**Purpose**

To construct for a linear multivariable system described by a state-space model (A,B,C,D) a regular pencil (A - lambda*B ) which f f has the invariant zeros of the system as generalized eigenvalues. The routine also computes the orders of the infinite zeros and the right and left Kronecker indices of the system (A,B,C,D).

SUBROUTINE AB08NZ( EQUIL, N, M, P, A, LDA, B, LDB, C, LDC, D, LDD, $ NU, RANK, DINFZ, NKROR, NKROL, INFZ, KRONR, $ KRONL, AF, LDAF, BF, LDBF, TOL, IWORK, DWORK, $ ZWORK, LZWORK, INFO ) C .. Scalar Arguments .. CHARACTER EQUIL INTEGER DINFZ, INFO, LDA, LDAF, LDB, LDBF, LDC, LDD, $ LZWORK, M, N, NKROL, NKROR, NU, P, RANK DOUBLE PRECISION TOL C .. Array Arguments .. INTEGER INFZ(*), IWORK(*), KRONL(*), KRONR(*) COMPLEX*16 A(LDA,*), AF(LDAF,*), B(LDB,*), BF(LDBF,*), $ C(LDC,*), D(LDD,*), ZWORK(*) DOUBLE PRECISION DWORK(*)

**Mode Parameters**

EQUIL CHARACTER*1 Specifies whether the user wishes to balance the compound matrix (see METHOD) as follows: = 'S': Perform balancing (scaling); = 'N': Do not perform balancing.

N (input) INTEGER The number of state variables, i.e., the order of the matrix A. N >= 0. M (input) INTEGER The number of system inputs. M >= 0. P (input) INTEGER The number of system outputs. P >= 0. A (input) COMPLEX*16 array, dimension (LDA,N) The leading N-by-N part of this array must contain the state dynamics matrix A of the system. LDA INTEGER The leading dimension of array A. LDA >= MAX(1,N). B (input) COMPLEX*16 array, dimension (LDB,M) The leading N-by-M part of this array must contain the input/state matrix B of the system. LDB INTEGER The leading dimension of array B. LDB >= MAX(1,N). C (input) COMPLEX*16 array, dimension (LDC,N) The leading P-by-N part of this array must contain the state/output matrix C of the system. LDC INTEGER The leading dimension of array C. LDC >= MAX(1,P). D (input) COMPLEX*16 array, dimension (LDD,M) The leading P-by-M part of this array must contain the direct transmission matrix D of the system. LDD INTEGER The leading dimension of array D. LDD >= MAX(1,P). NU (output) INTEGER The number of (finite) invariant zeros. RANK (output) INTEGER The normal rank of the transfer function matrix. DINFZ (output) INTEGER The maximum degree of infinite elementary divisors. NKROR (output) INTEGER The number of right Kronecker indices. NKROL (output) INTEGER The number of left Kronecker indices. INFZ (output) INTEGER array, dimension (N) The leading DINFZ elements of INFZ contain information on the infinite elementary divisors as follows: the system has INFZ(i) infinite elementary divisors of degree i, where i = 1,2,...,DINFZ. KRONR (output) INTEGER array, dimension (MAX(N,M)+1) The leading NKROR elements of this array contain the right Kronecker (column) indices. KRONL (output) INTEGER array, dimension (MAX(N,P)+1) The leading NKROL elements of this array contain the left Kronecker (row) indices. AF (output) COMPLEX*16 array, dimension (LDAF,N+MIN(P,M)) The leading NU-by-NU part of this array contains the coefficient matrix A of the reduced pencil. The remainder f of the leading (N+M)-by-(N+MIN(P,M)) part is used as internal workspace. LDAF INTEGER The leading dimension of array AF. LDAF >= MAX(1,N+M). BF (output) COMPLEX*16 array, dimension (LDBF,N+M) The leading NU-by-NU part of this array contains the coefficient matrix B of the reduced pencil. The f remainder of the leading (N+P)-by-(N+M) part is used as internal workspace. LDBF INTEGER The leading dimension of array BF. LDBF >= MAX(1,N+P).

TOL DOUBLE PRECISION A tolerance used in rank decisions to determine the effective rank, which is defined as the order of the largest leading (or trailing) triangular submatrix in the QR (or RQ) factorization with column (or row) pivoting whose estimated condition number is less than 1/TOL. If the user sets TOL to be less than SQRT((N+P)*(N+M))*EPS then the tolerance is taken as SQRT((N+P)*(N+M))*EPS, where EPS is the machine precision (see LAPACK Library Routine DLAMCH).

IWORK INTEGER array, dimension (MAX(M,P)) DWORK DOUBLE PRECISION array, dimension (MAX(N,2*MAX(P,M))) ZWORK DOUBLE PRECISION array, dimension (LZWORK) On exit, if INFO = 0, ZWORK(1) returns the optimal value of LZWORK. LZWORK INTEGER The length of the array ZWORK. LZWORK >= MAX( 1, MIN(P,M) + MAX(3*M-1,N), MIN(P,N) + MAX(3*P-1,N+P,N+M), MIN(M,N) + MAX(3*M-1,N+M) ). An upper bound is MAX(s,N) + MAX(3*s-1,N+s), with s = MAX(M,P). For optimum performance LZWORK should be larger. If LZWORK = -1, then a workspace query is assumed; the routine only calculates the optimal size of the ZWORK array, returns this value as the first entry of the ZWORK array, and no error message related to LZWORK is issued by XERBLA.

INFO INTEGER = 0: successful exit; < 0: if INFO = -i, the i-th argument had an illegal value.

The routine extracts from the system matrix of a state-space system (A,B,C,D) a regular pencil A - lambda*B which has the f f invariant zeros of the system as generalized eigenvalues as follows: (a) construct the (N+P)-by-(N+M) compound matrix (B A); (D C) (b) reduce the above system to one with the same invariant zeros and with D of full row rank; (c) pertranspose the system; (d) reduce the system to one with the same invariant zeros and with D square invertible; (e) perform a unitary transformation on the columns of (A - lambda*I B) in order to reduce it to ( C D) (A - lambda*B X) ( f f ), with Y and B square invertible; ( 0 Y) f (f) compute the right and left Kronecker indices of the system (A,B,C,D), which together with the orders of the infinite zeros (determined by steps (a) - (e)) constitute the complete set of structural invariants under strict equivalence transformations of a linear system.

[1] Svaricek, F. Computation of the Structural Invariants of Linear Multivariable Systems with an Extended Version of the Program ZEROS. System & Control Letters, 6, pp. 261-266, 1985. [2] Emami-Naeini, A. and Van Dooren, P. Computation of Zeros of Linear Multivariable Systems. Automatica, 18, pp. 415-430, 1982.

The algorithm is backward stable (see [2] and [1]).

In order to compute the invariant zeros of the system explicitly, a call to this routine may be followed by a call to the LAPACK Library routine ZGGEV with A = A , B = B and N = NU. f f If RANK = 0, the routine ZGEEV can be used (since B = I). f

**Program Text**

* AB08NZ EXAMPLE PROGRAM TEXT * Copyright (c) 2002-2017 NICONET e.V. * * .. Parameters .. DOUBLE PRECISION ZERO PARAMETER ( ZERO = 0.0D0 ) INTEGER NIN, NOUT PARAMETER ( NIN = 5, NOUT = 6 ) INTEGER NMAX, MMAX, PMAX PARAMETER ( NMAX = 10, MMAX = 10, PMAX = 10 ) INTEGER MPMAX PARAMETER ( MPMAX = MAX( MMAX, PMAX ) ) INTEGER LDA, LDB, LDC, LDD, LDAF, LDBF, LDQ, LDZ PARAMETER ( LDA = NMAX, LDB = NMAX, LDC = PMAX, $ LDD = PMAX, LDAF = NMAX+MPMAX, $ LDBF = NMAX+PMAX, LDQ = 1, LDZ = 1 ) INTEGER LDWORK PARAMETER ( LDWORK = 8*NMAX ) INTEGER LZWORK PARAMETER ( LZWORK = $ MAX( MIN( PMAX, MMAX ) + $ MAX( 3*MMAX - 1, NMAX ), $ MIN( PMAX, NMAX ) + $ MAX( 3*PMAX, NMAX+PMAX, NMAX+MMAX ), $ MIN( MMAX, NMAX ) + $ MAX( 3*MMAX, NMAX+MMAX ), 1 ) ) * .. Local Scalars .. DOUBLE PRECISION TOL INTEGER DINFZ, I, INFO, J, M, N, NINFZ, NKROL, NKROR, $ NU, P, RANK CHARACTER*1 EQUIL * .. Local Arrays .. COMPLEX*16 A(LDA,NMAX), AF(LDAF,NMAX+PMAX), ALPHA(NMAX), $ B(LDB,MMAX), BETA(NMAX), BF(LDBF,MMAX+NMAX), $ C(LDC,NMAX), D(LDD,MMAX), Q(LDQ,1), Z(LDZ,1), $ ZWORK(LZWORK) DOUBLE PRECISION DWORK(LDWORK) INTEGER INFZ(NMAX), IWORK(MPMAX+1), KRONL(NMAX+1), $ KRONR(NMAX+1) * .. External Subroutines .. EXTERNAL AB08NZ, ZGEGV * .. Intrinsic Functions .. INTRINSIC MAX * .. Executable Statements .. * WRITE ( NOUT, FMT = 99999 ) * Skip the heading in the data file and read the data. READ ( NIN, FMT = '()' ) READ ( NIN, FMT = * ) N, M, P, TOL, EQUIL IF ( N.LT.0 .OR. N.GT.NMAX ) THEN WRITE ( NOUT, FMT = 99972 ) N ELSE READ ( NIN, FMT = * ) ( ( A(I,J), J = 1,N ), I = 1,N ) IF ( M.LT.0 .OR. M.GT.MMAX ) THEN WRITE ( NOUT, FMT = 99971 ) M ELSE READ ( NIN, FMT = * ) ( ( B(I,J), J = 1,M ), I = 1,N ) IF ( P.LT.0 .OR. P.GT.PMAX ) THEN WRITE ( NOUT, FMT = 99970 ) P ELSE READ ( NIN, FMT = * ) ( ( C(I,J), J = 1,N ), I = 1,P ) READ ( NIN, FMT = * ) ( ( D(I,J), J = 1,M ), I = 1,P ) * Check the observability and compute the ordered set of * the observability indices (call the routine with M = 0). CALL AB08NZ( EQUIL, N, 0, P, A, LDA, B, LDB, C, LDC, D, $ LDD, NU, RANK, DINFZ, NKROR, NKROL, INFZ, $ KRONR, KRONL, AF, LDAF, BF, LDBF, TOL, $ IWORK, DWORK, ZWORK, LZWORK, INFO ) * IF ( INFO.NE.0 ) THEN WRITE ( NOUT, FMT = 99998 ) INFO ELSE WRITE ( NOUT, FMT = 99994 ) ( KRONL(I), I = 1,P ) IF ( NU.EQ.0 ) THEN WRITE ( NOUT, FMT = 99993 ) ELSE WRITE ( NOUT, FMT = 99992 ) N - NU WRITE ( NOUT, FMT = 99991 ) WRITE ( NOUT, FMT = 99990 ) DO 20 I = 1, NU WRITE ( NOUT, FMT = 99989 ) $ ( AF(I,J), J = 1,NU ) 20 CONTINUE END IF END IF * Check the controllability and compute the ordered set of * the controllability indices (call the routine with P = 0) CALL AB08NZ( EQUIL, N, M, 0, A, LDA, B, LDB, C, LDC, D, $ LDD, NU, RANK, DINFZ, NKROR, NKROL, INFZ, $ KRONR, KRONL, AF, LDAF, BF, LDBF, TOL, $ IWORK, DWORK, ZWORK, LZWORK, INFO ) * IF ( INFO.NE.0 ) THEN WRITE ( NOUT, FMT = 99998 ) INFO ELSE WRITE ( NOUT, FMT = 99988 ) ( KRONR(I), I = 1,M ) IF ( NU.EQ.0 ) THEN WRITE ( NOUT, FMT = 99987 ) ELSE WRITE ( NOUT, FMT = 99986 ) N - NU WRITE ( NOUT, FMT = 99985 ) WRITE ( NOUT, FMT = 99990 ) DO 40 I = 1, NU WRITE ( NOUT, FMT = 99989 ) $ ( AF(I,J), J = 1,NU ) 40 CONTINUE END IF END IF * Compute the structural invariants of the given system. CALL AB08NZ( EQUIL, N, M, P, A, LDA, B, LDB, C, LDC, D, $ LDD, NU, RANK, DINFZ, NKROR, NKROL, INFZ, $ KRONR, KRONL, AF, LDAF, BF, LDBF, TOL, $ IWORK, DWORK, ZWORK, LZWORK, INFO ) * IF ( INFO.NE.0 ) THEN WRITE ( NOUT, FMT = 99998 ) INFO ELSE WRITE ( NOUT, FMT = 99984 ) NU IF ( NU.GT.0 ) THEN * Compute the invariant zeros of the given system. * Complex Workspace: need 2*NU. * Real Workspace: need 8*NU. WRITE ( NOUT, FMT = 99983 ) CALL ZGEGV( 'No vectors', 'No vectors', NU, AF, $ LDAF, BF, LDBF, ALPHA, BETA, Q, LDQ, $ Z, LDZ, ZWORK, LZWORK, DWORK, INFO ) * IF ( INFO.NE.0 ) THEN WRITE ( NOUT, FMT = 99997 ) INFO ELSE WRITE ( NOUT, FMT = 99981 ) DO 60 I = 1, NU WRITE ( NOUT, FMT = 99980 ) ALPHA(I)/BETA(I) 60 CONTINUE WRITE ( NOUT, FMT = 99982 ) END IF END IF NINFZ = 0 DO 80 I = 1, DINFZ IF ( INFZ(I).GT.0 ) THEN NINFZ = NINFZ + INFZ(I)*I END IF 80 CONTINUE WRITE ( NOUT, FMT = 99978 ) NINFZ IF ( NINFZ.GT.0 ) THEN DO 100 I = 1, DINFZ WRITE ( NOUT, FMT = 99977 ) INFZ(I), I 100 CONTINUE END IF WRITE ( NOUT, FMT = 99976 ) NKROR IF ( NKROR.GT.0 ) WRITE ( NOUT, FMT = 99975 ) $ ( KRONR(I), I = 1,NKROR ) WRITE ( NOUT, FMT = 99974 ) NKROL IF ( NKROL.GT.0 ) WRITE ( NOUT, FMT = 99973 ) $ ( KRONL(I), I = 1,NKROL ) END IF END IF END IF END IF * STOP * 99999 FORMAT (' AB08NZ EXAMPLE PROGRAM RESULTS',/1X) 99998 FORMAT (' INFO on exit from AB08NZ = ',I2) 99997 FORMAT (' INFO on exit from ZGEGV = ',I2) 99994 FORMAT (' The left Kronecker indices of (A,C) are ',/(20(I3,2X))) 99993 FORMAT (/' The system (A,C) is completely observable ') 99992 FORMAT (/' The dimension of the observable subspace = ',I3) 99991 FORMAT (/' The output decoupling zeros are the eigenvalues of th', $ 'e matrix AF. ') 99990 FORMAT (/' The matrix AF is ') 99989 FORMAT (20(1X,F9.4,SP,F9.4,S,'i ')) 99988 FORMAT (//' The right Kronecker indices of (A,B) are ',/(20(I3,2X) $ )) 99987 FORMAT (/' The system (A,B) is completely controllable ') 99986 FORMAT (/' The dimension of the controllable subspace = ',I3) 99985 FORMAT (/' The input decoupling zeros are the eigenvalues of the', $ ' matrix AF. ') 99984 FORMAT (//' The number of finite invariant zeros = ',I3) 99983 FORMAT (/' The finite invariant zeros are ') 99982 FORMAT (/' which correspond to the generalized eigenvalues of (l', $ 'ambda*BF - AF).') 99981 FORMAT (/' real part imag part ') 99980 FORMAT (1X,F9.4,SP,F9.4,S,'i ') 99978 FORMAT (//' The number of infinite zeros = ',I3) 99977 FORMAT ( I4,' infinite zero(s) of order ',I3) 99976 FORMAT (/' The number of right Kronecker indices = ',I3) 99975 FORMAT (/' Right Kronecker (column) indices of (A,B,C,D) are ', $ /(20(I3,2X))) 99974 FORMAT (/' The number of left Kronecker indices = ',I3) 99973 FORMAT (/' The left Kronecker (row) indices of (A,B,C,D) are ', $ /(20(I3,2X))) 99972 FORMAT (/' N is out of range.',/' N = ',I5) 99971 FORMAT (/' M is out of range.',/' M = ',I5) 99970 FORMAT (/' P is out of range.',/' P = ',I5) END

AB08NZ EXAMPLE PROGRAM DATA 6 2 3 0.0 N (1.0,0.0) (0.0,0.0) (0.0,0.0) (0.0,0.0) (0.0,0.0) (0.0,0.0) (0.0,0.0) (1.0,0.0) (0.0,0.0) (0.0,0.0) (0.0,0.0) (0.0,0.0) (0.0,0.0) (0.0,0.0) (3.0,0.0) (0.0,0.0) (0.0,0.0) (0.0,0.0) (0.0,0.0) (0.0,0.0) (0.0,0.0) (-4.0,0.0) (0.0,0.0) (0.0,0.0) (0.0,0.0) (0.0,0.0) (0.0,0.0) (0.0,0.0) (-1.0,0.0) (0.0,0.0) (0.0,0.0) (0.0,0.0) (0.0,0.0) (0.0,0.0) (0.0,0.0) (3.0,0.0) (0.0,0.0) (-1.0,0.0) (-1.0,0.0) (0.0,0.0) (1.0,0.0) (-1.0,0.0) (0.0,0.0) (0.0,0.0) (0.0,0.0) (1.0,0.0) (-1.0,0.0) (-1.0,0.0) (1.0,0.0) (0.0,0.0) (0.0,0.0) (1.0,0.0) (0.0,0.0) (0.0,0.0) (0.0,0.0) (1.0,0.0) (0.0,0.0) (1.0,0.0) (0.0,0.0) (1.0,0.0) (0.0,0.0) (0.0,0.0) (1.0,0.0) (0.0,0.0) (0.0,0.0) (1.0,0.0) (0.0,0.0) (0.0,0.0) (0.0,0.0) (0.0,0.0) (0.0,0.0) (0.0,0.0)

AB08NZ EXAMPLE PROGRAM RESULTS The left Kronecker indices of (A,C) are 1 2 2 The dimension of the observable subspace = 5 The output decoupling zeros are the eigenvalues of the matrix AF. The matrix AF is -1.0000 +0.0000i The right Kronecker indices of (A,B) are 2 3 The dimension of the controllable subspace = 5 The input decoupling zeros are the eigenvalues of the matrix AF. The matrix AF is -4.0000 +0.0000i The number of finite invariant zeros = 2 The finite invariant zeros are real part imag part 2.0000 +0.0000i -1.0000 +0.0000i which correspond to the generalized eigenvalues of (lambda*BF - AF). The number of infinite zeros = 2 The orders of the infinite zeros are 1 1 The number of right Kronecker indices = 0 The number of left Kronecker indices = 1 The left Kronecker (row) indices of (A,B,C,D) are 2