapter 5 Irreducible Realization of Proper Rational Matrices - 0.1067 - 0.0737 0.1058 - 0.2178 H † 1 / 2 Yˆ S Yr - 0.9386 - 0.0613 - 0.3415 - 0.0149 - 0.1251 0.0328 0.1718 Uˆ † U r S 1 / 2 - 0.0432 - 0.2185 0.0432 0.1603 0.0737 - 0.2107 - 0.0737 - 0.0864 0.2178 0.1092 - 0.2178 - 0.1112 0.0613 0.4551 - 0.0613 - 1.7406 0.0149 - 1.1356 - 0.0149 - 0.0081 0.5171 - 1.3847 0.2441 - 0.6349 - 1.1176 - 0.0440 - 0.1031 0.2161 - 0.2259 - 0.4086 - 0.3976 0.0961 - 0.3109 0.9525 0.2259 0.4086 0.3976 - 1.2497 0.0369 0.2155 - 0.1904 0.1588 - 1.0139 - 0.1604 0.0772 †~ ˆ† ˆ A Y TU - 0.2227 - 0.1800 - 0.2888 0.8076 0.1246 - 0.1181 0.1354 - 0.4476 - 1.0915 0.6121 - 0.8443 - 0.0770 ˆ C I q ,qmY - 0.7539 - 1.2598 - 0.0551 - 0.0034 - 1.2803 0.3355 - 0.0471 1.4124 B UˆI Tp , pm 0.4652 - 0.5711 - 0.3121 - 0.2519 2 0 E H (0) 4 3 51 Ali Karimpour Sep 2009