FB01QD

Time-varying square root covariance Kalman filter (dense matrices)

[Specification] [Arguments] [Method] [References] [Comments] [Example]

Purpose

  To calculate a combined measurement and time update of one
  iteration of the time-varying Kalman filter. This update is given
  for the square root covariance filter, using dense matrices.

Specification
      SUBROUTINE FB01QD( JOBK, MULTBQ, N, M, P, S, LDS, A, LDA, B,
     $                   LDB, Q, LDQ, C, LDC, R, LDR, K, LDK, TOL,
     $                   IWORK, DWORK, LDWORK, INFO )
C     .. Scalar Arguments ..
      CHARACTER         JOBK, MULTBQ
      INTEGER           INFO, LDA, LDB, LDC, LDK, LDQ, LDR, LDS, LDWORK,
     $                  M, N, P
      DOUBLE PRECISION  TOL
C     .. Array Arguments ..
      INTEGER           IWORK(*)
      DOUBLE PRECISION  A(LDA,*), B(LDB,*), C(LDC,*), DWORK(*),
     $                  K(LDK,*), Q(LDQ,*), R(LDR,*), S(LDS,*)

Arguments

Mode Parameters

  JOBK    CHARACTER*1
          Indicates whether the user wishes to compute the Kalman
          filter gain matrix K  as follows:
                              i
          = 'K':  K  is computed and stored in array K;
                   i
          = 'N':  K  is not required.
                   i

  MULTBQ  CHARACTER*1                    1/2
          Indicates how matrices B  and Q    are to be passed to
                                  i      i
          the routine as follows:
          = 'P':  Array Q is not used and the array B must contain
                                 1/2
                  the product B Q   ;
                               i i
          = 'N':  Arrays B and Q must contain the matrices as
                  described below.

Input/Output Parameters
  N       (input) INTEGER
          The actual state dimension, i.e., the order of the
          matrices S    and A .  N >= 0.
                    i-1      i

  M       (input) INTEGER
          The actual input dimension, i.e., the order of the matrix
           1/2
          Q   .  M >= 0.
           i

  P       (input) INTEGER
          The actual output dimension, i.e., the order of the matrix
           1/2
          R   .  P >= 0.
           i

  S       (input/output) DOUBLE PRECISION array, dimension (LDS,N)
          On entry, the leading N-by-N lower triangular part of this
          array must contain S   , the square root (left Cholesky
                              i-1
          factor) of the state covariance matrix at instant (i-1).
          On exit, the leading N-by-N lower triangular part of this
          array contains S , the square root (left Cholesky factor)
                          i
          of the state covariance matrix at instant i.
          The strict upper triangular part of this array is not
          referenced.

  LDS     INTEGER
          The leading dimension of array S.  LDS >= MAX(1,N).

  A       (input) DOUBLE PRECISION array, dimension (LDA,N)
          The leading N-by-N part of this array must contain A ,
                                                              i
          the state transition matrix of the discrete system at
          instant i.

  LDA     INTEGER
          The leading dimension of array A.  LDA >= MAX(1,N).

  B       (input) DOUBLE PRECISION array, dimension (LDB,M)
          The leading N-by-M part of this array must contain B ,
                                                     1/2      i
          the input weight matrix (or the product B Q    if
                                                   i i
          MULTBQ = 'P') of the discrete system at instant i.

  LDB     INTEGER
          The leading dimension of array B.  LDB >= MAX(1,N).

  Q       (input) DOUBLE PRECISION array, dimension (LDQ,*)
          If MULTBQ = 'N', then the leading M-by-M lower triangular
                                           1/2
          part of this array must contain Q   , the square root
                                           i
          (left Cholesky factor) of the input (process) noise
          covariance matrix at instant i.
          The strict upper triangular part of this array is not
          referenced.
          If MULTBQ = 'P', Q is not referenced and can be supplied
          as a dummy array (i.e., set parameter LDQ = 1 and declare
          this array to be Q(1,1) in the calling program).

  LDQ     INTEGER
          The leading dimension of array Q.
          LDQ >= MAX(1,M) if MULTBQ = 'N';
          LDQ >= 1        if MULTBQ = 'P'.

  C       (input) DOUBLE PRECISION array, dimension (LDC,N)
          The leading P-by-N part of this array must contain C , the
                                                              i
          output weight matrix of the discrete system at instant i.

  LDC     INTEGER
          The leading dimension of array C.  LDC >= MAX(1,P).

  R       (input/output) DOUBLE PRECISION array, dimension (LDR,P)
          On entry, the leading P-by-P lower triangular part of this
                              1/2
          array must contain R   , the square root (left Cholesky
                              i
          factor) of the output (measurement) noise covariance
          matrix at instant i.
          On exit, the leading P-by-P lower triangular part of this
                                 1/2
          array contains (RINOV )   , the square root (left Cholesky
                               i
          factor) of the covariance matrix of the innovations at
          instant i.
          The strict upper triangular part of this array is not
          referenced.

  LDR     INTEGER
          The leading dimension of array R.  LDR >= MAX(1,P).

  K       (output) DOUBLE PRECISION array, dimension (LDK,P)
          If JOBK = 'K', and INFO = 0, then the leading N-by-P part
          of this array contains K , the Kalman filter gain matrix
                                  i
          at instant i.
          If JOBK = 'N', or JOBK = 'K' and INFO = 1, then the
          leading N-by-P part of this array contains AK , a matrix
                                                       i
          related to the Kalman filter gain matrix at instant i (see
                                                         -1/2
          METHOD). Specifically, AK  = A P     C'(RINOV')    .
                                   i    i i|i-1 i      i

  LDK     INTEGER
          The leading dimension of array K.   LDK >= MAX(1,N).

Tolerances
  TOL     DOUBLE PRECISION
          If JOBK = 'K', then TOL is used to test for near
                                            1/2
          singularity of the matrix (RINOV )   . If the user sets
                                          i
          TOL > 0, then the given value of TOL is used as a
          lower bound for the reciprocal condition number of that
          matrix; a matrix whose estimated condition number is less
          than 1/TOL is considered to be nonsingular. If the user
          sets TOL <= 0, then an implicitly computed, default
          tolerance, defined by TOLDEF = P*P*EPS, is used instead,
          where EPS is the machine precision (see LAPACK Library
          routine DLAMCH).
          Otherwise, TOL is not referenced.

Workspace
  IWORK   INTEGER array, dimension (LIWORK),
          where LIWORK = P if JOBK = 'K',
          and   LIWORK = 1 otherwise.

  DWORK   DOUBLE PRECISION array, dimension (LDWORK)
          On exit, if INFO = 0, DWORK(1) returns the optimal value
          of LDWORK.  If INFO = 0 and JOBK = 'K', DWORK(2) returns
          an estimate of the reciprocal of the condition number
                                     1/2
          (in the 1-norm) of (RINOV )   .
                                   i

  LDWORK  The length of the array DWORK.
          LDWORK >= MAX(1,N*(P+N)+2*P,N*(N+M+2)),     if JOBK = 'N';
          LDWORK >= MAX(2,N*(P+N)+2*P,N*(N+M+2),3*P), if JOBK = 'K'.
          For optimum performance LDWORK should be larger.

Error Indicator
  INFO    INTEGER
          = 0:  successful exit;
          < 0:  if INFO = -i, the i-th argument had an illegal
                value;
                                                     1/2
          = 1:  if JOBK = 'K' and the matrix (RINOV )   is singular,
                                                   i           1/2
                i.e., the condition number estimate of (RINOV )
                                                             i
                (in the 1-norm) exceeds 1/TOL.  The matrices S, AK ,
                            1/2                                   i
                and (RINOV )    have been computed.
                          i

Method
  The routine performs one recursion of the square root covariance
  filter algorithm, summarized as follows:

   |  1/2                      |     |         1/2          |
   | R      C x S      0       |     | (RINOV )     0     0 |
   |  i      i   i-1           |     |       i              |
   |                      1/2  | T = |                      |
   | 0      A x S    B x Q     |     |     AK       S     0 |
   |         i   i-1  i   i    |     |       i       i      |

       (Pre-array)                      (Post-array)

  where T is an orthogonal transformation triangularizing the
  pre-array.

  The state covariance matrix P    is factorized as
                               i|i-1
     P     = S  S'
      i|i-1   i  i

  and one combined time and measurement update for the state X
                                                              i|i-1
  is given by

     X     = A X      + K (Y - C X     ),
      i+1|i   i i|i-1    i  i   i i|i-1

                       -1/2
  where K = AK (RINOV )     is the Kalman filter gain matrix and Y
         i    i      i                                            i
  is the observed output of the system.

  The triangularization is done entirely via Householder
  transformations exploiting the zero pattern of the pre-array.

References
  [1] Anderson, B.D.O. and Moore, J.B.
      Optimal Filtering.
      Prentice Hall, Englewood Cliffs, New Jersey, 1979.

  [2] Verhaegen, M.H.G. and Van Dooren, P.
      Numerical Aspects of Different Kalman Filter Implementations.
      IEEE Trans. Auto. Contr., AC-31, pp. 907-917, Oct. 1986.

  [3] Vanbegin, M., Van Dooren, P., and Verhaegen, M.H.G.
      Algorithm 675: FORTRAN Subroutines for Computing the Square
      Root Covariance Filter and Square Root Information Filter in
      Dense or Hessenberg Forms.
      ACM Trans. Math. Software, 15, pp. 243-256, 1989.

Numerical Aspects
  The algorithm requires

        3    2                               2   2
  (7/6)N  + N  x (5/2 x P + M) + N x (1/2 x M + P )

  operations and is backward stable (see [2]).

Further Comments
  None
Example

Program Text

*     FB01QD EXAMPLE PROGRAM TEXT
*     Copyright (c) 2002-2017 NICONET e.V.
*
*     .. Parameters ..
      INTEGER          NIN, NOUT
      PARAMETER        ( NIN = 5, NOUT = 6 )
      INTEGER          NMAX, MMAX, PMAX
      PARAMETER        ( NMAX = 20, MMAX = 20, PMAX = 20 )
      INTEGER          LDA, LDB, LDC, LDK, LDQ, LDR, LDS
      PARAMETER        ( LDA = NMAX, LDB = NMAX, LDC = PMAX,
     $                   LDK = NMAX, LDQ = MMAX, LDR = PMAX,
     $                   LDS = NMAX )
      INTEGER          LDWORK
      PARAMETER        ( LDWORK = MAX( NMAX*(PMAX+NMAX)+2*PMAX,
     $                                 NMAX*(NMAX+MMAX+2), 3*PMAX ) )
*     .. Local Scalars ..
      DOUBLE PRECISION TOL
      INTEGER          I, INFO, ISTEP, J, M, N, P
      CHARACTER*1      JOBK, MULTBQ
*     .. Local Arrays ..
      DOUBLE PRECISION A(LDA,NMAX), B(LDB,MMAX), C(LDC,NMAX),
     $                 DIAG(PMAX), DWORK(LDWORK), K(LDK,PMAX),
     $                 Q(LDQ,MMAX), R(LDR,PMAX), S(LDS,NMAX)
      INTEGER          IWORK(PMAX)
*     .. External Functions ..
      LOGICAL          LSAME
      EXTERNAL         LSAME
*     .. External Subroutines ..
      EXTERNAL         DCOPY, FB01QD
*     .. 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, JOBK, TOL, MULTBQ
      IF ( N.LE.0 .OR. N.GT.NMAX ) THEN
         WRITE ( NOUT, FMT = 99994 ) N
      ELSE
         READ ( NIN, FMT = * ) ( ( S(I,J), J = 1,N ), I = 1,N )
         READ ( NIN, FMT = * ) ( ( A(I,J), J = 1,N ), I = 1,N )
         IF ( M.LE.0 .OR. M.GT.MMAX ) THEN
            WRITE ( NOUT, FMT = 99993 ) M
         ELSE
            READ ( NIN, FMT = * ) ( ( B(I,J), J = 1,M ), I = 1,N )
            IF ( LSAME( MULTBQ, 'N' ) ) READ ( NIN, FMT = *)
     $                               ( ( Q(I,J), J = 1,M ), I = 1,M )
            IF ( P.LE.0 .OR. P.GT.PMAX ) THEN
               WRITE ( NOUT, FMT = 99992 ) P
            ELSE
               READ ( NIN, FMT = * ) ( ( C(I,J), J = 1,N ), I = 1,P )
               READ ( NIN, FMT = * ) ( ( R(I,J), J = 1,P ), I = 1,P )
*              Save the strict lower triangle of R in its strict upper
*              triangle and the diagonal in the array DIAG.
               DO 10 I = 2, P
                  CALL DCOPY( I, R(I,1), LDR, R(1,I), 1 )
   10          CONTINUE
               CALL DCOPY( P, R, LDR+1, DIAG, 1 )
*              Perform three iterations of the (Kalman) filter recursion
*              (in square root covariance form).
               ISTEP = 1
   20          CONTINUE
                  CALL FB01QD( JOBK, MULTBQ, N, M, P, S, LDS, A, LDA,
     $                         B, LDB, Q, LDQ, C, LDC, R, LDR, K, LDK,
     $                         TOL, IWORK, DWORK, LDWORK, INFO )
                  ISTEP = ISTEP + 1
                  IF ( INFO.EQ.0 .AND. ISTEP.LE.3 ) THEN
*                    Restore the lower triangle of R.
                     DO 30 I = 2, P
                        CALL DCOPY( I, R(1,I), 1, R(I,1), LDR )
   30                CONTINUE
                     CALL DCOPY( P, DIAG, 1, R, LDR+1 )
                     GO TO 20
                  END IF
*
               IF ( INFO.NE.0 ) THEN
                  WRITE ( NOUT, FMT = 99998 ) INFO
               ELSE
                  WRITE ( NOUT, FMT = 99997 )
                  DO 40 I = 1, N
                     WRITE ( NOUT, FMT = 99995 ) ( S(I,J), J = 1,N )
   40             CONTINUE
                  IF ( LSAME( JOBK, 'K' ) ) THEN
                     WRITE ( NOUT, FMT = 99996 )
                     DO 60 I = 1, N
                        WRITE ( NOUT, FMT = 99995 ) ( K(I,J), J = 1,P )
   60                CONTINUE
                  END IF
               END IF
            END IF
         END IF
      END IF
      STOP
*
99999 FORMAT (' FB01QD EXAMPLE PROGRAM RESULTS',/1X)
99998 FORMAT (' INFO on exit from FB01QD = ',I2)
99997 FORMAT (' The square root of the state covariance matrix is ')
99996 FORMAT (/' The Kalman gain matrix is ')
99995 FORMAT (20(1X,F8.4))
99994 FORMAT (/' N is out of range.',/' N = ',I5)
99993 FORMAT (/' M is out of range.',/' M = ',I5)
99992 FORMAT (/' P is out of range.',/' P = ',I5)
      END
Program Data
 FB01QD EXAMPLE PROGRAM DATA
   4     2     2     K     0.0     N
   0.0000  0.0000  0.0000  0.0000
   0.0000  0.0000  0.0000  0.0000
   0.0000  0.0000  0.0000  0.0000
   0.0000  0.0000  0.0000  0.0000
   0.2113  0.8497  0.7263  0.8833
   0.7560  0.6857  0.1985  0.6525
   0.0002  0.8782  0.5442  0.3076
   0.3303  0.0683  0.2320  0.9329
   0.5618  0.5042
   0.5896  0.3493
   0.6853  0.3873
   0.8906  0.9222
   1.0000  0.0000
   0.0000  1.0000
   0.3616  0.5664  0.5015  0.2693
   0.2922  0.4826  0.4368  0.6325
   0.9488  0.0000
   0.3760  0.7340
Program Results
 FB01QD EXAMPLE PROGRAM RESULTS

 The square root of the state covariance matrix is 
  -1.2936   0.0000   0.0000   0.0000
  -1.1382  -0.2579   0.0000   0.0000
  -0.9622  -0.1529   0.2974   0.0000
  -1.3076   0.0936   0.4508  -0.4897

 The Kalman gain matrix is 
   0.3638   0.9469
   0.3532   0.8179
   0.2471   0.5542
   0.1982   0.6471

Return to index