
Citation 
 Permanent Link:
 https://ufdc.ufl.edu/UF00082425/00001
Material Information
 Title:
 New results in indirect adaptive control
 Creator:
 Ossman, Kathleen A. K., 1959 ( Dissertant )
Kamen, Edward W. ( Thesis advisor )
Bullock, Thomas E. ( Reviewer )
Hearn, Donald W. ( Reviewer )
Peebles, Peyton Z. ( Reviewer )
Svoronos, Spyros ( Reviewer )
 Place of Publication:
 Gainesville, Fla.
 Publisher:
 University of Florida
 Publication Date:
 1986
 Copyright Date:
 1986
 Language:
 English
Subjects
 Subjects / Keywords:
 Adaptive control ( jstor )
Automats ( jstor ) Estimators ( jstor ) Governing laws clause ( jstor ) Matrices ( jstor ) Perceptron convergence procedure ( jstor ) Point estimators ( jstor ) Preliminary estimates ( jstor ) Signals ( jstor ) State estimation ( jstor ) Dissertations, Academic  Electrical Engineering  UF Electrical Engineering thesis, Ph.D.
 Genre:
 bibliography ( marcgt )
nonfiction ( marcgt )
Notes
 Abstract:
 An adaptive regulator which does not require a persistently
exciting input is derived for multiinput multioutput linear discrete
time systems. The assumptions made on the unknown plant are (1) an
upper bound on the system order is known, (2) the system parameters
belong to known bounded intervals, and (3) the plant is stabilizable for
all possible values of the unknown system parameters ranging over the
known intervals. A recursive parameter estimator is described which
forces the estimates of the system parameters to converge to the known
intervals asymptotically. Using this parameter estimator, an adaptive
LQ regulator is developed which results in a globally stable adaptive
closedloop system in the sense that the system inputs and outputs
converge to zero asymptotically. The results are then extended to the
case of tracking and/or deterministic disturbance rejection using the
internal model principle. Simulations illustrating the performance of
the adaptive controller for several discretetime systems are included.
 Thesis:
 Thesis (Ph.D.)University of Florida, 1986.
 Bibliography:
 Includes bibliographic references (leaves 100102).
 General Note:
 Vita.
 Statement of Responsibility:
 by Kathleen A. K. Ossman.
Record Information
 Source Institution:
 University of Florida
 Holding Location:
 University of Florida
 Rights Management:
 Copyright Kathleen A. K. Ossman. Permission granted to the University of Florida to digitize, archive and distribute this item for nonprofit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
 Resource Identifier:
 030575333 ( ALEPH )
17647544 ( OCLC ) AEW2312 ( NOTIS )

Downloads 
This item has the following downloads:

Full Text 
NEW RESULTS IN INDIRECT ADAPTIVE CONTROL
By
KATHLEEN A. K. OSSMAN
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN
PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
1986
ACKNOWLEDGEMENTS
I would like to acknowledge the following people for their contributions to the completion of this dissertation. Special thanks go to my advisor, Dr. Edward W. Kamen, for his numerous helpful suggestions
and creative ideas. Thanks also go to the members of my supervisory committee: Dr. T. E. Bullock, Dr. D. W. Hearn, Dr. P. Z. Peebles, and Dr. S. Svoronos.
This work was supported in part by the U. S. Army Research Office, Research Triangle Park, N.C., under Contract No. DAAG2984K0081.
TABLE OF CONTENTS
ACKNOWLEDGEMENTS .
ABSTRACT . CHAPTERS
I INTRODUCTION .
II BACKGROUND . * . . . . . . . . . . . . . .
III SYSTEM DEFINITIONS ANO ASSUMPTIONS . IV PARAMETER ESTIMATION .
V ADAPTIVE REGULATOR. . oo. .
Uniform Stabilizability of the Estimated System .
Adaptive Observer. .
Feedback Gain Sequence . . . . .
Stability of the Adaptive ClosedLoop System.
VI APPLICATION TO TRACKING AND DISTURBANCE REJECTIONo. VII SIMULATIONS. VIII DISCUSSION .o. REFERENCES. . BIOGRAPHICAL SKETCH .o. o. . . . . . .
ii iv
1
4
12 18 36 36 41 44 55 61 66 96 100
103
Abstract of Dissertation Presented to the Graduate School of The University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
NEW RESULTS IN INDIRECT ADAPTIVE CONTROL
By
KATHLEEN A. K. OSSMAN
August 1986
Chairman: Dr. Edward W. Kamen
Major Department: Electrical Engineering
An adaptive regulator which does not require a persistently
exciting input is derived for multiinput multioutput linear discretetime systems. The assumptions made on the unknown plant are (1) an upper bound on the system order is known, (2) the system parameters belong to known bounded intervals, and (3) the plant is stabilizable for
all possible values of the unknown system parameters ranging over the known intervals. A recursive parameter estimator is described which forces the estimates of the system parameters to converge to the known intervals asymptotically. Using this parameter estimator, an adaptive LQ regulator is developed which results in a globally stable adaptive closedloop system in the sense that the system inputs and outputs converge to zero asymptotically. The results are then extended to the case of tracking and/or deterministic disturbance rejection using the internal model principle. Simulations illustrating the performance of the adaptive controller for several discretetime systems are included.
CHAPTER I
INTRODUCTION
Adaptive control, the problem, of controlling a system whose parameters are unknown prior to or changing during system operation, has been a major research topic during the past three decades. Although
originally intended for timevarying. or nonlinear systems, most of the stability results on adaptive controllers have been limited to linear timeinvariant systems. The majority of adaptive controllers can be classified as either direct or indirect.
In direct adaptive control, no attempt is made to estimate the unknown system parameters. Instead, the controller parameters are updated directly using real time input/output information. One specific example of direct adaptive control is model reference adaptive control (MRAC) where the unknown plant is forced to behave asymptotically like some prechosen reference model. Several individuals have developed
this approach, see for example [13]. The assumptions needed to prove global stability of model reference adaptive controllers are (1) the
unknown plant is minimum phase, (2) the relative degree of the plant is known and (3) the sign of the plant gain is known.
Indirect adaptive controllers are applicable to nonminimum phase as well as minimum phase systems. In indirect adaptive control, the
unknown system parameters are estimated in real time using available input/output information then the control law is computed using the most
recent parameter estimates. A number of individuals have introduced indirect adaptive controllers, see for example [412]. The assumptions
used to prove global stability of indirect adaptive controllers are (1)
an upper bound on system order is known, (2) the estimated system is uniformly stabilizable, and (3) the parameter estimator possesses certain key properties. Since the adaptive controller developed in this
dissertation can be classified as indirect, each of these assumptions will be discussed in more detail.
The importance of the first assumption was illustrated in [13141 by Rohrs and associates who investigated the behavior of adaptive controllers in the presence of unmodeled dynamics. Rohrs demonstrated through computer simulations that' all existing adaptive controllers could go unstable if the order of the system was underestimated. In
response to Rohrs' findings, several individuals [1519] have achieved promising results in the development of robust adaptive controllers.
Verification of the second assumption, which requires the existence
of a stabilizing feedback control law for the estimated system, is difficult because it depends on. the parameter estimates which are generated in real time. Unless something more can be said about where the parameter estimates are going, the assumption cannot be verified a priori and, consequently, global stability cannot be ensured. The first approach towards satisfying this assumption was the use of a persistently exciting external input which allows perfect identification of the plant. This approach along with more recent results which do not
require persistent excitation are discussed in more detail in Chapter .Ii.
The third assumption requires that the parameter estimator possess certain key properties which are explained in detail in Chapter II. It will suffice to mention the existence of several parameter estimation
algorithms, .such as the recursive leastsquares, projection, and orthogonalized projection algorithins discussed in ï¿½6], which do indeed have the required properties.
In this dissertation, an indirect adaptive regulator which does not
require a persistently exciting input is derived for multiinput multioutput discretetime systems. In Chapter II, more explicit background material on parameter estimators and indirect adaptive controllers is presented. Chapter III contains the system definitions and the assumptions which include (1) an upper bound on system order is known,
(2) the unknown system parameters belong to known bounded intervals, and
(3) the plant is stabilizable for all possible values of the unknown parameters ranging over the known intervals. A parameter estimator is
derived in Chapter IV which has the required properties for proving global stability of indirect adaptil/e controllers and, in addition, forces the estimates of the parameters to converge to the known bounded intervals. In Chapter V, it is first shown that the assumptions listed in Chapter II ensure the parameter estimator described in Chapter IV will generate a uniformly stabilizable estimated system. Using this
parameter estimator, an adaptive LQ regulator is developed which, when applied to the unknown plant, results in a globally stable closedloop system in the sense that the system inputs and outputs converge to zero asymptotically. The results are then extended in Chapter VI to the case of tracking and/or deterministic disturbance rejection using the internal model principle ï¿½20]. Chapter V.II includes simulations of the adaptive LQ controller for both SISO and MIMO discretetime systems. A
discussion of the results and further work to be investigated is included in Chapter VII.
CHAPTER II
BACKGROUND
In indirect adaptive control, the system parameters are estimated in real time using available input/output information. The control law
is then computed for the estimated system using the most recent parameter estimates. A wellknown problem in proving global stability of indirect adaptive control algorithms is the estimated system may not be controllable or even stabilizable for certain parameter estimates which are referred to as singular points. Global stability of the indirect
adaptive controller can only be ensured if there is some finite point in
time after which the parameter estimates are not arbitrarily close to a singular point. An example of a parameter estimator and an adaptive regulator will further illustrate the problem of singular points in indirect adaptive control.
Consider the SISO discretetime system described by the input/output difference equation:
q p
y(k) =jllajy(kj) + I b.u(kj). (21)
j=1
In (21), y(k) is the system output and u(k) is the control input. It
is assumed that an upper bound, n, on q and p is known but some or all of the system parameters aj and b. are unknown. The system described by
(21) can be rewritten in the following form which is convenient for parameter estimation:
y(k) = eT ,(kl)
where
T a . an b1 . bn] (22)
T (k1) = [y(k1) . y(kn) u(k1) .u(kn)].
The vector e consists of all the system parameters and *(kl) is a regression vector of past inputs and outputs. Throughout the following discussion, u ï¿½ o will designate the vector Euclidean norm defined by IlxH = (xTx)
Of all the parameter estimation schemes discussed in [6], the leastsquares algorithm has the fastest convergence rate and is most easily modified to handle output disturbances and slowly timevarying systems. The algorithm results from minimizing the quadratic cost function:
N
JNO) = 1/2 1 +YR VIe 2 2(06(0))Tp(0) (66(O))o
k=1
The cost function consists of the sum of the squares of the prediction errors plus an additional term which takes initial conditions into account. The matrix P(O) can be interpreted as a measure of confidence in the initial parameter estimate e(O). Designating e(k) as the estimate of the system parameters at time k, the leastsquares algorithm is described by
e(k) = e(k1) + TP(k1)0(k1) y(k)  eT(k1) (k1)]
1 + * (k1)P(k1) (k1)
P(k) = P(k1)  P(k1);(kl) T(k'l)P(k1), P(O) = pT(0) > 0. (23)
1 + T (k1)P(k1).(k1)
The key properties of the leastsquares scheme used for proving global stability of an indirect adaptive controller are
i. uiO(k)u
i i. iio(k)e(k1)u a 0 as k +
iii. lim e2(k) = 0
k+ 1 + ,T(k1)P(k1),(k1)
where e(k) = y(k)  oT(kI) (k1).
The error e(k) is referred to as the prediction error since it is simply the difference between the actual output at time k and the predicted output at time k using the most recent parameter estimate. These properties do not depend on the type of control input chosen or on the boundedness of the system input and output. The leastsquares algorithm
also has the property that the parameter estimates converge, although not necessarily to the true values. This property is not included among the key properties because it is not necessary for stability proofs.
An example of indirect adaptive control in the SISO case is Samson and Fuchs' LQ controller discussed in C12] and briefly outlined here. A
statespace observer realization of the system described by (21) is given by
x(k+l) = Fx(k) + Gu(k) (24)
y(k) = Hx(k)
where
a, b
a2 1b2
F . G . H= [1 0 . 0].
anI 1 bnI
an 0 bn
Samson and Fuchs assume that (F,G) is stabilizable. This assumption
allows for nonminimum phase systems and systems with stable common poles and zeros. The adaptive LQ control law used by Samson and Fuchs [12] is
given by
u(k) = L(k) (k), (25)
where u(k) is the input to the given plant, x(k) is the state estimate,
and L(k) is a stabilizing feedback gain for the estimated system (F(k), G(k)).
The state estimate (k) is generated from an adaptive observer. Letting F(k) and G(k) represent the current estimates for matrices F and G, the adaptive observer for the system is given by
X(k+l) = F(k)2(k) + G(k)u(k) + M(k) (y(k)9(k)) (26)
AA
y(k) = Hx(k)
where
al(k)
M(k) = . , ai(k) = estimate of ai at time k.
an(k)
The feedback gain sequence is computed at each time point k by solving one step of a Riccati difference equation:
R(k+l) = Q + LT(k)L(k) + (F(k)  G(k)L(k))TR(k)(F(k)  G(k)L(k))
L(k) = [GT(k)R(k)G(k) + I1 GT(k)R(k)F(k) (27)
In (27), Q and the initial value R(O) are arbitrary positive definite symmetric matrices.
In order to prove global stability of the closedloop system, Samson and Fuchs make two assumptions in addition to the stabilizability assumption on (F,G) mentioned previously. First, the parameter estimator must possess the following three properties:
i. ue(k)o < M <  for all k
ii. ne(k)  e(km)n 0 as k + for any finite m iii. Ie(k) 4 c(k)ni@(kl)ll + 8(k)
where a(k) and a(k) converge to zero.
An example of a parameter estimator possessing these properties is the
leastsquares estimator discussed previously. Clearly, properties (i) and (ii) listed above are equivalent to properties (i) and (ii) given for the leastsquares algorithm. It is not quite as obvious that Samson
and Fuchs' property (iii) is equivalent to property (iii) of the leastsquares algorithm. The proof is given in Chapter IV.
The second assumption for proving global stability is that the estimated system (F(k),G(k)) must be uniformly stabilizable.
Definition: The system (F(k),G(k)) is uniformly stabilizable if there exists an integer r > 1, a constant q and a uniformly bounded sequence L(k) such that
t+r1
nII (F(k)G(k)L(k))u1
where
t+r1
I (F(k)G(k)L(k)) = (F(t+r)  G(t+r)L(t+r))x
k=t
(F(t+r1)  G(t+r1)L(t+r1)) . (F(t)  G(t)L(t)).
Theorem: Assuming the system (F,G) is stabilizable, the parameter estimator possesses the three required properties, and the estimated
system (F(k),G(k)) is uniformly stabilizable, the control law u(k) =
L(k)^(k) described by (25)(27) will result in a globally stable closedloop system in the sense that the system input and output
converge to zero for any initial states in the plant and observer (Samson and Fuchs C12]).
In order to prove global stability, Samson and Fuchs must assume the estimated system (F(k),G(k)) is uniformly stabilizable. Unfortunately, this stabilizability condition cannot be checked a priori which
means global stability of the adaptive regulator cannot be guaranteed. If the leastsquares estimator is used, the assumptio n that (F(k),G(k)) is uniformly stabilizable is equivalent to requiring that (F.,G.) is
stabilizable where F~ irn FMk and G. = lim GMk.
As previously mentioned, the problem of avoiding singular limit points is common to all indirect adaptive controllers. Until the recent work of LozanoLeal and Goodwin ï¿½21], De Larminat ï¿½22] and Kreisselileier ï¿½23), the only way to avoid a singular limit point was to require a persistently exciting input which would force the parameter estimates to
converge 'to the true system parameters. This approach has been developed by a number of individuals, see for example [49]. The idea is to use an external input with many different frequencies which enriches the input/output information allowing perfect identification of the system parameters. As discussed in [1516], indirect adaptive controllers which require a persistently exciting input are not robust. Since the
adaptive closedloop system is inherently timevarying and nonlinear, any uncertainty in the plant could counteract the exciting input. The
problem of choosing a persistently exciting input in the presence of unmodeled plant dynamics has. not been completely resolved.
LozanoLeal and Goodwin ï¿½21] developed an estimation scheme which gives nonsingular parameter estimates at each point in time and in the limit for SISO linear discretetime systems. They modify (when necessary) the parameter estimates generated from the leastsquares estimation algorithm with data normalization. This modification allows them
to prove global stability of an adaptive pole placement scheme without requiring a persistently exciting input. It is, however, possible for
the parameter estimates to converge to a point that is near a singular point, in which case the controller gains may be large. Also, a MIMO version of the algorithm is not availa ble at the present time.
De Larminat [22] has also proposed a parameter estimation scheme that does not yield singular points. He assumes a priori knowledge of a
space G which contains the actual system parameters and is devoid of singular points. He then introduces a modification to the estimates obtained from the standard leastsquares algorithm. The modification is
only required for a finite period of time and produces estimates which belong to G for all time. Although his modification prevents convergence near a singular point, De Larminat's recursive procedure is not as explicit as the one proposed by LozanoLeal and Goodwin [21].
As in the case of De Larminat, Kreisselmeier ï¿½23] assumes prior information on the system parameters is available. Specifically, he assumes that the components a. of the system parameter vector e lie within a known bounded interval [,min, mrax, liHe then modifies the
1 1
identification scheme for SISO continuoustime systems to force the parameter estimate of e. to converge to the set Cmm max formerr i. Kreisselmeier also assumes that for each e with e0 C min 0rnax I
there is a feedback control system with a prescribed degree of stability. This condition can be checked a priori since it does not involve the parameter estimates. With these assumptions, Kreisselmeier proves global stability of the SISO continuoustime adaptive controller.
CHAPTER III
SYSTEM DEFINITIONS AND ASSUMPTIONS
The system to be regulated is the multiinput multioutput linear discretetime system described by
y(k) =  A.Y(kj) + B.u(kj) (31)
j=1 3 j=1 J
In (31), y(k) is the mx1 output vector and u(k) is the rxl control input vector. It is assumed that an upper bound, n, on p and q is known but all or some of the entries in the matrices A. and B. are unknown. The system described by (31) can be rewritten in the following form convenient for parameter estimation:
y(k) = pT4(kl) (32)
where
pT = [ A, An B1 , . Bn]
T(k1) = [yT(k1) . yT(kn) uT(k1) . uT(kn)]
The n(m+r)xm matrix P consists of all the system parameters and the n(m+r)xl vector *(kl) is a regression vector of past inputs and outputs. In certain applications, some of the entries in the system matrix P will be known a priori. Since it is not necessary to estimate known parameters, a scheme for separating the known parameters from the unknown parameters is advantageous. This is accomplished by rewriting (32) in the following form:
y(k) = T a(k1) + T(kl) (33)
The Nxm matrix e contains all of the unknown entries in P while the matrix p contains only those entries of P which are known a priori. The vectors a (k1) and *b(k1) are regression vectors whose components come from (kl). Clearly, the decomposition described in (33) is not unique for multiinput multioutput systems. In order to minimize the
required computations for parameter estimation, the size of a (k1)
should be made as small as possible. An example will serve to
illustrate these concepts.
Example 1: Consider the linear discretetime system with two outputs
and a single input described by
=k' yl(k) 1 a] y(k1) + 3 u(k)
Y2(k)  2 y2(k1) i
The parameter a is assumed to be unknown. The system can be rewritten
in the form
y(k) = PT(kR1)
where
1 a 3yl(k1) 1 PT= (kl) = yg(ki)
2 [ u(ki)
A decomposition for this system which minimizes the required computations for parameter estimation is given by
y(k) = Y2(k1) + 2(k)
OT a a(k1) : Y2(k1) T 1 ] b(kl) = (k1).
Again, consider the linear discretetime system described by (33). It is assumed that each component eij of the unknown system parameter matrix o belongs to a known bounded interval L m 'in max
This assumption is reasonable in those applications where some a priori information is available on the system. Clearly, the unknown system
parameter matrix e belongs to a subspace of ]RNxm which is defined by the known bounded intervals. In order to use the necessary concepts of boundedness, compactness and convergence, a suitable norm must be defined on IRNxm. The vector norm which will be used throughout this dissertation is the Euclidean norm defined by
Uxii : (xTx) 1/2
The matrix norm which is induced by the vector Euclidean norm is given
by
T 12
IMn = Ex (MTM)]
TT
where max (M M) is the eigenvalue of MTM with the greatest
magnitude. By the above assumption on the entries of e, the unknown system parameter matrix e belongs to a known compact subspace n of ]RNxm given by
Nxm min max
:O R Oijceoij ,6ijJ}
The concept of compactness (i.e., closed and bounded) is relative to the matrix norm previously defined.
The only additional assumption made on the system is a pointwise stabilizability assumption over 0. Specifically, the system described by (33) is assumed to be stabilizable for each eE0. This assumption can be verified a priori because it depends only on the known set a, not on the parameter estimates. The stabilizability condition can be tested using the transfer function matrix of the system. Let D(z1) and N(z") denote the polynomial matrices defined by
n n
O(z"1) = I + I A.(z'J), N(z'1) : B(zJ).
j=1 j l
The coefficients of D(z1) and N(z"I) depend on the system parameters and therefore can be viewed as functions of o. This dependancy is made explicit by denoting D(z1) and N(z"1) as D(zl,O) and N(z1,o) respectively. The transfer function matrix of the system described by (31) is then given by
W(z1,6) = DI(z1,e)N(zI,e).
The stability assumption is that the system described by (31) with transfer function matrix W(z"1, o) can be stabilized by dynamic output feedback for all oca. From well known results [24], this assumption is equivalent to the following rank condition:
rank [D(z,e) N(z,e)] = m for Izi < 1 and all eeo.
where m is the number of system outputs and
n n
D(z,e) = I + I Ajzj, N(z,o) = I B'zj j=1 J j=1 j
As previously mentioned, this stabilizability condition can be checked
since it does not depend on the parameter estimates. An example will
illustrate the test.
Example 2: Consider the system described in Example 1.
[1z az 3z1
ED(z,e) N(z,e)] = J
I 2z z
In order to satisfy the stabilizability assumption, the rank of [D(z,e) N(z,e)] must equal 2 for Izi 4 1 and ec. The determinants of
the 2x2 submatrices of [D(z,e) N(z,e)] are (2+a)z2  3z+1, z((6a)z3), z(14z). All three minors are zero only when z = 1/4 and a = 6. Hence the rank condition is violated only when z = 1/4 and a = 6. Therefore, the stabilizability assumption will be satisfied for any interval which does not contain the point I6}.
The amount of effort involved in checking the rank condition in (34) will of course depend on the number of unknown parameters. It may
be possible to use a rootfinding algorithm for polynomials in several variables (i.e., z and eij) to check the rank of [D(z,e) N(z,o)] for all GES. It should also be noted that by overestimating the max (p,q) in (31), the system cannot be controllable, but may be stabilizable. In
this case, adaptive pole placement would not be stabilizing feedback could be computed using the discussed in Chapter II.
To summarize, the assumptions on the rinput discretetime system described by (31)  (33) are
possible, but a Riccati approach moutput linear
Al: An upper bound, n, on p and q in (31) is known; A2: the components oij of the unknown system parameter matrix e belong
to known bounded intervals [min ,max] ;
A3: the system described by (33) is stabilizable for all e belonging NxMmin _max.
to n where n = {e Nxm: Oij E Lij ,eij D.
CHAPTER IV
PARAMETER ESTIMATION Consider the rinput moutput linear discretetime system described
y(k) = eT a (k1) + T b(k1).
(41)
As previously discussed in Chapter III, it is assumed that each component eij of the unknown system parameter matrix e belongs to a known bounded interval min max eij , eij.
The parameter estimation algorithm is given by
P(kl) a (k1)
e(k) = O(k1) P(kl)f(e(ki)) + 2 T x (42)
nk1 + a(kl)P(kl)a (ki)
Ey T(k)  *T (kl)ip T(ki)e(k1)]
P(k) = P(k1) 
P(k1)@a (kl) T(k1) P(k1)
2 +T
nk.1 +a(k1)P(k1)a (k1)
0 < P(O) = pT(0) < 21, (43)
0i(k)  eTix when 6ij(kI) > eiax fij(kl) e i(k1)  emn when e (k)
0 when 0. .(k1) E[emin 0maxI ij1 ij , ij
(44)
JI when the determinant of P(k) > e
nk 1 =where e. is any small positive number. (45)
"max (l111 a (kl)ii) otherwise
The parameter estimation algorithm described by (42)  (45) differs from the recursive leastsquares estimator in three ways. First, the initial covariance matrix P(Q) in (43) must be less than 21,
whereas the leastsquares algorithm allows for any positive definite initial covariance matrix. This condition is needed to ensure that the parameter estimator described by (42)  (45) possesses the properties listed in Theorem 4.1 which are necessary for proving global stability of the adaptive regulator discussed in Chapter V.
The second difference is the data normalization introduced through the term nk1 defined in (45). In the leastsquares algorithm,
T'k1 is simply equal to one. Again, the data normalization is necessary for proving that the parameter estimator has the desirable properties given in Theorem 4.1. Simulations have shown that if e is not
sui tably small, the normal izati on has a detrimental ef fect on the transient response of the system. This effect can be minimized by
choosing e to be very small in (45). Simulations which show the effect of the data normalization and the choice of e on system transient response are discussed in Chapter VII.
The major difference between the estimator given by (42)  (45) and the recursive leastsquares estimator is the addition of a "correction term" P(k1)f(e(k1)) in (42). This term forces the estimates
Oij(k) of the components of the system parameter matrix e to converge to the sets Lo j, o. ]. This property, when combined with the stabilizability assumption over Q discussed in Chapter III, eliminates the need
for a persistently exciting input. It is important to note that the
algorithm is not a projection algorithm. That is, eij(k) does not
rm eri maxl
belong to the set e ie'i. ] for every k but does converge to the set in the limit as k + .
The idea for the correction term came from Kreisselmeier [23]. In
fact, f(e(k1)) is defined exactly the same as in [23]; however, the parameter estimator described here for MIMO discretetime systems is quite different from the adaptive observer Kreisselmeier uses for SISO continuoustime systems.
The following theorem shows the estimator described by (42) (45) has many of the same properties as recursive leastsquares. In
addition, the estimates eij(k) converge to the known intervals ruin max
Oi , eT. ]. The prediction error e(k) which appears in the theorem ij 13
is defined as
e(k) = y(k)  eT (kl)a(ki) T T(k).
Theorem 4.1: The estimation algorithm (42)  (45) has the following properties
i. P(k) converges to a positive semidefinite matrix P. < 21;
ii. I eij(k) i < M <  for all k and i = 1,2,.,N, j = 1,2,.,m;
iii. 2 T le~k) + 0 as k +
nk_1 + a (kl)P(kl)a (k1)
iv. ue(k)u < a(k)# (k1)ln + b(k), where a(k) and b(k) are positive scalar sequences which converge to zero;
v. f(e(k)) + 0 as k + w, which implies that oij(k) converges to the
set m ij 'max ] for i = 1,2,.,N, i = 1,2,.,m;
vi. iei(k)  eij(kp)I + 0 as k +  for any integer p, i = 1,2,.,N, j = 1,2,004,m;
mi max
vii. If eij m 0i. or 0. for every i and j, then . .(k) converges Smin maxto a point in L ein mx.
Proof: Property (i) is a wellknown property first proved by Samson [10) with nk_1 equal to one. The proof is included here for the sake of completeness. Equation (43) implies that 0 4 xTP(k)x~xTP(k1)x for all xe RN, and for all k > 0. Therefore, for each x RN, the sequence xTP(k)x is monotone decreasing and must converge. Denoting
Pij(k) as the ijthm element of P(k), it follows that
(e. + e.)TP(k)(e. + e )  eTP(k)ei eTP(k)e.
P.(k) = 1 i e{2 1 eL T (47)
where
0
ei h0
1 ~ ith position.
0
0
Each of the terms on the right hand side of the equation (47) must converge which implies Pij(k) converges for every i and j. Noting that
P(O) < 21, it then follows that P(k) must converge to a positive semidefinite matrix P. < 21. To prove (ii), let '4(k) = o(k)o. Rewriting
(42) in terms of "e(k) gives
'(k1)  P(k1) a(k'1),Ta(k'1)6(k'1)
2 T P(k1)f(6(k1)). (48)
nk1 + a(k1)P(k1)a(ki)
Multiplying both sides of the equation (43) on the right by P1(k1) gives
P(k1)oa (k1)T (k1) P(k)P1l(k1) =Ia a0(49)
k + *T(k'l)P(k'1)a (ki)
Combining (48) and (49) yields
e(k) = P(k)P' (k1)O(k1)  P(k1)f(e(k1)). (410)
Now define the Lyapunov functional V(k) = Tr( P(k)p (k)'(k)) 0, where "Tr" denotes the trace operation. Inserting equation (410) into the expression for V(k) gives
IT 1 T1
V(k) = Tr ce (k1)P (k1)i'(k)]  Tr[f (e(k1))P(kl)p 1(k)(k)].
(411)
Inserting the expression (48) For "e(k) in the first term on the right hand side of (411) yields
V(k) = V(k1) 
hne(k) 112
2 + T (kl)p(kl) (ki)
nk. I+
 Tr [VT(k1)f(o(k1))]
Tr[fT(e(k1))P(kl)P' I (k)'(k)].
Equation (410) right hand side
V(k) = V(k1) 
for 16(k) is then inserted into the last term on the
of (412) giving
ne(k) 112
2 T~k~~1
"ki_ + Oa(kl)P(k1) a(k1)
2Tr[eT(k1)f(4(k1))] (413)
+ Tr[fT(o(k1))P(k1)p'I(k)P(k1)f(o(k1))]. It follows from the definition of f(O(k1)) in equation (44) that
Tr[fT (o(k1))"'(k1)] > Tr[f T(o(k1))f(e(k1))1. (414)
(412)
Combining equations (413) and (414) gives
V(k) 4 V(k1)  2 T
nk1 + a(kl)P(kl) a (k1) (415)
Tr[f T(e(k1))C21P(k1)P 1(k)P(k1)]f(o(k1))].
If the matrix 21  P(k1)P'1(k)P(k) is positive definite for k sufficiently large, then V(k) will be a monotone decreasing sequence which must converge since V(k) is bounded below by zero. In order to show
that the matrix 21  P(kl)Pl(k)P(k1) is positive definite for k sufficiently large, first assume that the determinant of P(k) does not converge to zero. In this case, P'1(k) converges to P1 which implies that 21  P(k1)P'l(k)P(k1) converges to 21  P . The matrix 21  P. is positive definite by property (i) in Theorem 4.1. Now suppose that
the determinant of P(k) does converge to zero. Using (43) and the
matrix inversion lemma gives
a(kl) T(kl)
P1(k) = . P(k1) a (
2
n ki
Multiplying both sides of this equation on the left and right by P(k1) and adding 21 to both sides give
P(kl) a (kl) a (kl)P(k1)
21  P(kI)P'I(k)P(k1) 2 21P(k1)
2
n k 1(416)
Since P(k) converges, equation (43) implies that the matrix
T
P(kl) a (k1) a(kl)P(k1)
2 +T
k_1 a(kl)P(kl)a ( ki)
converges to zero. For k sufficiently large, nk.1 = max (1,,ia (k1)ii) which implies that
T T
P(RI) ï¿½ (k1) Ta(k1) P(R1) P(RI) a(kl) Ta (kl) P(R1)
2 aT a >2
nk1 + a(kl)P(kl) a(k1) nk1 + ,maxP(O)R4 a(kI)i2
P(R1 )a (k,I )OT (k1) P (k1) m2ax a >0
nlk_1(1 + XmaxP(O))
P(k1)a (k1)0T(k1)P(k1)
It then follows that the matrix 2 converges
nk.1
to zero since it is bounded above by a matrix which is converging to zero. Thus by equation (416), the matrix 21  P(k1)P1(k)P(k1) also
converges to 21  P. in the case when the determinant of P(k) converges to zero. Therefore, the matrix 21  P(k1)P1(k)P(k1) is positive definite for k sufficiently large since the matrix converges to the positive definite matrix 21  P. By (415), this implies that V(k) is a monotone decreasing sequence for k sufficiently large and must therefore converge. It then follows that ij(k) is bounded (property (ii))
by noting that
V(k) = Tr(T(k)Pl(k)e(k)) > Exmin P'l(k)] Tr( T(k)o6(k))
N m 2
V(k) > 1/2 Tr(CT(k)^d(k)) 1/2 1 1 [e .(k) e 2 i=1 j=1 Jij
Now solving (415) recursively gives the relationship
q ne(j g )12
0 < V(q) 4 V(O)  1 2(
q j=1 Tj1 + a a (417)
 I Tr[fT(e(jl))C21  P(j1)P'I(j)P(j1)]f(e(j1))].
j=1
For sufficiently large k, the terms within the summations are nonnegative and must converge to zero since V(k) is nonnegative. Properties
(iii) and (v) then follow easily. Property (iv) follows immediately
from (iii) by first noting that
2le(k) II 1 b(k)
2 + oT(k.1)P(kl)o (k1)) 1/2 where b(k) is a positive scalar sequence which converges to zero. Multiplying both sides of the inequality by
2 1/2
(n2_1 + a(k1)P(k1)oa(k1)) and noting that nk1 < 1 + l0a (k1)i gives
gIe(k)ii 4 b(k)(nk2_ + T 1/2
< b(k)(Nk_1 + (maxP(k1))l/2 "ha(k.1)I)
4 a(k)ula(k1)o + b(k)
where a(k) and b(k) are positive scalar sequences which converge to zero. To prove (vi), first subtract '(k1) from both sides of equation (48). Premultiplying both sides of the resulting equation by its transpose and noting that e(k)  T(k1)oa (k1) gives
Tr[(6(k)O(k1)) T(e(k)e(k1))] = Tr[(A'(k)  eA(k1))T(eI(k) '(k1))]: e(k)o T (kl)P(kl)2a (kl) T (k) Tr [ 2a a TI
Tr kI+ ,Ta(k l)P(kl) a(ki))2
2 T1
[fT((ki)) P(k 2 (k) a
+2Tr T (kl)P(kl) (a
nk.1 + a l
+ Tr[fT (e(k1))P(k1) 2f(o(k1))]. (418)
Also
Ta (kI)P(kl) 2a (kI) Ta (kiP(k)a (kI)
2 T a(k_) 4 [maxP(O)] 2 T
nkI + a(kI)P(k) a nk+a(k)P(k)a(k1)
SIXmax P(O) (419)
Combining (418) and (419) gives Tr[(e(k)e(k1)) T(o(k)e(k1))] 4 [xmaxP(O)] l e(k)ii 2
max 1+ (k.1)P(kl) a (ki)
P(k) 2 a(kl) T(k'l)_'(k'l) + 2Tr [fT(o(k1)) a a
+maOTr+fT (k)P(k1) (ki)
+ 2 (P(O))Tr~f T(o(ki))f(e(ki))]. (420)
By properties (iii) and (v), the first and third terms on the right side of the inequality (420) converge to zero. The second term on the right side of the inequality will also converge to zero by property (v) provided that the matrix
P(kl) 2 a (k1) T(k1)
a a
2 7Ta .11,kl)@
nk1 + ,T(kI)P(kl) (ki)
is bounded. This matrix is shown to be bounded by first noting that
p2k1)@(k1) T(k1)
p 2 (kl)a (k1) T2kk12)a a (k)2
(i)a 1a (k1) (ki ~(kI)ii2
0 42 T(kil)P(k1) (ki) 2
nk1 + @a '' ank1
Ibk kiI 2 + Xmin P(k1)
Clearly, the numerator on the right hand side of the inequality is bounded above. The denominator is bounded below by one if the determinant of P(k) converges to zero, and by X min(P ) if the determinant
of P(k) does not converge to zero. Thus, all three terms on the right
hand side of the inequality (420) converge to zero. Since
N m
Tr[(e(k)  6(k1))T(o(k)  o(k1))] =I X coij (k)  ai(k1)]
i=1 j=i 1J
it follows that e ij (k)  i (k1) must converge to zero which proves
property (vi) in the case where p = 1. The proof for arbitrary p follows from the Schwarz inequality. To prove (vii), equation (410) is
first solved recursively giving
n1
e(n) = P(n)P1(o)'(0)  P(n) P' (i+l)P(i)f(e(i))
i=O
It then follows that
e(m)  e(n) = '(m)  1'(n) = (P(m)  P(n))P1 (0)'(0)
m1
P(m) I P I(i + 1)P(i)f(e(i)) i=0
n1 1
+P(n) P (i +1)P(i)f(e(i)).
i =0
Without loss of generality, assume that m is greater than n. Replacing the matrix P(n) preceding the summation with P(m) + (P(n)  P(m)) gives
e(m)  e(n) = (P(m)  P(n))P'I(0)Y(O)  P(m) i P1(i + 1)P(i)f(O(i)) i~n
n1
+ (P(n)  P(m)) I P(i + 1)P(i)f(o(i)).
i=O
Using the properties of the induced matrix norm defined in Chapter II gives
le ijn(m)  0i (n)l < 0le(m)  e(n)!
< ,P(m)  P(n)u oP1 (0)"O'(O),
m1 1
+ HP(m) ii 1 I1P (i+l)P(i)R n f(o(i))ll i=n
(421)
nI1
+ ,P(n)  P(m)ll Z uipl(i+1)P(i)i llf(e(i))1i.
i =0
If the right hand side of equation (421) can be made arbitrarily small for m and n sufficiently large, then each component 0i(k) of the matrix
sequence e(k) will be a Cauchy sequence. It will then follow using
property (v) that oij(k) converges to a point in the interval
mai omaxl
18 mi m " In order to prove that the right hand side of the inii '0 .1
equality (421) can be made arbitrarily small, two things must be
P1(
shown: lp (k+l)P(k)iu < v <  for all k, and I f(e(k))ii < L < .
k=1
In order to prove oP (k+1)P(k)in < v <  for all k, first assume that
the determinant of P(k) does not converge to zero. In this case,
P1(k+1) converges to P1 which implies that P1(k+1)P(k) converges to I. Therefore, P P'1(K+1)P(k) ii is bounded for all k when the determinant of P(k) does not converge to zero. Now suppose the determinant of P(k) does converge to zero. Applying the matrix inversion lemma to equation (43) gives
I + a(k) T(k)P(k)
Pl(k+l)P(k) a 2a
k
Using the properties of the induced matrix norm, it follows that
SNa(k) T(k) n
HP 1(k+l)P(k)H < iti + 2 aP(k)ii (422)
n'k
From (45), there exists a finite M such that nk = max (1,uca (k)it) for k > M. It then follows that all the components of the matrix
fa ( k) T(k)
2a have magnitude less than or equal to one for k > M which in
nlk
,a (k) a(k)
turn implies 2 is bounded for all k. Noting also that
nk
iiIii = 1 and P(k)n < 4, equation (422) implies P1 (k+1)P(k)ii is
also bounded for all k in the case when the determinant of P(k) converges to zero. Therefore, there exists a finite v such that
iliP (k+i)P(k)il < v < for all k. Applying this result to the inequality (421) gives
mni
eij (m)eij(n)i 4 iiP(m)P(n)l IiP1 (o)A'(0)II + 2v m Ilf(o(i))ii i=n
n1
+ RP(m)P(n)ov I uf(e(i))i. (423)
i=O
In order to complete the convergence proof, it must be shown that
Sf(O(k))n: < L < . It follows from equation (413) that
k=O
V(k) .< V(ki) 2 T e(k)a 2  2Tr[9'T(kl)f(6(kl))]
V1 kI + a(k1)P(k')O a(ki)
+ 'max[P(kl)Pl(k)P(kl)]Tr[fT(e(kl))f(O(kl))]. (424)
Now assume < e
Nowi asm j< ij j,.
6ij is an interior point of me min max
Lij ' i )" Using the definition of
f(e(k1)) in (44), if f(ii (kl)) * 0 then I"ii(k1)I > q > 0 where q
m ,max_ , min ] for i=1, ., N,
= m n ij ij ï¿½ij ij "'"
j=1, ., m. Also from the definition of f(e(k1)), it must be true that the sign of e'ij(k1) is equal to the sign of fij(e(k1)) for i=11 ., N and j=1, ., m. These facts when combined with equation (424), imply
i2 N m
V(k) < V(ki)  2  2q I I lfij(e(k))l
"nk1+ 1 T(k1)P(k1)Oa (k1) i= j=1
+max [P(kI)P1(k)P(kl)Tr[fT (e(ki))f(e(ki))].
+a
Rearranging terms in this inequality results in
ie(k) u 2
2 + T(kl)P(kl)o a(k1)
nk1 a aa
N m
+ 2q I I fij(e(k1))
i=1 j=1
V(k1)  V(k) + Xmax[P(kl)P (k)P(k1)]Tr[fT (e(k1))f(e(k1))].
Summing all the terms in this inequality from k equals one to M and N m
noting that nf(e(k1))oi 4 1 I ifi(e(k1))i gives
i=1 j=1
M le(k) 112
k=1 nk12 + ,Ta(k.1)P(k.1)a(k1)
M
+ 2q I 1lf(6(k1))11
k=1
(425)
M
V(O)  V(M) + Z maxEP(k1)P'l(k)P(k1)]Tr[fT(e(k1))f(e(k1))]ï¿½ k=1
If the right hand side of the inequality (425) is bounded in the limit as M approaches infinity, then the norm of f(e(k1)) will be summable. The term V(M) is bounded since V(k) converges. Also
XmaxCP(kI)PI(k)P(k1)]
is bounded since the matrix P(kI)PI(k)P(k1)
converges to P. It remains to be shown that 00 T
I Tr[fT(e(kl))f(6(k1))] < ï¿½ k=1
Equation (417) implies
k Tr[fT((k1))E21P(kl)Pl(k)P(k1)]f((k1))] <
k=1
Also, there exists a finite N such that the matrix [21P(k1)P1(k)P(k1)] is positive definite for k > N because [21P(k1)P1(k)P(k1)] converges to the positive definite matrix 21P. Since
0 < min [21P(k1)P'l(k)P(k)P(k1)]Tr[f T(o(k1))f(o(k1))]
k=N
' Z Tr[fT(e(k1))[21P(k1)P' (k)P(k1)]f(6(k1))] ,
k=N
it follows that Z Tr[fT(o(k1))f(o(k1))] < = Therefore, the right
k=l
hand side of the inequality (425) is bounded in the limit as M approaches infinity. This implies the existence of a finite L such that
I off(o(k1))o < L < =. Returning to the inequality (423), let
k=l
e > 0 be given. The sequence oij(k) will be a Cauchy sequence if IOij(m)eij(n)I is less than e for m and n sufficiently large. Since
P(k) converges to P=, there exists Nl(e) such that
iP(m)P(n)o < el for m,n > NI(e)
where el = (1/3) min [e/(nP (o)o(0)i1, e/(vL)].
Also, there exists N2(e) such that
m1
I Of(e(i))II < e/6v for m,n > N2(E). i=n
It then follows from (423) that leij (m)  ij(n)I < e for m, n >
max (N1,N2). Combining this with property (v), eij(k) converges to a point in mn max f0o
[Lij ,oi I f i, ., N and j=l, ., m.
As discussed in Chapter II, properties (ii)(iv) and (vi) are standard properties required for any parameter estimator used in indirect adaptive control. Property (v) forces the parameters to converge to a subset n of ]RNxm which contains no singular points. This additional property will ensure that the estimated system is uniformly stabilizable thus eliminating the need for a persistently exciting input. Property (vii), the convergence of eij(k) to a point in
omi n max i o rqie
ein'6 iJ' is not required for proving global stability of the
adaptive closedloop system. Property (vii) will be used in Chapter VI for tracking and disturbance rejection based on the internal model principle. It is interesting that unlike the case of least squares where e(k) converges to 0 + PP'1(o)y'(O) as k + , there is no closedform expression for the limit of e(k) as k +  when using the parameter estimator described by (42)(45). However, if P(k) converges to zero, then e(k) converges to the true system parameter matrix 8 for both the parameter estimator discussed in this chapter and the recursive leastsquares estimator. This is easily shown using the recursive solution of (410) which is given by
n1
e(n) = P(n)PI(0)"(0)  P(n) p1 (i+l)P(i)f(e(i)).
i=O
Taking the norm of both sides and using the properties of the induced matrix norm, it follows that
35
ii'(n) n < nP(n) ii nP1 (0)"6(0) + iP(n) it n f P1 (i+1)P(i )f(e(i )) .
i =0
nr"
It was shown previously that ii I P1(i+1)P(i)f(e(i))ii is bounded in i=O
the limit as n + =. Also P(n)i can be made arbitrarily small for n sufficiently large if P(k) converges to zero. Thus, 1'(n) converges to zero as n +  which implies O(k) converges to the true system parameter matrix e whenever P(k) converges to zero.
CHAPTER V
ADAPTIVE REGULATOR
The rinput moutput linear discretetime system described by
(31)(33) will be regulated using a MIMO version of Samson and Fuchs' adaptive LQ controller [12]. As discussed in Chapter Two, a necessary condition for the adaptive closedloop system to be globally stable is a
uniformly stabilizable estimated system. It will be shown that the
parameter estimator described by equations (42)(45) will give a
uniformly stabilizable estimated system provided that assumptions A1A3 summarized at the end of Chapter Three are satisfied.
Uniform Stabilizability of the Estimated System
Using the parameter estimator described by (42)(45), it will be shown that the estimated system can be stabilized using state feedback. As shown in [6], the statespace observer form realization for the system described by (31) is given by
x(k+1) = Fx(k) + Gu(k) (51)
y(k) = Hx(k)
where
A1 I BI
~A2 B
F = G = . H [ [I 0 . 0]
An1 I Bn1
An 0 Bn
The entries in F and G depend on the unknown system parameter matrix e. In fact, by the definition of e in (33), F and G can be viewed as continuous functions of e where e ranges over the parameter
Nxm: min max
space si = feP : ij c6eij ei .} This dependance can be made explicit by denoting F and G as F(e) and G(e) respectively. Assumption A3 in Chapter Three is equivalent to requiring that the system described by (F(e),G(e)) is pointwise stabilizable over a.
Definition: The pair (F(e),G(e)) is pointwise stabilizable over P. if for each o belonging to a, there exists a feedback matrix L(e) such that Ikmax (F(e)G(e)L(e))I < q(e) < 1 where Xmax(.) is the eigenvalue of
() with the greatest magnitude.
It follows from known results C24], that pointwise stabilizability of (F(e),G(e)) over s is equivalent to the rank condition (34). Thus, stabilizability over a can be checked using the polynomial matrices D(z,e) and N(z,e) comprising the system transfer function. Given E > 0, let sie be an extension of the parameter space Q defined as
Nxm min max
n =fee~ IR I j eeo. e' e. +ell. (52)
S13 13 1
Proposition 5.1: Suppose assumption A3 is satisfied so that (F(e),G(e)) is pointwise stabilizable over n. Then there exists an
e > 0 such that (F(e),G(e)) is also pointwise stabilizable over Q. Proof: Suppose (F(O),G(e)) is stabilizable over Q. For each e
belonging to n, there exists a bounded open neighborhood V, of e such that (F(e),G(O)) is stabilizable over V. This result is proved in
[25], by first showing that there exists a feedback matrix L(e) with entries which are continuous in a such that
Amax (F(e)G(e)L(e))I < q(o) < 1. The existence of the open
neighborhood V6 of e then follows because the eigenvalues of (F(e)G(o)L(e)) are continuous functions of 0. Since a is a compact
subset of Nxm and U V8 is an open cover of Q, there exists a
r
finite subcover U V0 of n. For k=1,2,., let ek (1/2 )k. For
j=l j
every k > 1, a  U V is a compact subset of RNxm and
ï¿½k j=1
r r
UV C2 UV0
k+1 j=1 j k j=1 .
r
Suppose a  U Va is nonempty for all k > 1. By the properties of
ek j=1 j
nested compact sets (see [26, p.38]) there is at least one point p which
r
belongs to nCk  U V for all k > 1. Therefore, pen for all
ï¿½k j=1 j
r
k > 1 which implies that pe. However, pj U V which forms an
j=l 8j
open cover of s and hence we have a contradiction. Thus,
r
k j=1 Ve must be empty for k > k which implies (F(a),G(e))
will be pointwise stabilizable over Q k for any k ) k1.
Let e(k) be defined as the estimate at time k of the system parameter matrix 0, and let (F(k),G(k)) denote the timevarying estimated system (F(e(k)),G(e(k))).
Definition: The estimated system (F(k),G(k)) is uniformly stabilizable as a timevarying system [27], if there exists a bounded feedback matrix L(k) such that (F(k)G(k)L(k)) is uniformly asympototically stable as a
timevarying system matrix.
The following theorem gives conditions under which the estimated system is uniformly stabilizable.
Theorem 5.1: Assuming A1A3 are satisfied, the parameter estimator described by (42)(45) will generate an estimated system (F(k),G(k)) which is uniformly stabilizable as a timevarying system.
Proof: By Proposition 5.1, there exists e > 0 such that (F(e),G(e)) is pointwise stabilizable for all e belonging to si. Using the results in Kamen and Khargonekar [25], since a is compact, there exists a feedback matrix L(e) with entries which are continuous in e and a positive constant q < 1 such that
Imax (F(o)G(e)L(e))I < q < 1 for all eEn
Property (v) in Theorem 4.1 implies that there exists a finite N such that e(k) belongs to Q, for all k > N. It then follows from the
result in Kamen and Khargonekar given above that there exists a feedback gain sequence L(k) continuous in e such that
I max (F(k)G(k)L(k))I < q < I for all k > N.
Since the entries in (F(k)G(k)L(k)) are continuous functions of O, property (vi) in Theorem 4.1 implies [F(k)G(k)L(k)][F(k1)G(k1)L(k1)] converges to zero as k goes to infinity. Desoer [28] proves A(k) will be uniformly asymptotically stable as a system matrix if (i) all eigenvalues of A(k) have magnitude less than some constant q < 1 for all k and (ii) A(k)A(k1) converges to zero as k goes to infinity. Using Desoer's results on stability of slowly timevarying systems, (F(k)G(k)L(k)) is uniformly asymptotically stable as a system matrix. Thus the estimated system (F(k),G(k)) is uniformly stabilizable as a timevarying system.
The adaptive control law for regulating the system described by (31) is of the form u(k) = L(k)^(k). The state estimate X(k) is
generated from an adaptive observer which will be discussed in the next section. The feedback gain matrix L(k) is chosen to stabilize the estimated system (F(k),G(k)). The existence of a stabilizing feedback
gain L(k) is guaranteed by Theorem 5.1. Several possible options for L(k) are given in the section following the adaptive observer section.
41
Adaptive Observer
The adaptive observer for the estimated system is given by
x(k+l) = F(k)2(K) + G(k)u(k) + M(k)(y(k)9(k))
9(k) = H(k)
A,(k) A2(k) F(k)=
An (k)
M(k)
7Al(k)"
A2(k)
An(k)
G(k) =
Bl (k) B2 (k)
Bn(k)
(53)
H = [1 0 . . . 0].
The adaptive observer defined by (53) is a deadbeat observer. The
output 9(k) of the adaptive observer can be expressed in the following form:
A
y(k) = [Al(k1) . An(kn) Bl(k1) . Bn(kn)] (kl)
(54)
where
This expression is easily derived by first rewriting (53) as
A
0 I
Bl(k1) B2(k1)
Bn(kl)"
u(k1)
"Al(k1)
A2(k1)
An(k1)
y(k1).
(55)
Multiplying equation (55) by H gives
9(k) = Hx(k) = E0 1 0 . O]X(k1) + Bl(k1)u(k1)  Al(k1)y(k1) Substituting into this equation for X(k1) using (55) yields
(k) = CQ 0 I . O]R(k2) + B1(k1)u(k1) + B2(kl)u(k2)
 Al(k1)y(kI)  A2(k2)y(k2)
Continuing the substitution using (55) gives
y(k) = [0 . 0 I] (kn+l) +
n1
SBj (kj)u(kj) j=l
I Z A.(kj)y(kj) j=l J
n n
I B (kj)u(kj)  I A (kj)y(kj) j=1 j=1
which is equivalent to (54).
The error between the output of the adaptive observer (53) and the output of the system (31) as defined as e(k) = :(k)  y(k).
The observer error e(k) has the same properties as the prediction error e(k) given in Theorem 4.1; that is, lle(k)n < a'(k)Hj(k1)o + b'(k) (56)
where a'(k) and b'(k) are positive scalar sequences which converge to zero. This is shown by first rewriting e(k) as
e(k) = A(k)  ceT (kl)a (ki) T(k1)]
+ [OT(kl) a(ki) + T % (k1)]  y(k)
'(k)(k) (k)  e T(kl) a(k1) + T (k1)]  e(k).
Taking the norm of both sides and using the triangle inequality gives
ll(k)i 4 OM(k)  (T (kl)a (k1) + (kl))il + ile(k)ll
Noting from (31)(33) that e T(k1)a (k1) + T b(k1) is equal to [AI(k1) . An(k1) B1(k1) . 9 n(ki)](ki) and using expression (54) gives
l 1(k)ii 4 ii[J (k) . J2n(k)] (kl)iu + iie(k)ii (57)
< 1[ 1 (k) . 2n]J2n ii i (k1)ii + iie(k)ii
Ai (ki) + Ai (ki) li~n
where Ji(k) =
Bi (ki)  Bi (ki) n+1i<2n.
It follows from property (vi) in Theorem 4.1 that aJi(k) converges to zero as k goes to infinity for i=1,.,2n. Applying this property and property (iv) of Theorem 4.1 to (57) gives the desired inequality (56). Property (56) is very important in the proof of global stability of the adaptive closedloop system.
Feedback Gain Sequence
The feedback gain sequence L(k) must be chosen so that (F(k)G(k)L(k)) is exponentially stable as a timevarying system matrix. As shown in Theorem 5.1, a stabilizing feedback does exist because the estimated system (F(k),G(k)) generated from the parameter estimator (42)(45) is uniformly stabilizable under assumptions A1A3. There are several options in the literature for choosing L(k), some of which will be discussed in this section.
One approach for computing L(k) is to stabilize the estimated system (F(k),G(k)) pointwise in time using control law strategies for timeinvariant systems such as pole placement or LQ control. Of course, in the case of pole placement, the stabilizability assumption over the parameter space Q (assumption A3) must be strengthened to reachability
over n. This approach has been developed by a number of individuals, see for example [48]. A drawback to this approach is that (F(k),G(k))
may not be reachable or stabilizable at certain points in time (i.e., 0(k) may be a singular point for some values of k). Recall that
Proposition 5.1 and property (v) of Theorem 4.1 only guarantee that (F(k),G(k)) is stabilizable for all k exceeding some finite N. Thus,
for time points prior to N, the chosen control law may not have a solution. Therefore, in this approach it is necessary to check whether
or not (F(k),G(k)) is reachable or stabilizable at each time point k, which adds to the number of online calculations. Also, a decision must
be made on how to modify the control law when singular points are encountered. A common choice in the literature is to update the control law only when (F(k),G(k)) is stabilizable.
Another approach first introduced by Kreisselmeier for continuoustime systems is to compute the feedback gain asymptotically [29]. This
approach was developed for SISO discretetime systems by Samson and Fuchs [1012] using an LQ control strategy, and later extended to MIMO discetetime systems by Ossman and Kamen [3032].
In [3032], the stabilizing feedback L(k) is computed at each time point k by solving in real time one step of a Riccati difference equation. More precisely, let
L(k) = [GT(k)RkG(k) + I]IGT(k)RkF(k) (58)
where R(k) is the solution to the Riccati difference equation
Rk+1 = Q + LT(k)L(k) + (F(k)  G(k)L(k))TRk(F(k)  G(k)L(k)).
(59)
In (59), both Q and the initial value Ro are positive definite symmetric matrices. In contrast to the first approach, where the algebraic Riccati equation would be solved for each k, this asymptotic approach requires the solution of only one iteration of the Riccati difference equation at each point in time. Obviously, this approach
offers considerable savings in online calculations. Also, it is not
necessary to check whether or not (F(k),G(k)) is stabilizable at each point k because L(k) given by (58)(59) is welldefined even for isolated singular points. In order to prove that (F(k)G(k)L(k)) is exponentially stable when L(k) is computed using (58)(59), it is first necessary to show u(L(k)L(k1))u converges to zero as k + .
Theorem 5.2: Suppose the parameter estimates are generated using the algorithm (42)(45). Then subject to assumptions A1A3, the feedback gain sequence given by (58)(59) has the property
HL(k)L(kI)i + 0 as k + .
Proof: The proof is rather lengthy and will therefore be divided into steps. Throughout the proof, Pk() will refer to the sequence generated from the timeinvariant Riccati difference equation (RDE) given by
Pk+l(0): Q + FT (6)P(O)F(8)  FT(e)Pk(o)G()[I + GT (M)Pk()G(o)]I
x GT (O)Pk(e)F(e), PO > 0.
The matrix P(e) will denote the solution to the algebraic Riccati equation (ARE) given by
P.(8) = Q + FT(e)P.(e)F(8)  FT ()P (e)G(o)[I + GT (6)P (e)G(e)]1
x G T(e)P(e)F(e).
Let Q denote the compact parameter space defined in (52) such that
(F(e),G(e)) is pointwise stabilizable for all a belonging to Q. Let
> 0 be given. It will first be shown that HRk  Rklii < ï¿½ for k
sufficiently large where Rk is computed using (59). Step 1: The sequence of matrices Rk given by equation (59) is
bounded. That is, Rk belongs to a compact subset D of NxN.
Assumption A3 implies (FT (e),G ()) is detectable for all e belonging
to the parameter space a. Using arguments similar to those in Theorem 5.1, it follows easily that (FT(k),GT(k)) is uniformly detectable.
Anderson and Moore [27], prove that Rk and L(k) given by (58)(59) will be bounded if (FT(k),GT(k)) is uniformly detectable.
1 2
Step 2: Let Pk(e) and Pk(e) denote the sequences generated from the RDE using initial conditions P0 and P0 respectively. Defining
APk(e) : P1()  P2(e), it then follows that
APT 1 T 1 1 T
APk+l(0) = F (0)(I  Pk(e)G(B)[G (e)Pk(e)G(O) + I] G (e))
x(APk(e) + APk(O)G(B)G T(O)P2 (e)G(e) + I]I G ()APk())
x (I P1(e)G(e)G T(6)P1()G(e) + I1 G T()) TF(e).
I keGe[T Ik
This relationship was introduced by Samson [10] in the singleinput
case. It can also be verified in the multiinput case, but due to the large amount of algebra involved the derivation will not be included here.
Step 3: For any PO > 0 there exists an N(e,Po) such that
liPk(e)  P=(e)il < E/6 for all Oe n and for all k > N.
Since (F(e),G(e)) is stabilizable for each o belonging to Q, it follows from wellknown results that Pk(e) converges to P.(e) pointwise over
It was shown in Green and Kamen ([33], Theorem 1) that "the
convergence is uniform over a. if n. is compact. Step 4: For all PO belonging to a compact subset D of NxN, there
exists an N(s) such that
OPN()  P=(e)u < e/6 for all Oet^.
ï¿½ m M
The compactness of 0 implies there exist matrices P0 and P0 such that
pm < p0 . M for all PO belonging to D. It follows from the
0 P0
relationship given in step two that
pm PM fralPO
Pk (e) < Pk(e) < Pk(e) for all PonD.
From step 3, there exist integers NI(EP ) and N2(e,PM) such that
IPk(e)  P.(e)ii < s/6 for all k > N1 and for all oe^;
lip (e)  P=(e)u < e/6 for all k > N2 and for all Bel .
Thus,
llPk(Q)  P(o)i < E/6 for all eï¿½12, for all Po O, and for all k > max(N1,N2).
Step 5:
for all
Given a > 0, there exists a 6(a) such that for all PkD and
A
lipk+1(8)  Pk+1( )Q < a
whenever lie  ^en < 6
This property follows easily from the fact that Pk+1(6) is a continuous function of e and the sets D and n, are compact.
e
Step 6: Let N be a fixed finite integer such that the condition in step four is satisfied. There exists a finite integer Ml(e) such that
nRk+N  Pkii < e/3 for all k > M
where Pk is the solution to the ARE when the matrices are evaluated at S(k). In order to prove this result, a family of sequences Pt follows
F(.) and G(.) is defined as
t
Pk = Rk for all k 4 t. Pk+1 = Q+FT ((t))P tF((t))FT ((t))pt G((t))[I + GT (e(t))P G(O(t))] 1
x GT(e(t))PtF(e(t)) for all k > t.
In other words, P is equal to Rk for all k up to time t. After time
t P is generated from the timeinvariant Riccati difference equation
,Pk
initialized by Rt and using the constant matrices F(.) and G(.) evaluated at e(t). For every 6(t) belonging to n , the sequence Pt tk
will be bounded for all k. In fact, Pk converges to P t whenever
k Ck
O(t) en,. It will first be shown that iR Pk+Ni < k /6 for k
Sk+N k+N< 1 fo k
sufficiently large. The matrix P k is simply the Nth step of a timek+N
invariant RDE intitialized at time k by Rk and using constant matrices F(.) and G(.) evaluated at e(k). The interpretation for
uRk+N Pku being small for k sufficiently large is that the timevarying RDE given by (59) will not vary much from the timeinvariant RDE over a finite interval of time N as long as e(k) does not change
much over the time interval N. Equation (59) can be rewritten as
R k+i+1 =Q+ F T ((k))Rk+iF(o(k))  F T(o(k))Rk+iG(e(k))
x [I + G T(o(k))Rk+iG(o(k))]"1 GT ((k))Rk+i F(e(k)) + Eki; Rk
for all k and for all ie[O,N].
The matrix Ek+i is given by
Ek+i FT (O(k+i))Rk+iF(e(k+i))  FT (e(k+i))Rk+i G(o(k+i)) x [I + GT (O(k+i))Rk+iG(o(k+i))]I GT ((k+i))Rk+iF(O(k+i))
 FT(o(k))Rk+iF(O(k)) + FT(o(k))Rk+iG(o(k))
x [I + GT(e(k))Rk+iG(e(k))]"1 GT(O(k))Rk+iF(o(k)) for all k and for all iECO,N].
Since Rk belongs to the compact set D for all k, it follows from step five that given any a > 0 there exists s(a) such that
,1Zk+i o < a whenever ie(k+i)  o(k)ii < 6 for ie[O,N].
k Qk Let Q"k =Rk  k
k
Qk+i +1
Using the equation introduced in step two gives
F T(e(k))(I  Rk+iG(e(k))[G T(e(k))Rk+iG(e(k)) + I]  16(k)))
k k T k
x (Qk+i + Qk+i G(o(k))[GT(o(k))Pk+iG(e(R)) + I]
T kT
x G T(e(k))Qk+i) (I  Rk+iG(e(k))G T(e(k))Rk+iG(e(k)) + I]T T kT k Q O
x GT(e(k) ))T + Zki ~k = 0 ie[0,N].
It follows from the boundedness of e(k), Rk, and sufficiently large) that
(for t
iF T(e(k))(I .Rk+iG(e(k))G T(e(k).)Rk+iG(e(k)) + 1]iGT(e(k)))II < KI< O and IG(e(k))[GT(e(k))Pk+i G(O(k)) + ] G T(e(k))N < K2 < . Thus for all k such that e(k)en,, we have
k K12 k i+ K2 k 0 2 + 1 k , ie0,N].
nQk+i+lu < K(Qk+i 2Qk+i k+i k
Clearly, there exists a al(e) such that if oE k+i 11 < 1 for all ie[O,N] then
k
Qk+N < C/6.
Let 6(a1) be chosen such that
HEk+iU < a 1for all ie[O,N] whenever ne(k+i)  e(k): < 6.
It follows from properties (v) and (vi) of Theorem 4.1 that there exists a finite integer M1 such that
el(k)gen for all k > M1
and ne(k+i)  e(k)n < 6(a ) for all ic [0,N] and for all k > M .
Therefore
k K
Qk+N =nRk+N  Pk+Nn < e/6 for all k > M1.
Using the results in step four, we obtain
k p
HPk+N k < e/6 for all k > M1. Thus
uRk+N  k IRk+N  Pk N + iP k k 1 < C/3 for all k > M1.
00k+N k+N  P0 10
which proves the result.
Step 7: There exists a finite integer M2(e) such that
0pk+l k H < E/3 for all k > M
The result follows immediately from the uniform continuity of P (.) over ^ and the convergence of Iie(k+i)  e(k)n to zero. Step 8: There exists a finite integer M(e) such that
nRk  R k1 < e for all k > M.
Applying the triangle inequality and using the results in steps six and seven gives
IRk+N+I  Rk+Nn 4 oRk+N+1  pk+lI + npk+li k 11 + lipk Rk+N1
for all k > max (M1, M2).
The result then follows easily by setting M = max(M1, M2) + N + 1. Step 9: If L(k) is computed using (58)(59), then HL(k) L(kl)ll converge to zero as k goes to infinity. This result follows easily
since fIRk  Rkl1l, IF(k)  F(k1) ii, and iG(k)  G(k)i all converge to zero.
Using the results of Theorem 5.2, it is now possible to prove that (F(k)G(k)L(k)) is exponentially stable as a timevarying system matrix. The exponential stability of (F(k)G(k)L(k)) is very important
for proving global stability of the adaptive closedloop system.
Theorem 5.3: Suppose e(k) is generated using the parameter estimator (42)(45) and the feedback gain sequence L(k) is computed using (58)(59). Then subject to assumptions A1A3, (F(k)G(k)L(k)) is
exponentially stable as a timevarying system matrix. Proof: Assumption A3 implies that (FT (),G T(e)) is detectable for each e belonging to Q. Using arguments similar to those in
Proposition 5.1 ind Theorem 5.1, it follows that (FT(k),GT(k)) is
uniformly detectable. In [27], Anderson and Moore prove that if
(FT(k),GT(k)) is uniformly detectable then (F(k)G(k)L(k))T is exponentially stable as a timevarying system matrix where L(k) is given by (58)(59). Since matrix products do not commute, it does not immediately follow that (F(k)G(k)L(k)) is exponentially stable. However, Samson and Fuchs [12] show A(k) will be exponentially stable as a timevarying system matrix if AT(k) is exponentially stable and if IA(k)A(k1)n converges to zero as k + . Since (F(k)G(k)L(k))T is exponentially stable, it then follows that (F(k)G(k)L(k)) will also be exponentially stable if u(F(k)G(k)L(k))  (F(k1)G(k1)L(k1)), converges to zero as k + . Property (vi) of Theorem 4.1 implies both IF(k)F(k1)ii and oG(k)G(k1)II converge to zero. It was shown in Theorem 5.2 that llL(k)L(k1)n also converges to zero. Since
F(k)G(k)L(k)(F(k1)G(k1)L(k1)) =
(F(k)F(k1))  G(k)(L(k)L(k1))  (G(k)(G(k1))L(k1) and both G(k) and L(k) are bounded, it follows that I(F(k)G(k)L(k))(F(k1)G(k1)L(k1))n does indeed converge to zero. Thus the matrix (F(k)G(k)L(k)) is exponentially stable as a timevarying system matrix.
Stability of the Adaptive ClosedLoop System
The control law chosen to regulate the system is given by
u(k) = L(k) (k) (510)
where x(k) is generated from the adaptive observer (53) and L(K) is a stabilizing feedback for the estimated system (F(k),G(k)). The following theorem shows that the adaptive regulator consisting of the
observer (53) and the control law (510) results in a globally stable closedloop system; that is, for any initial states in the plant and the observer, the input u(k) and the output y(k) converge to zero.
Theorem 5.4: Suppose that the parameter estimator (42)(45) is used so that there is a stabilizing feedback L(k) for the estimated system (F(k),G(k)) subject to assumptions A1A3. Then with the adaptive
regulator defined by (53) and (510), the resulting closedloop system is globally stable.
Proof: The proof is based on a MIMO extension of the results in [121. Letting (k) = (k)y(k) and using equations (53) and (510) gives
A(k+1) = (F(k)G(k)L(k)) (k)  M(k)A(k) y(k) = H%(k)  e(k).
Also, the regression vector 4(k) defined in (32) can be written in the form
(k) = S4(k1)
where
0 . 0 0 . 0
Im(n1)
S = 0 . 0 0 . 0
0
Ir(n1) :
0
Defining z(k+l) =A(kl gives
+ D(k) (k) + V(k)
H 0
D(R) = L(k)
, v(k) =
z(k+l) = A(k)z(k) + w(k)
e(k)
0 0
0
(511)
where
S D(k) v 1
A(k) = w(k) ) j
0 F(k)G(R)L(K) M(R)e(R)
The matrix S is a stable matrix since it is lowerblock triangular. The
matrix F(k)G(k)L(k) is stable by construction; therefore, A(k) in (511) is exponentially stable as a timevarying system matrix. In
addition, it follows easily from (56) and the boundedness of M(k) that
Iw(k)n < a(k)uz(k)n +b(k)
(512)
where a(k) and b(k) are positive scalar sequences which converge to zero as k + . Thus, (511) can be viewed as an exponentially stable timevarying system driven by an input w(k) which can grow no faster than linearly with the state, z(k). Following the proof in [12], it will be shown that Iiz(k) a converges to zero. Let A(k+N,k) denote the state transition matrix for (511) defined by
A(k+M,k) =
M1 11 A(k+i) i =0*
Since A(k) is an exponentially stable timevarying system matrix, it follows that
iiA(k+M,k)t < R 1 <  for all k,M > 0.
(513)
Also, given 0<6<1 there exists a finite p such that
nA(k+p,k)i < 6 for all k.
(514)
For the remainder of the proof, p will be a fixed constant such that(514) is satisfied. Taking the norm of both sides of (511) and using (512) gives
liz(k+1) e ItA(k)aItz(k) ii + nw(k) ii
(515)
(A(k)u + a(k))uz(k)ii + b(k)
Noting that HA(k)n, i(k), and E(k) are all bounded and p is fixed, it follows from equation (515) that there exist finite constants R2 and R3 such that
ttz(k+i)ii < R2z(k)ti + R3 for all is [O,p] and for all k. Solving equation (511) recursively gives
(516)
z(k+p) = A(k+p,k)z(k) +
p1 I A(k+p,k+j+i)w(k+j). j=0
Taking the norm of both sides and using (512) results in
iiz(k+p)i 4 1IA(k+p,k)lh iz(k)3
4 nA(k+p,k) li iiz(k)ii
p1
+ I fIA(k+p,k+j+1)nI iw(k+j)1u j=0
p1 + I nA(k+p,k+j+1)ii(a(k+j)ifz(k+j)ii + b(kj)). j=0
It follows from equations (513), (514), and (516) that
P p1
Iiz(k+p)ii .4 az(k)ii + max(1,Rj) I a(k+j)(R2 iz(k)ii + R3) + b(k+j) for j=0
all k.
Since a(k) converges to zero, there exists a finite N such that
(517)
i(k) < I p for all k > N.
6 + PR2max(1,R,)
Therefore, for all k > N,
~,pi
uz(k+p)i < roz(k)i + max(1,R1) P a(k+j)R3 +b(k+j) or equivalently, j=0
iz(k+p)ii 4 riiz(k)n + c(k)
(518)
where r < 1 and c(k) converges to zero as k + o. In order to show
(518) implies liz(k)o converges to zero, let e>O be given. Solving (518) recursively and using (516) gives
n1 j
iiz(k+i+np)ii rn iz(k+i)i + I rn c(k+i+jp) j=0
n1 n1j
iz(k+i+np)ii 4 r (R2iz(k)ii + R3) + I r c(k+i+jp) j=0
for all k > N, iE[O,p]. (519)
Since c(k) converges to zero, there exists a finite N, > N such that c(k) < for all k > N1.
(177
Applying this inequality to (519) gives nz(Nl+i+np)i; < rn(R2iz(N1)i + R3) + e/2, ie[O,p]. Also, since r < 1 there exists a finite N2 such that
r (R nz(N) + R) for all n > N2.
Thus
Hz(N1+i+np)ii < e for all n > N2, iï¿½[O,p]
which implies tz(k)ii < e for all k > N1 + N2P. By the definition of z(k), both the system input u(k) = L(k) (k) and the system output y(k) must converge to zero. Therefore the adaptive
60
closedloop system is globally stable using the adaptive regulator defined by (53) and (510).
CHAPTER VI
APPLICATION TO TRACKING AND DISTURBANCE REJECTION
The adaptive controller derived in Chapter V can. be applied to the problem of tracking with disturbance rejection using the internal model principle. The internal model principle has been discussed by a number of individuals, see for example [20].
Consider the rinput moutput plant described by
x(k+l) = Fx(k) + Gu(k) + Dv(k) (61)
y(k) = Hx(k)
where
A1 I B1
 A 2 18
A2I B2
F . . G: . H rI 0 . 0].
An_1 I Bn_1
An 0 Bn
The vector v(k) consists of the exogenous disturbance signals. As in
previous chapters, it is assumed that some or all of the entries in the matrices A. and B. are unknown and will be estimated using the parameter
estimator described by (42)(45). Letting r(k) denote an mvector
reference signal, the objective is to design an adaptive controller which stabilizes the system described by (61) and forces the tracking error e(k) = y(k)r(k) to converge to zero as k + .
As mentioned, the controller design will be based on the internal model principle. Assume that the disturbance vector v(k) and the reference signal r(k) both satisfy the difference equation:
z(k+q) + aq1z(k+q1) + . + aIZ(k+l) + a0z(k) = 0.
(62)
Let r be defined as
al
aq.1
The internal model can then be realized as
Xc(k+l) = AcXc(k) + Bce(k)
(63)
where Ac = block diagonal {r,
r *., r} mtupl e
Bc block diagonal {T, T tuple T
The following theorem gives the conditions under which it is possible to
design an adaptive controller which will stabilize the plant described
by (61) and will drive the tracking error e(k) to zero asymptotically.
Theorem 6.1: Suppose assumptions A1A3 are satisfied and in addition
zIF(e) 0 G(o)
rank = n+mq for all een and
BCH zIAc 0 for all z: Izi > 1
Nxmmi n max
where { = {s R e j ,6i j. Let xc(k) be generated from the
error driven system described in (63) and X(k) be generated from the adaptive observer given by (53). Then there exists a control law u(k)
Llx(k)L2xc(k) which when applied to the plant (61) results in a stable closedloop system and drives the tracking error e(k) = y(k) r(k) to zero.
Proof: Using equations (53) and (63) and setting v(k) and r(k) equal to zero, we obtain
[x(k+1) 1 F(k) [ x(k) 1 G(k)] [M(k)]
[c(k+) L BcH A [xc(kJ+ u(k) + LBcje(k) (64)
where e(k) = H'(k)  y(k). Assuming the given rank condition holds, it follows from property (v) of Theorem 4.1 that the system described by (64) is stabilizable. The stabilizing control law is of the form
u(k) = Ll(k)A(k) L2Xc(k). (65)
Let F(k) : and G(k) = [
BcH Ac r
As in the case of the adaptive regulator derived in Chapter V,a
stabilizing feedback gain L(k) = [L1(k) L2(k)] can be computed from one step of the Riccati difference equation
L(k) = [ T(k)Rk'G(k) + I"1 GT(k)RkF(k)
(66)
Rk+1 = Q + LT(k)L(k) + (F(k)6(k)L(k))TR k((k)r^(k)L(k)).
Again, Ro and Q are arbitrary positive definite symmetric matrices.
In order to use the existing results on tracking and disturbance rejection, the closedloop system consisting of (64) and the control law (65) must converge to a timeinvariant system. Properties (v) and (vii) of Theorem 4.1 imply that the pair (F(k),G(k)) converges to a stabilizable pair (F.,G). Using the results of Samson [o], L(k)
converges to L., the optimal LQ feedback for the pair (F.,G) given by L = [GTR G. + I]" GT RF
where the matrix R. is the solution to the ARE for (Fw,G.) . More
precisely,
R = Q + LTL + (FTG L )TR.(F .)
Applying the control law (65) to the system described results in an exponentially stable system driven by e(k).
results in Chapter Five, the observer error e(k) converges Also, it follows from property (ii) of Theorem 4.1 that bounded. Thus, the closedloop system consisting of (64)
by (64) From the to zero. M(k) is and the
control law u(k) = Llx(k)L2xc(k) is exponentially stable and converges to a timeinvariant system. Finally, from the results in Chen [20], if the control law (65) is applied to the plant (61), the resulting closedloop system is still stable (not including the exogenous disturbance v(k)) and the system output y(k) converges to the reference signal r(k) as k +The adaptive controller described by (63) and (65)(66) causes the plant (61) to reject deterministic disturbances and to track a given reference signal r(k). Simulations of the adaptive controller for various types of external disturbance and reference signals are included in Chapter VII.
CHAPTER VII
SIMULATIONS
This chapter contains several simulations of both the adaptive regulator derived in Chapter V and the adaptive controller discussed in Chapter Six applied to S150 as well as MIMO discretetime systems.
Example 7.1: Consider the nonminimum phase discretetime system described by the following difference equation
y(k) = 2y(k1)O.99y(k2)+O.5u(k1)+3u(k2).
The adaptive regulator defined by the observer (53) and the control law (510) with L(k) computed using (58)(59) was implemented. Three
different algorithms were used to estimate the system parameters: the
estimation scheme described by (42)(45), the recursive leastsquares algorithm defined by (23), and the estimation scheme given by (42)(45) with data normalization (i.e., ok1 = max(l, i, (kl) Hi) for all k). In all three cases, the initial "covariance" P(O) was chosen to be I and the initial state of the plant was [I 0jT. Also, Q and Roin
(59) were chosen to be I. The parameter ranges, initial parameter
estimates and steady state estimates for all three estimation schemes are displayed in Table 71. As seen from Table 71, the estimation
scheme described by (42)(45) forces the estimates of the parameters to converge to the given ranges; whereas, recursive least squares does
a, a2 bi b2
Actual Parameters 2.0 0.99 0.5 3.0
Initial Estimates 1.95 1.0 0.75 6.0
Parameter Ranges [1.9, 2.0] C2.0, 4.0] C0.5, 2.01 C2.5, 9.5]1
Estimation Scheme 2.00 1.00 .0.530 3.04
(42)(45)
Recursive Least 2.11 1.05 0.594 3.19
Squares
Estimation Scheme 2.00 0.991 0.571 3.14
(42)(45) with
ax(1,ie(k)ii)
nk = m
not. Figures 7173 contain plots of the output response, control input, and estimated parameters for each of the three cases. A
comparison of Figures 71 and 72 show the estimation algorithm defined by (42)(45) has a slightly better transient response than least squares. Figure 73 illustrates the detrimental effect of the data normalization on the transient response of the system. As previously
mentioned, in (45) should be chosen very small to minimize the effect of data normalization on the system transient response.
Table 71
DATA
~I P.
 i i'i. *1
20 30 40 50 60 70 ' k
It iiIiI I.'
I'
 ,***~~ ~ ii
ii III II
20 30 40 50 60 7o0 {k
Figure 71
Estimation Scheme (42)(45)
y(k)
20 15 10
5
5
10
15
20
u(k)
20 15 10
5
5
10
15
20
al(k)
b2 (k)
b1 (k)
20 30 40 50 60 70 '(k(d)
Figure 71 continued
Estimation Scheme (42)(45)
20
15 10
5
5
10
15
20
u(k)
20 15
10
5
5
10
15
20
30 4om 5o0o T (b)
Figure 72
Recursive Least Squares
I i I I
11 ' j,, , . . I I lI ,,
Jill 'i
al(k)
_ 20 30 40 50 _a2(k o
70 (k)
b2(k) bl(k)
20 30 40 50 6070(k)
Figure 72 continued Recursive Least Squares
20 30 40 50 60 70 (k)
(a)
1114 It *.
Ii' ~ IJ'~~20
30 40 50 60 70 (k)
Figure 73
Estimation Scheme (42)(45)
with nk= max(I, IVP(k)II )
y(k)
15
10
5
5
10
15
20
u(k)
I
Example 7.2: Consider the nonminimum phase stabilizable but not
reachable discretetime system described by
y(k) = 2.6y(k1)2.13y(k2)+0.54y(k3)+u(kl)+1.5u(k2)u(k3).
The adaptive controller defined by the adaptive observer (53) and the control law (65) with L(k) computed from (66) was implemented to force the output to track the reference signal r(k) = 10. No disturbance was introduced in this example. The internal model was given by
Xc(k+l) = Xc(k) 4. e(k)
with e(k) = y(k)10. The parameters were estimated using (42)(45). The initial conditions were chosen to be: P(O) = I, x(O) = [0 1 O]T, xc(O)=O, and Ro=I. The initial parameter estimates, parameter ranges,
and steady state values of the estimates are displayed in Table 72. Plots of the output response, control input, and parameter estimates are shown in Figure 74.
Equation (44) indicates that the initial "covariance" matrix P(O) should satisfy O
y(k)=6.5xlO8 and the control input was u(k)=2.6x1O11. The parameter
estimates at k=20 were al(k)=451, a2(k)=943, a3(k)=484, bl(k)=1.0, b2(k)=447, and b3(k)=898. Thus, choosing P(O) smaller than 21 is not
merely a technicality in the proof of Theorem 4.1, but is a necessary condition for preserving stability of the adaptive closedloop system.
Tabl e 72 DATA
_ _ _ _ _ a, a2 a3J b, b2 j b3
Actual Parameters 2.6 2.13 0.54f . 1.5 1.0
Initial Estimates 2.5 1.0 0.5f 2.0 1.6 r 1.5
Parameter Ranges [2.5,2.8] C5,31 C2,1J C4,8] C1.5,1.74 C3,01
Estimation Scheme 2.5 1.92 0.429 0.9991 1.60 0.795
(42)(45) ______ ___________
y(k)
40 30 [ I Il
11101 20 30 40 50 60 70 (k)
IOi
20
30
40
(a)
20 30 40 50 60 70 (k)
(b)
Figure 74
Estimation Scheme (42)(45)
40 30
20 10
10
20
30
40
a.,(k)
a 3 ( k)
20 30 40 50 60 70
a 2(k)
b 2(Ck)
U ir A
20 30 40 50 60 70 b 3(k)
(d)
Figure 74 continued
Estimation Scheme (42)(45)
Example 7.3: Consider the discretetime system described by
y(k) = 2y(k1)5y(k2)u(k1)0.5u(k2)
The adaptive controller defined by the observer (53) and the control law (65) with L(k) computed from (66) was implemented to force the output to track the reference signal r(k)=5. A step disturbance defined
by
[ O]T K < 40
V(k) [3 O]T K > 40
was introduced. The internal model was given by
Xc(k+l) = xc(k) + e(k)
with e(k)=y(k)5. The parameters were estimated using the algorithm defined by (42)(45). The initial conditions were chosen to be: P(O)=I, x(O)=[1 O]T, Xc(O)=O, Ro=I. The initial parameter estimates, parameter ranges, and steady state estimated values are displayed in Table 73. The output response, control input, and parameter estimates are shown in Figure 75. For comparison purposes, plots of the output responses and control input when the parameters are known exactly are shown in Figure 76. As expected, the transient response is considerably worse when the system parameters must be estimated.
However, once the system is in steady state, the adaptive system seems to respond equally well to a sudden step disturbance as the system using exact parameters.
Table 73 DATA
a, a2 bI b2
Actual Parameters 2.0 5.0 1.0 0.5
Initial Estimates 0.5 6.0 1.0 0.0
Parameter Ranges [3, 21 [8, 4] C1.1, 0.9] [1, 1.0]
Estimation Scheme 1.64 4.3 0.90 0.478
(42)(45)
20 40 60 80 100 120
140 (k)
u(k
80 60
40 20
 fj ' :"I .
60 .
80
(b)
Figure 75 Estimation Scheme (42)(45)
20 15 10
5
5
10
15
20
,. 20 40
..
60 80 100 120
z 4U 60 80 100 120
I 4 b 2(k) 140 bl(k)
(d)
Figure 75 continued
Estimation Scheme (42)(45)
14n
al(k)
a2(k)
. l
140
y(k)
20 15 10
.,
_________________________ II
~ I,
is
20 40 60 80 100
140 (k)
(a).
uk
. 20 40'.60
100
140 (k)
Figure 76
Exact Parameters
5
10
15
20
20
40
60
80
Example 7.4: Consider the nonminimum phase discretetime system
described by
y(k)=2y(k1)0.99y(k2)+0.5u(k1)+3u(k2)
The adaptive controller defined by the observer (53) and the control
law (65) with L(k) computed using (66) was implemented to force the system output to track the reference signal r(k)=5 and reject the sinusoidal disturbance v(k)=2sin (7rk/2). The internal model was given by
Xc(k+1) 0 0 J Xc(k) + 0 e(k)
with e(k) = y(k)5. The system parameters were estimated using the
algorithm (42)(45). The initial conditions were chosen to be P(O)=I, x(O)= [1 O]T, xc(O)=[O 0 O]T, and Ro=I. The initial parameter estimates, parameter ranges, and steady state estimated values are displayed in Table 74. Figure 77 includes plots of the system output response, control input, and estimated system parameters. For comparison purposes, both the system output and control input are shown in Figure 78 for the case when the parameters are known exactly.
Table 74 DATA
a, a2 b1 b2
Actual Parameters 2 0.99 0.5 3.0
Initial Estimates 1.95 1.0 0.75 5.0
Parameter Ranges ï¿½1.9, 2.0] C2.0, 4.0] C0.5, 2.0] [2.5, 7.5]
Estimation Scheme 1.91 0.920 0.406 2.83
(42)(45)
I,,. 
30 40 50 60 70 (k)
40 50 60 70 (k)
(b)
Figure 77
Estimation Scheme (42)(45)
A~k
A^
I I. III. I
ii I.
I i
I.,,
 .i I .
10 11'I t20
Iit
30 
40
u (I )
10
20
30
40
. i ll ', ". 2 ' 3. . 0o
t 1
~ U1
20 3 4n0 50 60 7 0
b.,(k)
h. (ic
10 20 30 40 50 60 70
(d)
Figure 77 continued
Estimation Scheme (42)(45)
F
*1
 Ii .
I
f f, I
3
I.U
b W
20 30 40 50 60 1 70 (k)
20 30 40 50 60 70 (k)
Figure 78
Exact Parameters
y(k)
40 30 20 10
10
20
30
40
I ~
" LU
u(k)
10
i
Example 7.5: Consider the single input two output discretetime system described by
y(k) al y (k)1 + 3u(k1)
_y2(k) 1 2 Y2 (k1) bj\
where a and b are unknown parameters. As discussed in Chapter Three,
the required computations for parameter estimation can be reduced by separating the known parameters from the unknown parameters. This is
accomplished by rewriting the system equations as
y(k) = T, a(ki) + *T b(k1)
with
T 'Y k1 * 1 0 3' [y1(k1)'1
F (k1) = 9 T % (k1) y2(k1)I.
ab) Lu(k1) ] 2 0 L u(k1)j
The adaptive regulator defined by the observer (53) and the control law (510) with L(k) computed using (58)(59) was implemented. The system parameter matrix e was estimated using the algorithm (42)(45). The
initial conditions were chosen to be: P(O)=I, x(O)=[1 1]T, and Ro=I.
The initial estimates, parameter ranges, and steady state estimates are
given in Table 75. Plots of the output response, control input, and parameter estimates are shown in Figure 79.
a b
Actual Parameters 4.0 1.0
Initial Estimates 6.0 1.0
Parameter Ranges [2.0, 10.01 E4.0, 2.01
Estimation Scheme 3.981 1.001
(42)(45)
Table 75
DATA
yl(k)
40 30
20 L 10L
10
20
30
40
Y2 (k)
40 30 20 10
10
20
30
40
10 20 30 40 50 60 70 (k)
(b)
Figure 79
Estimation Scheme (42)(45)
10 20 30 40 50 60 70 (k)
(a)
20 30 40 50 60 70 (k)
80 60 40 20
20
40
60
80
8
6
4
2
2
4
6
8
" 10 20 30 40 50 60 70 (k)
(d)
Figure 79 continued Estimation Scheme (42)(45)
u(k)
a(k)
b(k)

Example 7.6: Consider the multiinput multioutput discretetime system described by
E k)1 [I a,21[y1 (ki) +1 [ b12 1 ul (ki)1
Y2(k) L1 2 J Y2 (ki) b21 b22 u2 (ki)
The order of required computations for parameter estimation is reduced by rewriting the system in the form
y(k) T Ta (ki) + T b(kI)
with
OT [a12 0 b12 , iaT(k.1) = Y2(ki) ul(k1) u2(kl)],
0 b21 b22
= [ i 1 1bT(k1) : Eyl(k1) Y2(k1) ul(k1)].
1 2 0
The adaptive regulator defined by the observer (53) and the control law
(510) with L(k) computed using (58)(59) was applied to the system. The algorithm given by (42)(45) was used to estimate the system parameter matrix e. The initial conditions for the simulation were: P(O)=I, x(O)=[1 O]T, and Ro=I. The initial parameter estimates,
parameter ranges, and steady state estimates are displayed in Table 76. Figure 710 contains plots of the output response, control input, and parameter estimates.
Tabl e 76
DATA
____________12_ a2121 2
Actual Parameters 0.75 2.0 3.0 0.5
Initial Estimates 0.65 1.0 2.5 1.0
Parameter Ranges ï¿½0.5, 0.8] [5.0, 3.0] [1.0, 4.0] [1.0, 3.0)
Estimation Scheme 0.785 1.58 2.97 0.320
(42)(45) 1
y1 (k)
r
10
5
5
101
20 30 40 50 60 70 (k)
(a)
Y2 (k)
10
5
5
10
1'i 10 j i
20 30 40 50 60 70 (k)
(b)
Figure 710
Estimation Scheme (42)(45)
Ul(k) 10
5
10 u2 (k)
20 30 40 50 60 70 (k)
(d)
Figure 710 continued
Estimation Scheme (42)(45)
20 30 40 50 60 70 (k)
(c)
,., "10
5
10
IIm .
(k)
77=
10 20 30 40 50 60 70 b 22 (k)
M
10 20 30 40 50 60 70
(e)
Figure 710 continued
Estimation Scheme (42)(45)
CHAPTER VIII
OI.SCUSSION
A globally stable adaptive LQ controller which does not require persistent excitation was introduced for multiinput multioutput linear
discretetime systems. The assumptions made on the plant were (1) an upper bound on system order is known, (2) the unknown system parameters belong to known bounded intervals, and (3) the plant is stabilizable for
all values of the unknown parameters ranging over the known bounded intervals. When applied to the unknown plant using this parameter estimator, the adaptive LQ controller ensures the system inputs and outputs will remain bounded, and forces the output to track a given reference signal in the presence of a deterministic external disturbance. Some remaining considerations include robustness,
application to timevarying or nonlinear systems, and analysis of transient response.
In many control applications, the order of the model will be lower than that of the plant. As mentioned in the introduction, Rohrs et al. [13141 demonstrated that most adaptive controllers could go unstable if the order of the system was underestimated, even if the modelling errors were small. Since knowledge of an upper bound on system order is one of
the assumptions made on the adaptive controller presented in this dissertation, it is likely that this controller will also go unstable for systems with unmodelled dynamics. There have been several
approaches in the literature towards development of robust adaptive controllers, some of which will be briefly discussed here.

Full Text 
91
Example 7.6: Consider the multiinput multioutput discretetime system
described by
Vl(k)'
'l
"a12
*1
(k1)*
1
b12
ux (k1)
y2
1
2
/2
(k1)
b2i
b22 _
u2 (k1)
The order of required computations for parameter estimation is reduced
by rewriting the system in the form
y(k) eTa(kl) + iT<(b( k1)
with
9
T =
1
1
0
2
>12
)22
1
0
, <>aT(kl) = [y2(kl) u1(kl) u2(kl)],
, 4>bT(k1) = Cyx(kl) y2(k1) u^k1)].
The adaptive regulator defined by the observer (53) and the control law
(510) with L(k) computed using (58)(59) was applied to the system.
The algorithm given by (42)(45) was used to estimate the system
parameter matrix e. The initial conditions for the simulation were:
P(0)=I, x(0)=[l 0]^, and RQ=I. The initial parameter estimates,
parameter ranges, and steady state estimates are displayed in Table
76. Figure 710 contains plots of the output response, control input,
and parameter estimates.
58
P1
z(k+p) = A(k+p,k)z(k) + J A(k+p,k+j+l)w(k+j).
j=0
Taking the norm of both sides and using (512) results in
P1
nz(k+p)n < iiA(k+p,k)n nz(k) n + Â£ nA(k+p,k+j+l) n nw(k+j)n
j=0
P1
< nA(k+p,k)n tiz(k) it + l nA(k+p,k+j+l) n(a(k+j) nz(k+j) 11 +b(k+j)).
j=0
It follows from equations (513), (514), and (516) that
p P1 _
I!z(k+p) ii < 6iiz(k) ii + max(l,R!j l a(k+j)(R9iiz(k) II + R~) + b(k+j) for
1 j=0
all k. (517)
Since I(k) converges to zero, there exists a finite N such that
a(k) <
1
6 + pRgdiaxil ,R^)
for all k > N.
Therefore, for all k > N,
p P1 _
iz(k+p)n < rnz(k)ii + max(l,RÂ¡) l a(k+j)R, + b(k+j) or equivalently.
1 j=0 J
iz(k+p)ii < rnz(k)n + c(k)
(518)
where r < 1 and c(k) converges to zero as k  . In order to show
(518) implies uz(k)n converges to zero, let e>0 be given. Solving
(518) recursively and using (516) gives
7
where
F =
"al
1
bl
a2
1
b2
G =
"anl
1
bnl
"an
0
bn
H = [1 0
0].
Samson and Fuchs assume that (F,G) is stabilizable. This assumption
allows for nonminimum phase systems and systems with stable common poles
and zeros. The adaptive IQ control law used by Samson and Fuchs [12] is
given by
u(k) = L(k)x(k), (25)
where u(k) is the input to the given plant, x(k) is the state estimate,
and L(k) is a stabilizing feedback gain for the estimated system (F(k),
fi(k)).
The state estimate x(k) is generated from an adaptive observer.
Letting F(k) and G(k) represent the current estimates for matrices F and
G, the adaptive observer for the system is given by
x(k+l) = F(k)$(k) + G(k)u(k) + M(k) (y(k)y(k)) (26)
y(k) = H?(k)
82
Example 7.4: Consider the nonminimum phase discretetime system
described by
y(k)=2y(kl)0.99y(k2)+0.5u(kl)+3u(k2)
The adaptive controller defined by the observer (53) and the control
law (65) with L(k) computed using (66) was implemented to force the
system output to track the reference signal r(k)=5 and reject the
sinusoidal disturbance v(k)=2sin (irk/2). The internal model was given
by
xc(k+D 
1
o
1
o'
1
o
1
0
0
1
xe00 +
0
1
1
4
1
1
l*
1
e(k)
with e(k) = y(k)5. The system parameters were estimated using the
algorithm (42)(45). The initial conditions were chosen to be P(0)=I,
x(0)= [1 Of. xc(0)=[0 0 0]^, and R0=I. The initial parameter
estimates, parameter ranges, and steady state estimated values are
displayed in Table 74. Figure 77 includes plots of the system output
response, control input, and estimated system parameters. For
comparison purposes, both the system output and control input are shown
in Figure 78 for the case when the parameters are known exactly.
29
It then follows that
e(m) e(n) = '(m) een) = (P(m) P(n))P*(O)e,(0)
m1 1
P(m) l P'Mi + l)P(i)f(0(i))
1=0
+P(n) Y P1(1 +l)P(1)f(0(D).
1=0
Without loss of generality, assume that m is greater than n. Replacing
the matrix P(n) preceding the summation with P(m) + (P(n) P(m)) gives
i m1 1
0(m) 0(n) = (P(m) P(n))P"1(O)0(O) P(m) Y P1(i + 1)P(i)f(0(i))
i=n
+ (P(n) P(m)) l P1(i + l)P(i)f(0(i)).
1=0
Using the properties of the induced matrix norm defined in Chapter II
gives
0ij(m) 0ij(n) < lo(nn) 0(n) i
< iP(m) P(n) ii nP"1(0)?(0)n
. m1 n
+ nP(m> n l Pi(i+l)P(i) i iif(e(i)) ii
i=n
(421)
nl ,
+ nP(n) P(m) ii l nP (i+l)P(i)n If (0(i)) ii .
i=0
If the right hand side of equation (421) can be made arbitrarily small
for m and n sufficiently large, then each component e.^(k) of the matrix
* J
90
u(k)
Figure 79 continued
Estimation Scheme (42)(45)
64
stabilizing feedback gain L(k) = [L^(k) L?(k)] can be computed from one
step of the Riccati difference equation
L(k) = [GT(k)RkG(k) + I]"1 GT(k)RkF(k) (66)
Rk+1 = Q + LT(k)L(k) + (F(k)4(k)L(k))TRk(F(k)G(k)L(k)).
Again, RQ and Q are arbitrary positive definite symmetric matrices.
In order to use the existing results on tracking and disturbance
rejection, the closedloop system consisting of (64) and the control
law (65) must converge to a timeinvariant system. Properties (v) and
(vii) of Theorem 4.1 imply that the pair (F(k),G(k)) converges to a
stabilizable pair (F ,G ). Using the results of Samson CLO], L(k)
00 00
converges to L the optimal LQ feedback for the pair (F ,G ) given by
00 00
L = [GTR G + I]1 GT R F
00 00 00 CO 00 00 00
where the matrix R is the solution to the ARE for (F ,G ) More
00 00
precisely,
R = Q + LTL + (F G L )TR (F G L ).
Applying the control law (65) to the system described by (64)
results in an exponentially stable system driven by e(k). From the
results in Chapter Five, the observer error e(k) converges to zero.
Also, it follows from property (ii) of Theorem 4.1 that M(k) is
bounded. Thus, the closedloop system consisting of (64) and the
TABLE OF CONTENTS
ACKNOWLEDGEMENTS ii
ABSTRACT iv
CHAPTERS
I INTRODUCTION 1
II BACKGROUND 4
III SYSTEM DEFINITIONS AND ASSUMPTIONS 12
IV PARAMETER ESTIMATION 18
V ADAPTIVE REGULATOR 36
Uniform Stabilizability of the Estimated System 36
Adaptive Observer 41
Feedback Gain Sequence 44
Stability of the Adaptive ClosedLoop System 55
VI APPLICATION TO TRACKING AND DISTURBANCE REJECTION 61
VII SIMULATIONS 66
VIII DISCUSSION 96
REFERENCES 100
BIOGRAPHICAL SKETCH 103
i i i
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
f t'}'c~> * J
S. /Svoronos
Professor of Chemical Engineering
This dissertation was submitted to the Graduate Faculty of
the College of Engineering and to the Graduate School and
was accepted as partial fulfillment of the requirements for
the degree of Doctor of Philosophy.
August 1986
Dean, College of Engineering
Dean, Graduate School
45
over fl. This approach has been developed by a number of individuals,
see for example [48]. A drawback to this approach is that (F(k),G(k))
may not be reachable or stabilizable at certain points in time (i.e.,
0(k) may be a singular point for some values of k). Recall that
Proposition 5.1 and property (v) of Theorem 4.1 only guarantee that
(F(k),G(k)) is stabilizable for all k exceeding some finite N. Thus,
for time points prior to N, the chosen control law may not have a
solution. Therefore, in this approach it is necessary to check whether
or not (F(k),G(k)) is reachable or stabilizable at each time point k,
which adds to the number of online calculations. Also, a decision must
be made on how to modify the control law when singular points are
encountered. A common choice in the literature is to update the control
law only when (F(k),G(k)) is stabilizable.
Another approach first introduced by Kreisselmeier for continuous
time systems is to compute the feedback gain asymptotically [29]. This
approach was developed for SISO discretetime systems by Samson and
Fuchs [1012] using an LQ control strategy, and later extended to MIMO
discetetime systems by Ossman and Kamen [3032].
In [3032], the stabilizing feedback L(k) is computed at each time
point k by solving in real time one step of a Riccati difference
equation. More precisely, let
L(k) = [GT(k)RkG(k) + I]_1GT(k)RkF(k) (58)
where R(k) is the solution to the Riccati difference equation
Rk+1 = Q + LT(k)L(k) + (F(k) G(k)L(k))TRk(F(k) G(k)L(k)).
(59)
57
Ml
A(k+M,k) = n A(k+i)
i=0'
Since A(k) is an exponentially stable timevarying system matrix, it
follows that
nA(k+M,k)I < R1 < for all k,M > 0. (513)
Also, given 0<5<1 there exists a finite p such that
nA(k+p,k)ii < 6 for all k. (514)
For the remainder of the proof, p will be a fixed constant such that
(514) is satisfied. Taking the norm of both sides of (511) and using
(512) gives
nz(k+l)n < A(k) iinz(k) n + nw(k)ii (515)
< ( h A( k) d + i(k)) nz(k) ii + b(k).
Noting that nA(k)n, a(k), and b(k) are all bounded and p is fixed, it
follows from equation (515) that there exist finite constants R2 and Rg
such that
11 z(k+i) 11 < R^t!z(k) 11 + Rg for all ie [0,p] and for all k. (516)
Solving equation (511) recursively gives
37
where
*
A1 1
_A2
32
G =
Vi
I
Bn1
An
0
Bn
H = [I 0 ... 0]
The entries in F and G depend on the unknown system parameter matrix
6. In fact, by the definition of e in (33), F and G can be viewed as
continuous functions of 9 where 9 ranges over the parameter
space n = {9e ]R^xm: 9..e[9?!n 9*il?x]}. This dependance can be made
1J T J T 3
explicit by denoting F and G as F(e) and G(e) respectively. Assumption
A3 in Chapter Three is equivalent to requiring that the system described
by (F(9),G(9)) is pointwise stabilizable over n.
Definition: The pair (F(g),G(9)) is pointwise stabilizable over fi if
for each 9 belonging to n, there exists a feedback matrix L(e) such that
Uma(F(9)G(9)L(9))I < q(9) < 1 where x^ () is the eigenvalue of
maX max
() with the greatest magnitude.
It follows from known results [24], that pointwise stabilizability
of (F(0) ,G(e)) over n is equivalent to the rank condition (34). Thus,
stabilizability over a can be checked using the polynomial matrices
D(z,9) and N(z,0) comprising the system transfer function. Given e > 0,
let be an extension of the parameter space n defined as
Q
e
rn mNxrn rAmin AmaxL i,
{0e F : 9..e[9.. e, 9.. +e]}.
(52)
3
algorithms, such as the recursive leastsquares, projection, and
orthogonalized projection algorithms discussed in [6], which do indeed
have the required properties.
In this dissertation, an indirect adaptive regulator which does not
require a persistently exciting input is derived for multiinput multi
output discretetime systems. In Chapter II, more explicit background
material on parameter estimators and indirect adaptive controllers is
presented. Chapter III contains the system definitions and the
assumptions which include (1) an upper bound on system order is known,
(2) the unknown system parameters belong to known bounded intervals, and
(3) the plant is stabilizable for all possible values of the unknown
parameters ranging over the known intervals. A parameter estimator is
derived in Chapter IV which has the required properties for proving
global stability of indirect adaptive controllers and, in addition,
forces the estimates of the parameters to converge to the known bounded
intervals. In Chapter V, it is first shown that the assumptions listed
in Chapter II ensure the parameter estimator described in Chapter IV
will generate a uniformly stabilizable estimated system. Using this
parameter estimator, an adaptive LQ regulator is developed which, when
applied to the unknown plant, results in a globally stable closedloop
system in the sense that the system inputs and outputs converge to zero
asymptotically. The results are then extended in Chapter VI to the case
of tracking and/or deterministic disturbance rejection using the
internal model principle [20]. Chapter VII includes simulations of the
adaptive LQ controller for both SISO and MIMO discretetime systems. A
discussion of the results and further work to be investigated is
included in Chapter VII.
101
[13] C. Rohrs, "Adaptive Control in the Presence of Unmodeled
Dynamics," Ph.D. dissertation, Dept. Elec. Eng. Comput. Sci.,
Mass. Inst. Techno!., Aug. 1982.
[14] C. Rohrs, L. Valavani, M. Athans, G. Stein, "Robustness of
ContinuousTime Adaptive Control Algorithms in the Presence of
Unmodeled Dynamics," IEEE Trans. Automat. Contr., Vol. AC30, No.
9, pp. 881889, Sept. 1985.
[15] G. Kreisselmeier, "A Robust Indirect AdaptiveControl Approach,"
Int. J. Control, Vol. 43, No. 1, pp. 161175, 1986.
[16] G. Kreisselmeier and B.D.O. Anderson, "Robust Model Reference
Adaptive Control," IEEE Trans Automat. Contr., Vol. AC31, No. 2,
Feb. 1986.
[17] K.S. Narendra and A.M. Annaswamy, "Persistent Excitation and
Robust Adaptive Algorithms", in Proc. 3rd Workshop on Applications
of Adaptive System Theory, Yale University, New Haven, CT, pp. Il
ls, 1983.
[18] B.D.O. Anderson, "Exponential Convergence and Persistent
Excitation," in Proc. 21st IEEE Conf. Decision Contr., Orlando,
FL, pp. 1217, 1982.
[19] P.A. Ioannou and P.V. Kokotovic, "Singular Perturbations and
Robust Redesign of Adaptive Control," in Proc. 21st Conf. Decision
and Contr., Orlando, FL, 1982.
[20] C. Chen, Linear System Theory and Design, CBS College Publishing,
New York, N.Y., 1984.
[21] R. LozanoLeal and G.C. Goodwin, "A Globally Convergent Adaptive
Pole Placement Algorithm Without a Persistency of Excitation
Requirement," IEEE Trans. Automat. Contr., Vol. AC30, No. 8, pp.
795798, Aug. 1985.
[22] P. De Larminat, "On the Stability Condition in Indirect Adaptive
Control," Automtica, Vol. 20, pp. 793795, 1984.
[23] G. Kreisselmeier, "An Approach to Stable Indirect Adaptive
Control," Automtica, Vol. 21, pp. 425431, July 1985.
[24] P.P. Khargonekar and E.D. Sontag, "On the Relation Between Stable
Matrix Fraction Factorizations and Regulable Realizations of
Linear Systems over Rings," IEEE Trans. Automatic Control, Vol.
AC27, pp. 627638, 1982.
[25] E.W. Kamen and P.P. Khargonekar, "On the Control of Linear Systems
Whose Coefficients Are Functions of Parameters," IEEE Trans.
Automat. Contr., Vol. AC29, No. 1, pp. 2533, Jan. 1984.
[26] W. Rudin, Principles of Mathematical Analysis, McGrawHill, New
York, 1976.
95
8
6
4
2
2
4
6
8
10
20
30
40
50
60
b
a
12
12
(k)
XE)
70
(e)
8
6
4
2
2
4
6
8
UmXf.
10
20
30 40 50
(f)
60
b21(k)
70 b22(k)
Figure 710 continued
Estimation Scheme (42)(45)
85
Figure 77 continued
Estimation Scheme (42)(45)
41
Adaptive Observer
The adaptive observer for the estimated system is given by
x(k+l) = F(k)x(K) + G(k)u(k) + M(k)(y(k)y(k)) (53)
y(k) = H>t (k)
where
*A1(k) I
A2(k)
\w
B2(k)
F(k) =
I
_An(k) 0
G(k) =
Bn(k)
M(k)
Ax(k)
A2(k)
H = [I 0 .... 03.
The adaptive observer defined by (53) is a deadbeat observer. The
output y(k) of the adaptive observer can be expressed in the following
form:
y(k) = CAjU1)
An(kn) B1(k1)
Bn(kn)](j>(kl)
(54)
30
sequence e(k) will be a Cauchy sequence. It will then follow using
property (v) that e^. (k) converges to a point in the interval
[e?1n>e?!x]. In order to prove that the right hand side of the in
T J 1J
equality (421) can be made arbitrarily small, two things must be
1
shown: nP (k+l)P(k)ii < v < for all k, and Â£ nf(e(k))ii < L < .
1 k=1
In order to prove nP (k+l)P(k)n < v < for all k, first assume that
the determinant of P(k) does not converge to zero. In this case,
P_1(k+1) converges to P1 which implies that P^(k+l)P(k) converges to
oo
I. Therefore, i P"*(K+l)P(k) n is bounded for all k when the deter
minant of P(k) does not converge to zero. Now suppose the determinant
of P(k) does converge to zero. Applying the matrix inversion lemma to
equation (43) gives
P1(k+l)P(k) =
I + *a(k)^(k)P(k)
2
Using the properties of the induced matrix norm, it follows that
i a(k) J(k)n
iPA(k+l)P(k) i < iiln + 2 J iiP(k) I, (422)
nk
From (45), there exists a finite M such that nk = max (1,na(k) ii) for
k > M. It then follows that all the components of the matrix
+a(k)*a(k)
2
nk
have magnitude less than or equal to one for k > M which in
turn implies
^aOO^k)
2
nk
Hill = 1 and nP(k)n < 4,
is bounded for all k.
equation (422) implies
Noting also that
HP'^k+ljPk) ii is
8
where
M(k)
a^k)
, a^(k) = estimate of a^ at time k.
an(k)
The feedback gain sequence is computed at each time point k by
solving one step of a Riccati difference equation:
R(k+1) = Q + LT(k)L(k) + (F(k) G(k)L(k))TR(k)(F(k) G(k)L(k))
L(k) = [GT(k)R(k)G(k) + I]1 GT(k)R(k)F(k) (27)
In (27), Q and the initial value R(0) are arbitrary positive
definite symmetric matrices.
In order to prove global stability of the closedloop system,
Samson and Fuchs make two assumptions in addition to the stabilizability
assumption on (F,G) mentioned previously. First, the parameter
estimator must possess the following three properties:
i. ne(k)n < M < for all k
ii. ne(k) e(km) n + 0 as k > for any finite m
iii.e(k) < a(k) n<>(kl) n + $(k)
where a(k) and g(k) converge to zero.
92
Table 76
DATA
"a12
b12
b21
b22
Actual Parameters
0.75
2.0
3.0
0.5
Initial Estimates
0.65
1.0
2.5
1.0
Parameter Ranges
I1
00
o
A
in
o
i i
[5.0, 3.0]
[1.0, 4.0]
[1.0, 3.0)
Estimation Scheme
(42)(45)
0.785
1.58
2.97
0.320
BIOGRAPHICAL SKETCH
The author was born in Cincinnati, Ohio, on July 19, 1959. She
received a BSEE and a MSEE degree from the Georgia Institute of
Technology in March 1982 and December 1982 respectively. She expects to
receive a Ph.D. in electrical engineering from the University of Florida
in August 1986.
103
50
In other words, Pk is equal to Rk for all k up to time t. After time
t, pÂ£ is generated from the timeinvariant Riccati difference equation
initialized by Rt and using the constant matrices F(*) and G()
evaluated at 9(t). For every e(t) belonging to n the sequence pÂ£
will be bounded for all k. In fact, P^ converges to P^ whenever
K oo
9(t)eiU. It will first be shown that nRk+N P^r < e/6 for k
sufficiently large. The matrix p^ is simply the Nth step of a time
invariant ROE initialized at time k by Rk and using constant matrices
F() and G() evaluated at 9(k). The interpretation for
llRk+N Pk+N11 bei'ng sn,a",1 For k sufficiently large is that the time
varying RDE given by (59) will not vary much from the timeinvariant
ROE over a finite interval of time N as long as 9(k) does not change
much over the time interval N. Equation (59) can be rewritten as
Rk+i+i 55 Q + FT(0(k))Rk+iF(e(k)) FT(e(k))Rk+iG(e(k))
X [I + GT(9(k))Rk+.G(9(k))]1 GT(9(k))Rk+.F(9(k)) + Zk+.; RR
for all k and for all ie[0,N].
The matrix zk+i is given by
Ek+i = FT(0(k+i))Rk+iF(9(k+i)) PT(e(k+i))Rk+j G(9(k+i))
x [I + GT(9(k+i))Rk+.G(9(k+i))]"1 GT(9(k+i))Rk+.F(9(k+i))
FT(9(k))Rk+.F(9(k)) + FT(9(k))Rk+.G(9(k))
x [I + GT(9(k))Rk+.G(9(k))]'1 GT(9(k))Rk+.F(9(k)) for all k and
for all ie[0,N].
PAGE 1
1(: 5(68/76 ,1 ,1',5(&7 $'$37,9( &21752/ %\ .$7+/((1 $ . 2660$1 $ ',66(57$7,21 35(6(17(' 72 7+( *5$'8$7( 6&+22/ 2) 7+( 81,9(56,7< 2) )/25,'$ ,1 3$57,$/ )8/),//0(17 2) 7+( 5(48,5(0(176 )25 7+( '(*5(( 2) '2&725 2) 3+,/2623+< 81,9(56,7< 2) )/25,'$
PAGE 2
$&.12:/('*(0(176 , ZRXOG OLNH WR DFNQRZOHGJH WKH IROORZLQJ SHRSOH IRU WKHLU FRQWULEXWLRQV WR WKH FRPSOHWLRQ RI WKLV GLVVHUWDWLRQ 6SHFLDO WKDQNV JR WR P\ DGYLVRU 'U (GZDUG : .DPHQ IRU KLV QXPHURXV KHOSIXO VXJJHVWLRQV DQG FUHDWLYH LGHDV 7KDQNV DOVR JR WR WKH PHPEHUV RI P\ VXSHUYLVRU\ FRPPLWWHH 'U 7 ( %XOORFN 'U ' : +HDUQ 'U 3 = 3HHEOHV DQG 'U 6 6YRURQRV 7KLV ZRUN ZDV VXSSRUWHG LQ SDUW E\ WKH 8 6 $UP\ 5HVHDUFK 2IILFH 5HVHDUFK 7ULDQJOH 3DUN 1& XQGHU &RQWUDFW 1R '$$*. LL
PAGE 3
7$%/( 2) &217(176 $&.12:/('*(0(176 LL $%675$&7 LY &+$37(56 , ,1752'8&7,21 ,, %$&.*5281' ,,, 6<67(0 '(),1,7,216 $1' $668037,216 ,9 3$5$0(7(5 (67,0$7,21 9 $'$37,9( 5(*8/$725 8QLIRUP 6WDELOL]DELOLW\ RI WKH (VWLPDWHG 6\VWHP $GDSWLYH 2EVHUYHU )HHGEDFN *DLQ 6HTXHQFH 6WDELOLW\ RI WKH $GDSWLYH &ORVHG/RRS 6\VWHP 9, $33/,&$7,21 72 75$&.,1* $1' ',6785%$1&( 5((&7,21 9,, 6,08/$7,216 9,,, ',6&866,21 5()(5(1&(6 %,2*5$3+,&$/ 6.(7&+ L L L
PAGE 4
$EVWUDFW RI 'LVVHUWDWLRQ 3UHVHQWHG WR WKH *UDGXDWH 6FKRRO RI 7KH 8QLYHUVLW\ RI )ORULGD LQ 3DUWLDO )XOILOOPHQW RI WKH 5HTXLUHPHQWV IRU WKH 'HJUHH RI 'RFWRU RI 3KLORVRSK\ 1(: 5(68/76 ,1 ,1',5(&7 $'$37,9( &21752/ %\ .$7+/((1 $ . 2660$1 $XJXVW &KDLUPDQ 'U (GZDUG : .DPHQ 0DMRU 'HSDUWPHQW (OHFWULFDO (QJLQHHULQJ $Q DGDSWLYH UHJXODWRU ZKLFK GRHV QRW UHTXLUH D SHUVLVWHQWO\ H[FLWLQJ LQSXW LV GHULYHG IRU PXOWLLQSXW PXOWLRXWSXW OLQHDU GLVFUHWHn WLPH V\VWHPV 7KH DVVXPSWLRQV PDGH RQ WKH XQNQRZQ SODQW DUH f DQ XSSHU ERXQG RQ WKH V\VWHP RUGHU LV NQRZQ f WKH V\VWHP SDUDPHWHUV EHORQJ WR NQRZQ ERXQGHG LQWHUYDOV DQG f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
PAGE 5
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f ZKHUH WKH XQNQRZQ SODQW LV IRUFHG WR EHKDYH DV\PSWRWLFDOO\ OLNH VRPH SUHFKRVHQ UHIHUHQFH PRGHO 6HYHUDO LQGLYLGXDOV KDYH GHYHORSHG WKLV DSSURDFK VHH IRU H[DPSOH >@ 7KH DVVXPSWLRQV QHHGHG WR SURYH JOREDO VWDELOLW\ RI PRGHO UHIHUHQFH DGDSWLYH FRQWUROOHUV DUH f WKH XQNQRZQ SODQW LV PLQLPXP SKDVH f WKH UHODWLYH GHJUHH RI WKH SODQW LV NQRZQ DQG f WKH VLJQ RI WKH SODQW JDLQ LV NQRZQ ,QGLUHFW DGDSWLYH FRQWUROOHUV DUH DSSOLFDEOH WR QRQPLQLPXP SKDVH DV ZHOO DV PLQLPXP SKDVH V\VWHPV ,Q LQGLUHFW DGDSWLYH FRQWURO WKH XQNQRZQ V\VWHP SDUDPHWHUV DUH HVWLPDWHG LQ UHDO WLPH XVLQJ DYDLODEOH LQSXWRXWSXW LQIRUPDWLRQ WKHQ WKH FRQWURO ODZ LV FRPSXWHG XVLQJ WKH PRVW UHFHQW SDUDPHWHU HVWLPDWHV $ QXPEHU RI LQGLYLGXDOV KDYH LQWURGXFHG LQGLUHFW DGDSWLYH FRQWUROOHUV VHH IRU H[DPSOH >@ 7KH DVVXPSWLRQV
PAGE 6
n XVHG WR SURYH JOREDO VWDELOLW\ RI LQGLUHFW DGDSWLYH FRQWUROOHUV DUH f DQ XSSHU ERXQG RQ V\VWHP RUGHU LV NQRZQ f WKH HVWLPDWHG V\VWHP LV XQLIRUPO\ VWDELOL]DEOH DQG f WKH SDUDPHWHU HVWLPDWRU SRVVHVVHV FHUWDLQ NH\ SURSHUWLHV 6LQFH WKH DGDSWLYH FRQWUROOHU GHYHORSHG LQ WKLV GLVVHUWDWLRQ FDQ EH FODVVLILHG DV LQGLUHFW HDFK RI WKHVH DVVXPSWLRQV ZLOO EH GLVFXVVHG LQ PRUH GHWDLO 7KH LPSRUWDQFH RI WKH ILUVW DVVXPSWLRQ ZDV LOOXVWUDWHG LQ >@ E\ 5RKUV DQG DVVRFLDWHV ZKR LQYHVWLJDWHG WKH EHKDYLRU RI DGDSWLYH FRQWUROOHUV LQ WKH SUHVHQFH RI XQPRGHOHG G\QDPLFV 5RKUV GHPRQVWUDWHG WKURXJK FRPSXWHU VLPXODWLRQV WKDWfÂ¬ DOO H[LVWLQJ DGDSWLYH FRQWUROOHUV FRXOG JR XQVWDEOH LI WKH RUGHU RI WKH V\VWHP ZDV XQGHUHVWLPDWHG ,Q UHVSRQVH WR 5RKUVn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
PAGE 7
DOJRULWKPV VXFK DV WKH UHFXUVLYH OHDVWVTXDUHV SURMHFWLRQ DQG RUWKRJRQDOL]HG SURMHFWLRQ DOJRULWKPV GLVFXVVHG LQ >@ ZKLFK GR LQGHHG KDYH WKH UHTXLUHG SURSHUWLHV ,Q WKLV GLVVHUWDWLRQ DQ LQGLUHFW DGDSWLYH UHJXODWRU ZKLFK GRHV QRW UHTXLUH D SHUVLVWHQWO\ H[FLWLQJ LQSXW LV GHULYHG IRU PXOWLLQSXW PXOWL RXWSXW GLVFUHWHWLPH V\VWHPV ,Q &KDSWHU ,, PRUH H[SOLFLW EDFNJURXQG PDWHULDO RQ SDUDPHWHU HVWLPDWRUV DQG LQGLUHFW DGDSWLYH FRQWUROOHUV LV SUHVHQWHG fÂµ &KDSWHU ,,, FRQWDLQV WKH V\VWHP GHILQLWLRQV DQG WKH DVVXPSWLRQV ZKLFK LQFOXGH f DQ XSSHU ERXQG RQ V\VWHP RUGHU LV NQRZQ f WKH XQNQRZQ V\VWHP SDUDPHWHUV EHORQJ WR NQRZQ ERXQGHG LQWHUYDOV DQG f WKH SODQW LV VWDELOL]DEOH IRU DOO SRVVLEOH YDOXHV RI WKH XQNQRZQ SDUDPHWHUV UDQJLQJ RYHU WKH NQRZQ LQWHUYDOV $ SDUDPHWHU HVWLPDWRU LV GHULYHG LQ &KDSWHU ,9 ZKLFK KDV WKH UHTXLUHG SURSHUWLHV IRU SURYLQJ JOREDO VWDELOLW\ RI LQGLUHFW DGDSWLYH FRQWUROOHUV DQG LQ DGGLWLRQ IRUFHV WKH HVWLPDWHV RI WKH SDUDPHWHUV WR FRQYHUJH WR WKH NQRZQ ERXQGHG LQWHUYDOV ,Q &KDSWHU 9 LW LV ILUVW VKRZQ WKDW WKH DVVXPSWLRQV OLVWHG LQ &KDSWHU ,, HQVXUH WKH SDUDPHWHU HVWLPDWRU GHVFULEHG LQ &KDSWHU ,9 ZLOO JHQHUDWH D XQLIRUPO\ VWDELOL]DEOH HVWLPDWHG V\VWHP 8VLQJ WKLV SDUDPHWHU HVWLPDWRU DQ DGDSWLYH /4 UHJXODWRU LV GHYHORSHG ZKLFK ZKHQ DSSOLHG WR WKH XQNQRZQ SODQW UHVXOWV LQ D JOREDOO\ VWDEOH FORVHGORRS V\VWHP LQ WKH VHQVH WKDW WKH V\VWHP LQSXWV DQG RXWSXWV FRQYHUJH WR ]HUR DV\PSWRWLFDOO\ 7KH UHVXOWV DUH WKHQ H[WHQGHG LQ &KDSWHU 9, WR WKH FDVH RI WUDFNLQJ DQGRU GHWHUPLQLVWLF GLVWXUEDQFH UHMHFWLRQ XVLQJ WKH LQWHUQDO PRGHO SULQFLSOH >@ &KDSWHU 9,, LQFOXGHV VLPXODWLRQV RI WKH DGDSWLYH /4 FRQWUROOHU IRU ERWK 6,62 DQG 0,02 GLVFUHWHWLPH V\VWHPV $ GLVFXVVLRQ RI WKH UHVXOWV DQG IXUWKHU ZRUN WR EH LQYHVWLJDWHG LV LQFOXGHG LQ &KDSWHU 9,,
PAGE 8
&+$37(5 ,, %$&.*5281' ,Q LQGLUHFW DGDSWLYH FRQWURO WKH V\VWHP SDUDPHWHUV DUH HVWLPDWHG LQ UHDO WLPH XVLQJ DYDLODEOH LQSXWRXWSXW LQIRUPDWLRQ 7KH FRQWURO ODZ LV WKHQ FRPSXWHG IRU WKH HVWLPDWHG V\VWHP XVLQJ WKH PRVW UHFHQW SDUDPn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n SXW GLIIHUHQFH HTXDWLRQ T 3 f ,Q f \Nf LV WKH V\VWHP RXWSXW DQG XNf LV WKH FRQWURO LQSXW ,W LV DVVXPHG WKDW DQ XSSHU ERXQG Q RQ T DQG S LV NQRZQ EXW VRPH RU DOO RI WKH V\VWHP SDUDPHWHUV DM DQG EM DUH XQNQRZQ 7KH V\VWHP GHVFULEHG E\ f FDQ EH UHZULWWHQ LQ WKH IROORZLQJ IRUP ZKLFK LV FRQYHQLHQW IRU SDUDPHWHU HVWLPDWLRQ \Nf H7M!NOf
PAGE 9
ZKHUH f W!7NOf >\NOf \NQf XNOf XNQf@ 7KH YHFWRU FRQVLVWV RI DOO WKH V\VWHP SDUDPHWHUV DQG M!NOf LV D UHJUHVVLRQ YHFWRU RI SDVW LQSXWV DQG RXWSXWV 7KURXJKRXW WKH IROORZLQJ GLVFXVVLRQ Q fÂµ ZLOO GHVLJQDWH WKH YHFWRU (XFOLGHDQ QRUP GHILQHG E\ 7 A [ [ [f 2I DOO WKH SDUDPHWHU HVWLPDWLRQ VFKHPHV GLVFXVVHG LQ >@ WKH OHDVWVTXDUHV DOJRULWKP KDV WKH IDVWHVW FRQYHUJHQFH UDWH DQG LV PRVW HDVLO\ PRGLILHG WR KDQGOH RXWSXW GLVWXUEDQFHV DQG VORZO\ WLPHYDU\LQJ V\VWHPV 7KH DOJRULWKP UHVXOWV IURP PLQLPL]LQJ WKH TXDGUDWLF FRVW IXQFWLRQ N O 7KH FRVW IXQFWLRQ FRQVLVWV RI WKH VXP RI WKH VTXDUHV RI WKH SUHGLFWLRQ HUURUV SOXV DQ DGGLWLRQDO WHUP ZKLFK WDNHV LQLWLDO FRQGLWLRQV LQWR DFFRXQW 7KH PDWUL[ 3f FDQ EH LQWHUSUHWHG DV D PHDVXUH RI FRQILGHQFH LQ WKH LQLWLDO SDUDPHWHU HVWLPDWH f 'HVLJQDWLQJ Nf DV WKH HVWLn PDWH RI WKH V\VWHP SDUDPHWHUV DW WLPH N WKH OHDVWVTXDUHV DOJRULWKP LV GHVFULEHG E\ &\Nf H7NOfW!NOf@ 3Nf 3NLf 3NOf!NOfNOf3NOO S7NOf3NOf !NOf fÂµ 3f 37f ! f
PAGE 10
7KH NH\ SURSHUWLHV RI WKH OHDVWVTXDUHV VFKHPH XVHG IRU SURYLQJ JOREDO VWDELOLW\ RI DQ LQGLUHFW DGDSWLYH FRQWUROOHU DUH LQHNfQ0m IRU DOO N LLLLNMHNOf K fÂµ! DV N p LLLOLP c Nm ! NOf3NOf!Nf ZKHUH HNf \Nf HANOfW!NOf 7KH HUURU HNf LV UHIHUUHG WR DV WKH SUHGLFWLRQ HUURU VLQFH LW LV VLPSO\ WKH GLIIHUHQFH EHWZHHQ WKH DFWXDO RXWSXW DW WLPH N DQG WKH SUHGLFWHG RXWSXW DW WLPH N XVLQJ WKH PRVW UHFHQW SDUDPHWHU HVWLPDWH 7KHVH SURn SHUWLHV GR QRW GHSHQG RQ WKH W\SH RI FRQWURO LQSXW FKRVHQ RU RQ WKH ERXQGHGQHVV RI WKH V\VWHP LQSXW DQG RXWSXW 7KH OHDVWVTXDUHV DOJRULWKP DOVR KDV WKH SURSHUW\ WKDW WKH SDUDPHWHU HVWLPDWHV FRQYHUJH DOWKRXJK QRW QHFHVVDULO\ WR WKH WUXH YDOXHV 7KLV SURSHUW\ LV QRW LQFOXGHG DPRQJ WKH NH\ SURSHUWLHV EHFDXVH LW LV QRW QHFHVVDU\ IRU VWDELOLW\ SURRIV $Q H[DPSOH RI LQGLUHFW DGDSWLYH FRQWURO LQ WKH 6,62 FDVH LV 6DPVRQ DQG )XFKVn /4 FRQWUROOHU GLVFXVVHG LQ >@ DQG EULHIO\ RXWOLQHG KHUH $ VWDWHVSDFH REVHUYHU UHDOL]DWLRQ RI WKH V\VWHP GHVFULEHG E\ f LV JLYHQ E\ [NOf )[Nf *XNf f \Nf +[Nf
PAGE 11
ZKHUH ) DO EO D E fÂµ fÂµ * fÂµ fÂµ DQO fÂµ fÂµ EQO DQ EQ + > @ 6DPVRQ DQG )XFKV DVVXPH WKDW )*f LV VWDELOL]DEOH 7KLV DVVXPSWLRQ DOORZV IRU QRQPLQLPXP SKDVH V\VWHPV DQG V\VWHPV ZLWK VWDEOH FRPPRQ SROHV DQG ]HURV 7KH DGDSWLYH ,4 FRQWURO ODZ XVHG E\ 6DPVRQ DQG )XFKV >@ LV JLYHQ E\ XNf /Nf[Nf f ZKHUH XNf LV WKH LQSXW WR WKH JLYHQ SODQW [Nf LV WKH VWDWH HVWLPDWH DQG /Nf LV D VWDELOL]LQJ IHHGEDFN JDLQ IRU WKH HVWLPDWHG V\VWHP )Nf ILNff 7KH VWDWH HVWLPDWH [Nf LV JHQHUDWHG IURP DQ DGDSWLYH REVHUYHU /HWWLQJ )Nf DQG *Nf UHSUHVHQW WKH FXUUHQW HVWLPDWHV IRU PDWULFHV ) DQG * WKH DGDSWLYH REVHUYHU IRU WKH V\VWHP LV JLYHQ E\ [NOf )NfNf *NfXNf 0Nf \Nf\Nff f \Nf +"Nf
PAGE 12
ZKHUH 0Nf DANf DANf HVWLPDWH RI DA DW WLPH N DQNf 7KH IHHGEDFN JDLQ VHTXHQFH LV FRPSXWHG DW HDFK WLPH SRLQW N E\ VROYLQJ RQH VWHS RI D 5LFFDWL GLIIHUHQFH HTXDWLRQ 5Nf 4 /7Nf/Nf )Nf *Nf/Nff75Nf)Nf *Nf/Nff /Nf >*7Nf5Nf*Nf ,@fÂ¬ *7Nf5Nf)Nf f ,Q f 4 DQG WKH LQLWLDO YDOXH 5f DUH DUELWUDU\ SRVLWLYH GHILQLWH V\PPHWULF PDWULFHV ,Q RUGHU WR SURYH JOREDO VWDELOLW\ RI WKH FORVHGORRS V\VWHP 6DPVRQ DQG )XFKV PDNH WZR DVVXPSWLRQV LQ DGGLWLRQ WR WKH VWDELOL]DELOLW\ DVVXPSWLRQ RQ )*f PHQWLRQHG SUHYLRXVO\ )LUVW WKH SDUDPHWHU HVWLPDWRU PXVW SRVVHVV WKH IROORZLQJ WKUHH SURSHUWLHV L QHNfQ 0 p IRU DOO N LL QHNf HNPf Q DV N Â‘! m IRU DQ\ ILQLWH P LLL_HNf_ DNf Q_!NOf Q Nf ZKHUH DNf DQG JNf FRQYHUJH WR ]HUR
PAGE 13
$Q H[DPSOH RI D SDUDPHWHU HVWLPDWRU SRVVHVVLQJ WKHVH SURSHUWLHV LV WKH OHDVWVTXDUHV HVWLPDWRU GLVFXVVHG SUHYLRXVO\ &OHDUO\ SURSHUWLHV Lf DQG LLf OLVWHG DERYH DUH HTXLYDOHQW WR SURSHUWLHV Lf DQG LLf JLYHQ IRU WKH OHDVWVTXDUHV DOJRULWKP ,W LV QRW TXLWH DV REYLRXV WKDW 6DPVRQ DQG )XFKVn SURSHUW\ LLLf LV HTXLYDOHQW WR SURSHUW\ LLLf RI WKH OHDVW VTXDUHV DOJRULWKP 7KH SURRI LV JLYHQ LQ &KDSWHU ,9 7KH VHFRQG DVVXPSWLRQ IRU SURYLQJ JOREDO VWDELOLW\ LV WKDW WKH HVWLPDWHG V\VWHP )Nf*Nff PXVW EH XQLIRUPO\ VWDELOL]DEOH 'HILQLWLRQ 7KH V\VWHP )Nf*Nff LV XQLIRUPO\ VWDELOL]DEOH LI WKHUH H[LVWV DQ LQWHJHU U ! D FRQVWDQW T DQG D XQLIRUPO\ ERXQGHG VHTXHQFH /Nf VXFK WKDW WU O + )Nf*Nf/Nff, TO IRU DOO WH r N W ZKHUH WU Q )Nf*Nf/Nff )WUf *WUf/WUff[ N W )WUf *WUOf/WUOff )Wf *Wf/Wff 7KHRUHP $VVXPLQJ WKH V\VWHP )*f LV VWDELOL]DEOH WKH SDUDPHWHU HVWLPDWRU SRVVHVVHV WKH WKUHH UHTXLUHG SURSHUWLHV DQG WKH HVWLPDWHG V\VWHP )Nf*Nff LV XQLIRUPO\ VWDELOL]DEOH WKH FRQWURO ODZ XNf /Nf[Nf GHVFULEHG E\ ff ZLOO UHVXOW LQ D JOREDOO\ VWDEOH FORVHGORRS V\VWHP LQ WKH VHQVH WKDW WKH V\VWHP LQSXW DQG RXWSXW FRQYHUJH WR ]HUR IRU DQ\ LQLWLDO VWDWHV LQ WKH SODQW DQG REVHUYHU 6DPVRQ DQG )XFKV >@f
PAGE 14
,Q RUGHU WR SURYH JOREDO VWDELOLW\ 6DPVRQ DQG )XFKV PXVW DVVXPH WKH HVWLPDWHG V\VWHP )Nf*Nff LV XQLIRUPO\ VWDELOL]DEOH 8QIRUWXn QDWHO\ WKLV VWDELOL]DELOLW\ FRQGLWLRQ FDQQRW EH FKHFNHG D SULRUL ZKLFK PHDQV JOREDO VWDELOLW\ RI WKH DGDSWLYH UHJXODWRU FDQQRW EH JXDUDQWHHG ,I WKH OHDVWVTXDUHV HVWLPDWRU LV XVHG WKH DVVXPSWLRQ WKDW )Nf*Nff LV XQLIRUPO\ VWDELOL]DEOH LV HTXLYDOHQW WR UHTXLULQJ WKDW LV VWDELOL]DEOH ZKHUH ) OLP )Nf DQG * OLP *Nf RR .0} .+} $V SUHYLRXVO\ PHQWLRQHG WKH SUREOHP RI DYRLGLQJ VLQJXODU OLPLW SRLQWV LV FRPPRQ WR DOO LQGLUHFW DGDSWLYH FRQWUROOHUV 8QWLO WKH UHFHQW ZRUN RI /R]DQR/HDO DQG *RRGZLQ >@ 2H /DUPLQDW >@ DQG .UHLVVHOPHLHU >@ WKH RQO\ ZD\ WR DYRLG D VLQJXODU OLPLW SRLQW ZDV WR UHTXLUH D SHUVLVWHQWO\ H[FLWLQJ LQSXW ZKLFK ZRXOG IRUFH WKH SDUDPHWHU HVWLPDWHV WR FRQYHUJH WR WKH WUXH V\VWHP SDUDPHWHUV 7KLV DSSURDFK KDV EHHQ GHYHOn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n VDU\f WKH SDUDPHWHU HVWLPDWHV JHQHUDWHG IURP WKH OHDVWVTXDUHV HVWLPDn WLRQ DOJRULWKP ZLWK GDWD QRUPDOL]DWLRQ 7KLV PRGLILFDWLRQ DOORZV WKHP
PAGE 15
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n JHQFH QHDU D VLQJXODU SRLQW 'H /DUPLQDWn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n LW\ 7KLV FRQGLWLRQ FDQ EH FKHFNHG D SULRUL VLQFH LW GRHV QRW LQYROYH WKH SDUDPHWHU HVWLPDWHV :LWK WKHVH DVVXPSWLRQV .UHLVVHOPHLHU SURYHV JOREDO VWDELOLW\ RI WKH 6,62 FRQWLQXRXVWLPH DGDSWLYH FRQWUROOHU
PAGE 16
&+$37(5 ,,, 6<67(0 '(),1,7,216 $1' $668037,216 7KH V\VWHP WR EH UHJXODWHG LV WKH PXOWLLQSXW PXOWLRXWSXW OLQHDU GLVFUHWHWLPH V\VWHP GHVFULEHG E\ \22 r " $\NMf " %XNMf f ,Q f \Nf LV WKH P[O RXWSXW YHFWRU DQG XNf LV WKH U[O FRQWURO LQSXW YHFWRU ,W LV DVVXPHG WKDW DQ XSSHU ERXQG Q RQ S DQG T LV NQRZQ EXW DOO RU VRPH RI WKH HQWULHV LQ WKH PDWULFHV $M DQG %M DUH XQNQRZQ 7KH V\VWHP GHVFULEHG E\ f FDQ EH UHZULWWHQ LQ WKH IROORZLQJ IRUP FRQYHQLHQW IRU SDUDPHWHU HVWLPDWLRQ \Nf 37NOf f ZKHUH 37 &$@B $Q %Q@ !7Nf >\7NOf \7NQf X7NOf X7NQf@ 7KH QPUf[P PDWUL[ 3 FRQVLVWV RI DOO WKH V\VWHP SDUDPHWHUV DQG WKH QPUf[O YHFWRU M!NOf LV D UHJUHVVLRQ YHFWRU RI SDVW LQSXWV DQG RXWSXWV ,Q FHUWDLQ DSSOLFDWLRQV VRPH RI WKH HQWULHV LQ WKH V\VWHP PDWUL[ 3 ZLOO EH NQRZQ D SULRUL 6LQFH LW LV QRW QHFHVVDU\ WR HVWLPDWH NQRZQ SDUDPHWHUV D VFKHPH IRU VHSDUDWLQJ WKH NQRZQ SDUDPHWHUV IURP WKH XQNQRZQ SDUDPHWHUV LV DGYDQWDJHRXV 7KLV LV DFFRPSOLVKHG E\ UHZULWLQJ f LQ WKH IROORZLQJ IRUP
PAGE 17
\Nf r 7Â¯!DNOf ANOf f 7KH 1[P PDWUL[ H FRQWDLQV DOO RI WKH XQNQRZQ HQWULHV LQ 3 ZKLOH WKH PDWUL[ WS FRQWDLQV RQO\ WKRVH HQWULHV RI 3 ZKLFK DUH NQRZQ D SULRUL 7KH YHFWRUV Âµ Nf DQG ! Nf DUH UHJUHVVLRQ YHFWRUV ZKRVH FRPSRQHQWV FRPH IURP M!NOf &OHDUO\ WKH GHFRPSRVLWLRQ GHVFULEHG LQ f LV QRW XQLTXH IRU PXOWLLQSXW PXOWLRXWSXW V\VWHPV ,Q RUGHU WR PLQLPL]H WKH UHTXLUHG FRPSXWDWLRQV IRU SDUDPHWHU HVWLPDWLRQ WKH VL]H RI I! Nf 6W VKRXOG EH PDGH DV VPDOO DV SRVVLEOH $Q H[DPSOH ZLOO VHUYH WR LOOXVWUDWH WKHVH FRQFHSWV ([DPSOH &RQVLGHU WKH OLQHDU GLVFUHWHWLPH V\VWHP ZLWK WZR RXWSXWV DQG D VLQJOH LQSXW GHVFULEHG E\ \Nf \LNf D n\AN' \"Nf \NLf XNOf 7KH SDUDPHWHU D LV DVVXPHG WR EH XQNQRZQ 7KH V\VWHP FDQ EH UHZULWWHQ LQ WKH IRUP \Nf 37 ! Nf ZKHUH n D n Â‘\ANLfn !NOf \N' XNOfB $ GHFRPSRVLWLRQ IRU WKLV V\VWHP ZKLFK PLQLPL]HV WKH UHTXLUHG FRPSXWDWLRQV IRU SDUDPHWHU HVWLPDWLRQ LV JLYHQ E\
PAGE 18
\Nf \NLf H7 DLFLf \NLf NOf I!ENOf I!Nf $JDLQ FRQVLGHU WKH OLQHDU GLVFUHWHWLPH V\VWHP GHVFULEHG E\ f ,W LV DVVXPHG WKDW HDFK FRPSRQHQW HA RI WKH XQNQRZQ V\VWHP SDUDPHWHU PDWUL[ EHORQJV WR D NQRZQ ERXQGHG LQWHUYDO >LrQ D""r@ 7KLV DVVXPSWLRQ LV UHDVRQDEOH LQ WKRVH DSSOLFDWLRQV ZKHUH VRPH D SULRUL LQIRUPDWLRQ LV DYDLODEOH RQ WKH V\VWHP &OHDUO\ WKH XQNQRZQ V\VWHP SDUDPHWHU PDWUL[ EHORQJV WR D VXEVSDFH RI @5A[P ZKLFK LV GHILQHG E\ WKH NQRZQ ERXQGHG LQWHUYDOV ,Q RUGHU WR XVH WKH QHFHVVDU\ FRQFHSWV RI ERXQGHGQHVV FRPSDFWQHVV DQG FRQYHUJHQFH D VXLWDEOH QRUP PXVW EH GHILQHG RQ M51[P 7KH YHFWRU QRUP ZKLFK ZLOO EH XVHG WKURXJKRXW WKLV GLVVHUWDWLRQ LV WKH (XFOLGHDQ QRUP GHILQHG E\ ,; O [ [f 9 7KH PDWUL[ QRUP ZKLFK LV LQGXFHG E\ WKH YHFWRU (XFOLGHDQ QRUP LV JLYHQ E\ 7 A r &:0 0f@ fÂµ ZKHUH APD[AA LV WKH HLHQYDOXH RI 070 ZLWK WKH JUHDWHVW PDJQLWXGH %\ WKH DERYH DVVXPSWLRQ RQ WKH HQWULHV RI WKH XQNQRZQ V\VWHP SDUDPHWHU PDWUL[ EHORQJV WR D NQRZQ FRPSDFW VXEVSDFH Q RI @51[P JLYHQ E\
PAGE 19
IW r ^V ,5 1[P 9 U LPQ LQD[ / M` 7KH FRQFHSW RI FRPSDFWQHVV LH FORVHG DQG ERXQGHGf LV UHODWLYH WR WKH PDWUL[ QRUP SUHYLRXVO\ GHILQHG 7KH RQO\ DGGLWLRQDO DVVXPSWLRQ PDGH RQ WKH V\VWHP LV D SRLQWZLVH VWDELOL]DELOLW\ DVVXPSWLRQ RYHU IW 6SHFLILFDOO\ WKH V\VWHP GHVFULEHG E\ f LV DVVXPHG WR EH VWDELOL]DEOH IRU HDFK HIW 7KLV DVVXPSWLRQ FDQ EH YHULILHG D SULRUL EHFDXVH LW GHSHQGV RQO\ RQ WKH NQRZQ VHW IW QRW RQ WKH SDUDPHWHU HVWLPDWHV 7KH VWDELOL]DELOLW\ FRQGLWLRQ FDQ EH WHVWHG XVLQJ WKH WUDQVIHU IXQFWLRQ PDWUL[ RI WKH V\VWHP /HW ']rf DQG 1]rf GHQRWH WKH SRO\QRPLDO PDWULFHV GHILQHG E\ 28fÂ°f $ ]af 18f O %]fÂ¯Mf M O M O 7KH FRHIILFLHQWV RI ']Of DQG 1]Af GHSHQG RQ WKH V\VWHP SDUDPHWHUV DQG WKHUHIRUH FDQ EH YLHZHG DV IXQFWLRQV RI 7KLV GHSHQGDQF\ LV PDGH H[SOLFLW E\ GHQRWLQJ 'L]fÂ¬f DQG 1]f DV ']f DQG 1]fÂ°OHf UHVSHFWLYHO\ 7KH WUDQVIHU IXQFWLRQ PDWUL[ RI WKH V\VWHP GHVFULEHG E\ f LV WKHQ JLYHQ E\ :]f 'fÂ¯]f1]fÂ¯f 7KH VWDELOLW\ DVVXPSWLRQ LV WKDW WKH V\VWHP GHVFULEHG E\ f ZLWK WUDQVIHU IXQFWLRQ PDWUL[ :] f FDQ EH VWDELOL]HG E\ G\QDPLF RXWSXW IHHGEDFN IRU DOO HIW )URP ZHOO NQRZQ UHVXOWV >@ WKLV DVVXPSWLRQ LV HTXLYDOHQW WR WKH IROORZLQJ UDQN FRQGLWLRQ
PAGE 20
UDQN >']Hf 1]f@ P IRU _]_ DQG DOO HIO ZKHUH P LV WKH QXPEHU RI V\VWHP RXWSXWV DQG Q fÂµ Q ']Hf , O $]1]f O %]M M O $V SUHYLRXVO\ PHQWLRQHG WKLV VWDELOL]DELOLW\ FRQGLWLRQ FDQ EH FKHFNHG VLQFH LW GRHV QRW GHSHQG RQ WKH SDUDPHWHU HVWLPDWHV $Q H[DPSOH ZLOO LOOXVWUDWH WKH WHVW ([DPSOH &RQVLGHU WKH V\VWHP GHVFULEHG LQ ([DPSOH >']Hf 1]Hf@ ] ] D] ] O] ] ,Q RUGHU WR VDWLVI\ WKH VWDELOL]DELOLW\ DVVXPSWLRQ WKH UDQN RI >']f 1]f@ PXVW HTXDO IRU _]_ DQG HQ 7KH GHWHUPLQDQWV RI WKH [ VXEPDWULFHV RI >']f 1]f@ DUH Df] ]O ]Df]f ]O]f $OO WKUHH PLQRUV DUH ]HUR RQO\ ZKHQ ] 9 DQG D +HQFH WKH UDQN FRQGLWLRQ LV YLRODWHG RQO\ ZKHQ ] 9 DQG D 7KHUHIRUH WKH VWDELOL]DELOLW\ DVVXPSWLRQ ZLOO EH VDWLVILHG IRU DQ\ LQWHUYDO ZKLFK GRHV QRW FRQWDLQ WKH SRLQW ^` 7KH DPRXQW RI HIIRUW LQYROYHG LQ FKHFNLQJ WKH UDQN FRQGLWLRQ LQ f ZLOO RI FRXUVH GHSHQG RQ WKH QXPEHU RI XQNQRZQ SDUDPHWHUV ,W PD\ EH SRVVLEOH WR XVH D URRWILQGLQJ DOJRULWKP IRU SRO\QRPLDOV LQ VHYHUDO YDULDEOHV LH ] DQG LMf WR FKHFN WKH UDQN RI >']f 1]Hf@ IRU DOO H* ,W VKRXOG DOVR EH QRWHG WKDW E\ RYHUHVWLPDWLQJ WKH PD[ STf LQ f WKH V\VWHP FDQQRW EH FRQWUROODEOH EXW PD\ EH VWDELOL]DEOH ,Q
PAGE 21
WKLV FDVH DGDSWLYH SROH SODFHPHQW ZRXOG QRW EH SRVVLEOH EXW D VWDELOL]LQJ IHHGEDFN FRXOG EH FRPSXWHG XVLQJ WKH 5LFFDWL DSSURDFK GLVFXVVHG LQ &KDSWHU ,, 7R VXPPDUL]H WKH DVVXPSWLRQV RQ WKH ULQSXW PRXWSXW OLQHDU GLVFUHWHWLPH V\VWHP GHVFULEHG E\ f f DUH $O $ $ $Q XSSHU ERXQG Q RQ S DQG T LQ f LV NQRZQ WKH FRPSRQHQWV HLM RI WKH XQNQRZQ V\VWHP SDUDPHWHU PDWUL[ H EHORQJ WR NQRZQ ERXQGHG LQWHUYDOV >HAÂ¯4A@ WKH V\VWHP GHVFULEHG E\ f LV VWDELOL]DEOH IRU DOO H EHORQJLQJ WR Q ZKHUH Q ^HH 51[P H >HAQ A[@` r r n 
PAGE 22
&+$37(5 ,9 3$5$0(7(5 (67,0$7,21 &RQVLGHU WKH ULQSXW PRXWSXW OLQHDU GLVFUHWHWLPH V\VWHP GHVFULEHG E\ \Nf bDNOf ANOf f $V SUHYLRXVO\ GLVFXVVHG LQ &KDSWHU ,,, LW LV DVVXPHG WKDW HDFK FRPSRn QHQW RI WKH XQNQRZQ V\VWHP SDUDPHWHU PDWUL[ EHORQJV WR D NQRZQ ERXQGHG LQWHUYDO >rfÂ°Q ,,A;@ / LM LM 7KH SDUDPHWHU HVWLPDWLRQ DOJRULWKP LV JLYHQ E\ 3NO8 Nf Nf Nf 3NOfINOff A [ f ? ANOf3NOfr Nf >\7Nf !NOfAANOfHNOf@ 3Nf 3NOf 3NO+DNOfANOf3NOf ?O AcNO:N'ANOfn 3f 37f f nfÂµLMNm PD[ LM ZKHQ Nf ! A[ ILMNOf A fÂµXWNfÂ¬Lf ,nLQ ,ZKHQ Nf J" r ?n 9 f ZKHQ Nf H>AQ r1[@
PAGE 23
ZKHQ WKH GHWHUPLQDQW RI 3Nf ! H QNB ZKHUH V LV DQ\ VPDOO SRVLWLYH QXPEHU f PD[ LL_! Nf LLf RWKHUZLVH N D 7KH SDUDPHWHU HVWLPDWLRQ DOJRULWKP GHVFULEHG E\ f f GLIIHUV IURP WKH UHFXUVLYH OHDVWVTXDUHV HVWLPDWRU LQ WKUHH ZD\V )LUVW WKH LQLWLDO FRYDULDQFH PDWUL[ 3f LQ f PXVW EH OHVV WKDQ ZKHUHDV WKH OHDVWVTXDUHV DOJRULWKP DOORZV IRU DQ\ SRVLWLYH GHILQLWH LQLWLDO FRYDULDQFH PDWUL[ 7KLV FRQGLWLRQ LV QHHGHG WR HQVXUH WKDW WKH SDUDPHWHU HVWLPDWRU GHVFULEHG E\ f f SRVVHVVHV WKH SURSHUWLHV OLVWHG LQ 7KHRUHP ZKLFK DUH QHFHVVDU\ IRU SURYLQJ JOREDO VWDELOLW\ RI WKH DGDSWLYH UHJXODWRU GLVFXVVHG LQ &KDSWHU 9 7KH VHFRQG GLIIHUHQFH LV WKH GDWD QRUPDOL]DWLRQ LQWURGXFHG WKURXJK WKH WHUP A GHILQHG LQ f ,Q WKH OHDVWVTXDUHV DOJRULWKP QN L LV VLPSO\ HTXDO WR RQH $JDLQ WKH GDWD QRUPDOL]DWLRQ LV QHFHVn VDU\ IRU SURYLQJ WKDW WKH SDUDPHWHU HVWLPDWRU KDV WKH GHVLUDEOH SURSHUn WLHV JLYHQ LQ 7KHRUHP 6LPXODWLRQV KDYH VKRZQ WKDW LI H LV QRW VXLWDEO\ VPDOO WKH QRUPDOL]DWLRQ KDV D GHWULPHQWDO HIIHFW RQ WKH WUDQVLHQW UHVSRQVH RI WKH V\VWHP 7KLV HIIHFW FDQ EH PLQLPL]HG E\ FKRRVLQJ H WR EH YHU\ VPDOO LQ f 6LPXODWLRQV ZKLFK VKRZ WKH HIIHFW RI WKH GDWD QRUPDOL]DWLRQ DQG WKH FKRLFH RI H RQ V\VWHP WUDQVLHQW UHn VSRQVH DUH GLVFXVVHG LQ &KDSWHU 9,, 7KH PDMRU GLIIHUHQFH EHWZHHQ WKH HVWLPDWRU JLYHQ E\ f f DQG WKH UHFXUVLYH OHDVWVTXDUHV HVWLPDWRU LV WKH DGGLWLRQ RI D FRUUHFn WLRQ WHUP 3NOfIHNOff LQ f 7KLV WHUP IRUFHV WKH HVWLPDWHV MMNf RI WKH FRPSRQHQWV RI WKH V\VWHP SDUDPHWHU PDWUL[ WR FRQYHUJH WR WKH VHWV &H"OQ}k7A;8 7KLV SURSHUW\ ZKHQ FRPELQHG ZLWK WKH VWDELOL] DELOLW\ DVVXPSWLRQ RYHU Q GLVFXVVHG LQ &KDSWHU ,,, HOLPLQDWHV WKH QHHG
PAGE 24
IRU D SHUVLVWHQWO\ H[FLWLQJ LQSXW ,W LV LPSRUWDQW WR QRWH WKDW WKH DOJRULWKP LV QRW D SURMHFWLRQ DOJRULWKP 7KDW LV cMNf GRHV QRW EHORQJ WR WKH VHW >OQQHWfÂ¬"[@ IRU HYHU\ N EXW GRHV FRQYHUJH WR WKH VHW 7 LQ WKH OLPLW DV N }fÂµ } 7KH LGHD IRU WKH FRUUHFWLRQ WHUP FDPH IURP .UHLVVHOPHLHU >@ ,Q IDFW IHNOff LV GHILQHG H[DFWO\ WKH VDPH DV LQ >@ KRZHYHU WKH SDUDPHWHU HVWLPDWRU GHVFULEHG KHUH IRU 0,02 GLVFUHWHWLPH V\VWHPV LV TXLWH GLIIHUHQW IURP WKH DGDSWLYH REVHUYHU .UHLVVHOPHLHU XVHV IRU 6,62 FRQWLQXRXVWLPH V\VWHPV 7KH IROORZLQJ WKHRUHP VKRZV WKH HVWLPDWRU GHVFULEHG E\ f f KDV PDQ\ RI WKH VDPH SURSHUWLHV DV UHFXUVLYH OHDVWVTXDUHV ,Q DGGLWLRQ WKH HVWLPDWHV MMNf FRQYHUJH WR WKH NQRZQ LQWHUYDOV >"@Q}A[@ 7KH SUHGLFWLRQ HUURU HNf ZKLFK DSSHDUV LQ WKH WKHRUHP LV GHILQHG DV HNf \Nf H7NOfM!DNOf ANOf
PAGE 25
7KHRUHP 7KH HVWLPDWLRQ DOJRULWKP f f KDV WKH IROORZLQJ SURSHUWLHV L 3Nf FRQYHUJHV WR D SRVLWLYH VHPLGHILQLWH PDWUL[ 3A LL _ Nf_ 0 p IRU DOO N DQG L 1 M OP P LHNf QNO !ANOf 3 Nf !DNOf DV N m LY LLHNfQ DNfDM!NfQ ENf ZKHUH DNf DQG ENf DUH SRVLWLYH VFDODU VHTXHQFHV ZKLFK FRQYHUJH WR ]HUR Y IHNff DV N } ZKLFK LPSOLHV WKDW MNf FRQYHUJHV WR WKH VHW >,nQ!A[@ IRU L 1 M P n r ?9, _ Nf NSf M rÂ‘ DV N } IRU DQ\ LQWHJHU S n n L \\ fÂµ fÂµ W \1\ M AA A YLL ,I r A RU A[ IRU HYHU\ L DQG M WKHQ Nf FRQYHUJHV n fÂµ 8 A M fÂµ IMURL Q PD[ WR D SRLQW LQ > @r r M r M 3URRI 3URSHUW\ Lf LV D ZHOONQRZQ SURSHUW\ ILUVW SURYHG E\ 6DPVRQ >@ ZLWK WLAL HTXDO WR RQH 7KH SURRI LV LQFOXGHG KHUH IRU WKH VDNH RI FRPSOHWHQHVV (TXDWLRQ f LPSOLHV WKDW [A3Nf[[A3NOf[ IRU DOO [H 51 DQG IRU DOO N ! 7KHUHIRUH IRU HDFK [ H(1 WKH VHn TXHQFH [A3Nf[ LV PRQRWRQH GHFUHDVLQJ DQG PXVW FRQYHUJH 'HQRWLQJ 3MMNf DV WKH LMA/ HOHPHQW RI 3Nf LW IROORZV WKDW H Hf73NfHL HAf HOSNfHL H7SNfH 3LM22 f
PAGE 26
ZKHUH 2 2 2 m LWK SRVLWLRQ (DFK RI WKH WHUPV RQ WKH ULJKW KDQG VLGH RI WKH HTXDWLRQ f PXVW FRQYHUJH ZKLFK LPSOLHV 3MMNf FRQYHUJHV IRU HYHU\ L DQG M 1RWLQJ WKDW 3f LW WKHQ IROORZV WKDW 3Nf PXVW FRQYHUJH WR D SRVLWLYH VHPL GHILQLWH PDWUL[ 3A 7R SURYH LLf OHW nHNf NfH 5HZULWLQJ f LQ WHUPV RI HNf JLYHV SNO+DNOfANOfHNOf HLNf nHNOf 3NOfIHNOff f ?O AcN'3NOfrDNOf 0XOWLSO\LQJ ERWK VLGHV RI WKH HTXDWLRQ f RQ WKH ULJKW E\ 3NOf JLYHV 3Nf3fÂ°Nf , 3NOfMDNOfANOf f ?O Naf3^Nn.Nnf &RPELQLQJ f DQG f \LHOGV HNf 3Nf3NOfHfÂ¬NOf SNOfIHNOff f
PAGE 27
1RZ GHILQH WKH /\DSXQRY IXQFWLRQDO 9Nf 7U&H7Nf3fÂ¯NfANff ! ZKHUH 7U GHQRWHV WKH WUDFH RSHUDWLRQ ,QVHUWLQJ HTXDWLRQ f LQWR WKH H[SUHVVLRQ IRU 9Nf JLYHV 9Nf 7U >n7NOf3NOfHNf@ 7U>I7HNOff3NOf3fÂ¯NfHNf@ f ,QVHUWLQJ WKH H[SUHVVLRQ f IRU HNf LQ WKH ILUVW WHUP RQ WKH ULJKW KDQG VLGH RI f \LHOGV 9Nf 9NOf 7fÂ§668LO9O DNfSNfrDNfÂ¯f 7U >7NOfIHNOff@ 7U>I7HNOff3NOf3fÂ¯NfHnNf@ f (TXDWLRQ f IRU HNf LV WKHQ LQVHUWHG LQWR WKH ODVW WHUP RQ WKH ULJKW KDQG VLGH RI f JLYLQJ 9Nf 9NOf QHNf LL 9O AcNOf3NOfDNOf 7U>7NOfINOff@ f ,W IROORZV IURP WKH GHILQLWLRQ RI INOff LQ HTXDWLRQ f WKDW 7U>I7NOffIINOf@ ! 7U>I7NOffINOff@ f
PAGE 28
&RPELQLQJ HTXDWLRQV f DQG f JLYHV 9Nf 9NOf LLHNf LLA QNO AcNOf3.O+DNOf f 7U&I7HNOff&,3NOf3Nf3NOf@INOff@ ,I WKH PDWUL[ 3NOf3rNf3NOf LV SRVLWLYH GHILQLWH IRU N VXIILn FLHQWO\ ODUJH WKHQ 9Nf ZLOO EH D PRQRWRQH GHFUHDVLQJ VHTXHQFH ZKLFK PXVW FRQYHUJH VLQFH 9Nf LV ERXQGHG EHORZ E\ ]HUR ,Q RUGHU WR VKRZ WKDW WKH PDWUL[ 3NOf3Nf3NOf LV SRVLWLYH GHILQLWH IRU N VXIILFLHQWO\ ODUJH ILUVW DVVXPH WKDW WKH GHWHUPLQDQW RI 3Nf GRHV QRW FRQYHUJH WR ]HUR ,Q WKLV FDVH 3ANf FRQYHUJHV WR Sar ZKLFK LPSOLHV WKDW 3NOf3ONf3NOf FRQYHUJHV WR 3A 7KH PDWUL[ 3A LV SRVLWLYH GHILQLWH E\ SURSHUW\ Lf LQ 7KHRUHP 0RZ VXSSRVH WKDW WKH GHWHUPLQDQW RI 3Nf GRHV FRQYHUJH WR ]HUR 8VLQJ f DQG WKH PDWUL[ LQYHUVLRQ OHPPD JLYHV ANL+AN' 3fÂ¯Nf SANOf N 0XOWLSO\LQJ ERWK VLGHV RI WKLV HTXDWLRQ RQ WKH OHIW DQG ULJKW E\ 3NOf DQG DGGLQJ WR ERWK VLGHV JLYH 3NOf3Nf3NOf ,3Nf 3NO+DNO+cNOf3NOf Q N f
PAGE 29
6LQFH 3Nf FRQYHUJHV HTXDWLRQ f LPSOLHV WKDW WKH PDWUL[ 3NOfM!DNOf`!JNOf3^NOf 9L NfÂ«` FRQYHUJHV WR ]HUR )RU N VXIILFLHQWO\ ODUJH Q PD[ Q! NOf Qf .O D ZKLFK LPSOLHV WKDW 3NO+DNOfANOf3NOf 3NO+DNOfNOf3NOf QNO L}DNf3NfDNf ?O [PD[3fDNf_= 3NOfÂ« NOfNOf3NOf ! D D ! R 9OA 3NO8 NOfNOf3NOf ,W WKHQ IROORZV WKDW WKH PDWUL[ fÂ§A FRQYHUJHV 9O WR ]HUR VLQFH LW LV ERXQGHG DERYH E\ D PDWUL[ ZKLFK LV FRQYHUJLQJ WR ]HUR 7KXV E\ HTXDWLRQ f WKH PDWUL[ 3NOf3ONf3NOf DOVR FRQYHUJHV WR 30 LQ WKH FDVH ZKHQ WKH GHWHUPLQDQW RI 3Nf FRQYHUJHV WR ]HUR 7KHUHIRUH WKH PDWUL[ 3NOf3ANf3NOf LV SRVLWLYH GHILQLWH IRU N VXIILFLHQWO\ ODUJH VLQFH WKH PDWUL[ FRQYHUJHV WR WKH SRVLWLYH GHILQLWH PDWUL[ 3A %\ f WKLV LPSOLHV WKDW 9Nf LV D PRQRWRQH GHFUHDVLQJ VHTXHQFH IRU N VXIILFLHQWO\ ODUJH DQG PXVW WKHUHn IRUH FRQYHUJH ,W WKHQ IROORZV WKDW Nf LV ERXQGHG SURSHUW\ LLff n E\ QRWLQJ WKDW
PAGE 30
9Nf 7UH7NfSNfnHNff ! >$A3n9f@ 7UH7NfnHNff 7 1 P YNf ! 9 7U7N!UNff 9 A A &HMNf fÂ¬M@ 1RZ VROYLQJ f UHFXUVLYHO\ JLYHV WKH UHODWLRQVKLS T LHMfLfÂ« 9Tf 9f , fÂ§ fÂ§< M O QMO f , 7U>I7MOff>, 3MOf3Mf3MOf@IHMOff@ M O )RU VXIILFLHQWO\ ODUJH N WKH WHUPV ZLWKLQ WKH VXPPDWLRQV DUH QRQQHJDn WLYH DQG PXVW FRQYHUJH WR ]HUR VLQFH 9Nf LV QRQQHJDWLYH 3URSHUWLHV LLLf DQG Yf WKHQ IROORZ HDVLO\ 3URSHUW\ LYf IROORZV LPPHGLDWHO\ IURP LLLf E\ ILUVW QRWLQJ WKDW LLHNf QNL r,NLfSNLfrD_Lff ENf ZKHUH ENf LV D SRVLWLYH VFDODU VHTXHQFH ZKLFK FRQYHUJHV WR ]HUR 0XOWLSO\LQJ ERWK VLGHV RI WKH LQHTXDOLW\ E\ QeBL I!DNOf3NOfM!DNOff DQG QRWLQJ WKDW LLI!DNf LL JLYHV QHNfm ENfQA AcNOf3NOfrDNOff ENf9O 8PD[3!Off rDNOfLf ENfL 8PD[3ff ;f +DNOfQ ENf DNf }M}DNOf Q ENf
PAGE 31
ZKHUH DNf DQG ENf DUH SRVLWLYH VFDODU VHTXHQFHV ZKLFK FRQYHUJH WR ]HUR 7R SURYH YLf ILUVW VXEWUDFW HANOf IURP ERWK VLGHV RI HTXDn WLRQ f 3UDPXOWLSO\LQJ ERWK VLGHV RI WKH UHVXOWLQJ HTXDWLRQ E\ LWV WUDQVSRVH DQG QRWLQJ WKDW HNf H7NOfc! Nf JLYHV 7U>HNfNOff7NfNOff@ 7U>HNf nNOff7Nf fÂµ"NOff! HNfNOf3NOfr NOfH7Nf \ S fÂ§ a T fÂ§fÂ§fÂ§ a a a 9O ANOf3NOfADNOff B 7U W 3NOfr NOfNOfNOf I7 NOfffÂ§ 9O ADNfÂ«fSNfÂ¬OfrDFBf f 7U>I7HNOff3NOfIHNOff@ $OVR A,NOf3NOfr Nf I!DNOf3NO+ Nf &:!rf 777 QB ANOf3NOfrDNOf QNODLNSNnfrDNfÂ«f ; 3f PD[ n f &RPELQLQJ f DQG f JLYHV 7U&NfHNOffNfHNOff &[PD[3f@ LHNf LLn 9O9NfÂ¬f3NBfDNfÂ¯f 7U 7 3NOfÂ« NOfNOfNOf I7NOffA D D QNO mD N'3NOfANOf rD[ 32ff7U>I7NOffINOff@ f
PAGE 32
%\ SURSHUWLHV LLLf DQG Yf WKH ILUVW DQG WKLUG WHUPV RQ WKH ULJKW VLGH RI WKH LQHTXDOLW\ f FRQYHUJH WR ]HUR 7KH VHFRQG WHUP RQ WKH ULJKW VLGH RI WKH LQHTXDOLW\ ZLOO DOVR FRQYHUJH WR ]HUR E\ SURSHUW\ Yf SURn YLGHG WKDW WKH PDWUL[ 3NOfrDNOfANOf 9O rcNOf3NO+DNOf LV ERXQGHG 7KLV PDWUL[ LV VKRZQ WR EH ERXQGHG E\ ILUVW QRWLQJ WKDW 3NOfM!DNOfANOf ?O cNOf3NO+DNOf 8NO+nLNOf 3ÂNf D D LL}DNOfLL nN LLL!DNOf \ $PLQ3NOf PLQ &OHDUO\ WKH QXPHUDWRU RQ WKH ULJKW KDQG VLGH RI WKH LQHTXDOLW\ LV ERXQGHG DERYH 7KH GHQRPLQDWRU LV ERXQGHG EHORZ E\ RQH LI WKH GHWHUn PLQDQW RI 3Nf FRQYHUJHV WR ]HUR DQG E\ [ 3 f LI WKH GHWHUPLQDQW PLQ Rr RI 3Nf GRHV QRW FRQYHUJH WR ]HUR 7KXV DOO WKUHH WHUPV RQ WKH ULJKW KDQG VLGH RI WKH LQHTXDOLW\ f FRQYHUJH WR ]HUR 6LQFH 7U>HNf HNOff7Nf HNOff@ >HNf HANOf@ M O X LW IROORZV WKDW Nf HNf PXVW FRQYHUJH WR ]HUR ZKLFK SURYHV fÂµ SURSHUW\ YLf LQ WKH FDVH ZKHUH S 7KH SURRI IRU DUELWUDU\ S IROn ORZV IURP WKH 6FKZDU] LQHTXDOLW\ 7R SURYH YLLf HTXDWLRQ f LV ILUVW VROYHG UHFXUVLYHO\ JLYLQJ HQf "QfSOf4f 3Qf O 3LOf3LfILff L
PAGE 33
,W WKHQ IROORZV WKDW HPf HQf nkPf HfÂ¬HQf 3Pf 3Qff3fÂ¯r2fHf P 3Pf O 3n0L Of3LfILff 3Qf < 3 Of3fI'f :LWKRXW ORVV RI JHQHUDOLW\ DVVXPH WKDW P LV JUHDWHU WKDQ Q 5HSODFLQJ WKH PDWUL[ 3Qf SUHFHGLQJ WKH VXPPDWLRQ ZLWK 3Pf 3Qf 3Pff JLYHV L P Pf Qf 3Pf 3Qff32ffÂ«2f 3Pf < 3L f3LfILff L Q 3Qf 3Pff O 3L Of3LfILff 8VLQJ WKH SURSHUWLHV RI WKH LQGXFHG PDWUL[ QRUP GHILQHG LQ &KDSWHU ,, JLYHV LMPf LMQf_ ORQQf Qf L L3Pf 3Qf LL Q3f"fQ P Q Q3P! Q O Â‘3fÂ¯LLOf3Lf L LLIHLff LL L Q f QO Q3Qf 3Pf LL O Q3 LOf3LfQ ,I Lff LL L ,I WKH ULJKW KDQG VLGH RI HTXDWLRQ f FDQ EH PDGH DUELWUDULO\ VPDOO IRU P DQG Q VXIILFLHQWO\ ODUJH WKHQ HDFK FRPSRQHQW HANf RI WKH PDWUL[ r 
PAGE 34
VHTXHQFH HNf ZLOO EH D &DXFK\ VHTXHQFH ,W ZLOO WKHQ IROORZ XVLQJ SURSHUW\ Yf WKDW HA Nf FRQYHUJHV WR D SRLQW LQ WKH LQWHUYDO >H"Q!H"[@ ,Q RUGHU WR SURYH WKDW WKH ULJKW KDQG VLGH RI WKH LQ 7 HTXDOLW\ f FDQ EH PDGH DUELWUDULO\ VPDOO WZR WKLQJV PXVW EH VKRZQ Q3 NOf3NfLL Y } IRU DOO N DQG e QIHNffLL / p N ,Q RUGHU WR SURYH Q3 NOf3NfQ Y m IRU DOO N ILUVW DVVXPH WKDW WKH GHWHUPLQDQW RI 3Nf GRHV QRW FRQYHUJH WR ]HUR ,Q WKLV FDVH 3BNf FRQYHUJHV WR 3 ZKLFK LPSOLHV WKDW 3fÂ°ANOf3Nf FRQYHUJHV WR RR , 7KHUHIRUH L 3r.Of3Nf Q LV ERXQGHG IRU DOO N ZKHQ WKH GHWHUn PLQDQW RI 3Nf GRHV QRW FRQYHUJH WR ]HUR 1RZ VXSSRVH WKH GHWHUPLQDQW RI 3Nf GRHV FRQYHUJH WR ]HUR $SSO\LQJ WKH PDWUL[ LQYHUVLRQ OHPPD WR HTXDWLRQ f JLYHV 3NOf3Nf , rDNfANf3Nf 8VLQJ WKH SURSHUWLHV RI WKH LQGXFHG PDWUL[ QRUP LW IROORZV WKDW L }I!DNf NfQ L3fÂ¯$NOf3Nf L LLOQ fÂ§ LL3Nf , f QN )URP f WKHUH H[LVWV D ILQLWH 0 VXFK WKDW QN PD[ QM!DNf LLf IRU N ! 0 ,W WKHQ IROORZV WKDW DOO WKH FRPSRQHQWV RI WKH PDWUL[ DNfrDNf QN KDYH PDJQLWXGH OHVV WKDQ RU HTXDO WR RQH IRU N ! 0 ZKLFK LQ WXUQ LPSOLHV AD22ANf QN +LOO DQG Q3NfQ LV ERXQGHG IRU DOO N HTXDWLRQ f LPSOLHV 1RWLQJ DOVR WKDW +3nANOM3Â¯Nf LL LV
PAGE 35
DOVR ERXQGHG IRU DOO N LQ WKH FDVH ZKHQ WKH GHWHUPLQDQW RI 3Nf FRQn YHUJHV WR ]HUR 7KHUHIRUH WKHUH H[LVWV D ILQLWH Y VXFK WKDW Q3fÂ¯rNOf3Nf LL Y m IRU DOO N $SSO\LQJ WKLV UHVXOW WR WKH LQn HTXDOLW\ f JLYHV D PO _L fÂµ PfH Qf _ Q3Pf3Qf Q Q3LffLL Y O QIHLffQ L Q Q Q3Pf3QfQY e QIHLffD f L ,Q RUGHU WR FRPSOHWH WKH FRQYHUJHQFH SURRI LW PXVW EH VKRZQ WKDW O LOIHNff Q / } ,W IROORZV IURP HTXDWLRQ f WKDW N 9Nf 9NOf LLHNf 9L cNOf3N'rDNOf 7U>H NOfIHNOff@ [PD[>3NOf3Nf3NOf@7U>I7NOffINOff@ f 1RZ DVVXPH A; IRU L 1 DQG M P LH L fÂµ LV DQ LQWHULRU SRLQW RI >HO Qn"[@f 8VLQJ WKH GHILQLWLRQ RI nLM LM INOff LQ f LI ILMNOff r WKHQ _HLMNOf_ ! T ! ZKHUH T fÂµ Q ,7; PLULQ D fÂµ L r >mM mLM nLM}@ Â 1 fÂ« M O P $OVR IURP WKH GHILQLWLRQ RI IHNOff LW PXVW EH WUXH WKDW WKH VLJQ RI HANOf LV HTXDO WR WKH VLJQ RI ILQNOff IRU L O LM 1 DQG M O P 7KHVH IDFWV ZKHQ FRPELQHG ZLWK HTXDWLRQ f LPSO\ 9Nf 9NOf 1 P \ fÂ§ T O e _ILL NOff n OQn n n L O M O QHNf LLn QNO L!N'3NOfM}DNOf A[&3ÂNQ3nANÂ¯3Â¯NPU&AÂ¯H88QWHWNff@
PAGE 36
5HDUUDQJLQJ WHUPV LQ WKLV LQHTXDOLW\ UHVXOWV LQ ,, F ? 3LW X JT 9O ANOf3NOfDNOf 9NOf 9Nf ;A&A'3n 1 P , O L O M O ILMN'f Nf3NOf@7U>I7NOffINOff@ 6XPPLQJ DOO WKH WHUPV LQ WKLV LQHTXDOLW\ IURP N HTXDOV RQH WR 0 DQG 1 P QRWLQJ WKDW QIHNOffQ O O _IcBHNOff_ JLYHV M O 0 , QHNf LLr 0 W T O OOIHNOff N QNO rNf3NfrD_Ff N 0 f 9f 90f O fÂµ$PD[&3NOf3fÂ¯Nf3NOf@7U&I7NOffINOff@ N O ,I WKH ULJKW KDQG VLGH RI WKH LQHTXDOLW\ f LV ERXQGHG LQ WKH OLPLW DV 0 DSSURDFKHV LQILQLW\ WKHQ WKH QRUP RI IHNff ZLOO EH VXPPDEOH 7KH WHUP 90f LV ERXQGHG VLQFH 9Nf FRQYHUJHV $OVR $PD[&3LN'3A2238Of@ LV ERXQGHG VLQFH WKH PDWUL[ 3NOf3fÂ¯Nf3NOf FRQYHUJHV WR 3A ,W UHPDLQV WR EH VKRZQ WKDW O 7U>I7HNOffIHNOff@ N O (TXDWLRQ f LPSOLHV O 7U>I7NOff>,3NOf3nNf3NOf@IHNOff@ N O
PAGE 37
$OVR WKHUH H[LVWV D ILQLWH 1 VXFK WKDW WKH PDWUL[ >,3NOf3rNf3NOf@ LV SRVLWLYH GHILQLWH IRU N ! 1 EHFDXVH >,3NOf3ONf3NOf@ FRQYHUJHV WR WKH SRVLWLYH GHILQLWH PDWUL[ 30 6LQFH O [PLQ>,3NOf3Nf3Nf3NOf@7U>I7HNOffINOff@ N 1 PLQ , 7U>I7NOff>,3NOf3fÂ¯Nf3NOf@INOff@ N 1 LW IROORZV WKDW O 7U>IANOffINOff@ } 7KHUHIRUH WKH ULJKW N O KDQG VLGH RI WKH LQHTXDOLW\ f LV ERXQGHG LQ WKH OLPLW DV 0 DSn SURDFKHV LQILQLW\ 7KLV LPSOLHV WKH H[LVWHQFH RI D ILQLWH / VXFK WKDW O ,OINff LL / } 5HWXUQLQJ WR WKH LQHTXDOLW\ f OHW N O H ! EH JLYHQ 7KH VHTXHQFH AMNf ZLOO EH D &DXFK\ VHTXHQFH LI AMPfcMQf _ LV OHVV WKDQ H IRU P DQG Q VXIILFLHQWO\ ODUJH 6LQFH 3Nf FRQYHUJHV WR 3A WKHUH H[LVWV 1AHf VXFK WKDW Q3Pf3QfQ IRU PQ ! 1AHf ZKHUH eO f PLQ >HQ32f2f LL HY/f@ $OVR WKHUH H[LVWV 18f VXFK WKDW P O QIHLffQ HY IRU PQ ! 1"Hf L Q Â
PAGE 38
,W WKHQ IROORZV IURP f WKDW _HAPf AQf_ V IRU P Q ! PD[ 1MAf &RPELQLQJ WKLV ZLWK SURSHUW\ Yf HAMNf FRQYHUJHV WR D SRLQW LQ > @ IRU L O 1 DQG M O P , $V GLVFXVVHG LQ &KDSWHU ,, SURSHUWLHV LLfLYf DQG YLf DUH VWDQGDUG SURSHUWLHV UHTXLUHG IRU DQ\ SDUDPHWHU HVWLPDWRU XVHG LQ LQn GLUHFW DGDSWLYH FRQWURO 3URSHUW\ Yf IRUFHV WKH SDUDPHWHUV WR FRQYHUJH WR D VXEVHW RI @5A[P ZKLFK FRQWDLQV QR VLQJXODU SRLQWV 7KLV DGGLn WLRQDO SURSHUW\ ZLOO HQVXUH WKDW WKH HVWLPDWHG V\VWHP LV XQLIRUPO\ VWDELOL]DEOH WKXV HOLPLQDWLQJ WKH QHHG IRU D SHUVLVWHQWO\ H[FLWLQJ LQSXW 3URSHUW\ YLLf WKH FRQYHUJHQFH RI cMNf WR D SRLQW LQ >"@Q}A;@} LV QRW UHTXLUHG IRU SURYLQJ JOREDO VWDELOLW\ RI WKH DGDSWLYH FORVHGORRS V\VWHP 3URSHUW\ YLLf ZLOO EH XVHG LQ &KDSWHU 9, IRU WUDFNLQJ DQG GLVWXUEDQFH UHMHFWLRQ EDVHG RQ WKH LQWHUQDO PRGHO SULQFLSOH ,W LV LQWHUHVWLQJ WKDW XQOLNH WKH FDVH RI OHDVW VTXDUHV ZKHUH HNf FRQYHUJHV WR 3RR32fn2f DV N Â« m WKHUH LV QR FORVHG IRUP H[SUHVVLRQ IRU WKH OLPLW RI Nf DV N !Â‘ } ZKHQ XVLQJ WKH SDUDPn HWHU HVWLPDWRU GHVFULEHG E\ ff +RZHYHU LI 3Nf FRQYHUJHV WR ]HUR WKHQ HNf FRQYHUJHV WR WKH WUXH V\VWHP SDUDPHWHU PDWUL[ H IRU ERWK WKH SDUDPHWHU HVWLPDWRU GLVFXVVHG LQ WKLV FKDSWHU DQG WKH UHFXUVLYH OHDVWVTXDUHV HVWLPDWRU 7KLV LV HDVLO\ VKRZQ XVLQJ WKH UHFXUVLYH VROXWLRQ RI f ZKLFK LV JLYHQ E\ "Qf 3Q!S^fnHf 3Qf 3n Of3OfIHff L 4 7DNLQJ WKH QRUP RI ERWK VLGHV DQG XVLQJ WKH SURSHUWLHV RI WKH LQGXFHG PDWUL[ QRUP LW IROORZV WKDW
PAGE 39
QHQfQ L3Qf Q3fnHf Q Q3Qf X Q O 3LOf3L fI HLff Q L QfÂ¯U ,W ZDV VKRZQ SUHYLRXVO\ WKDW LL O 3 LOf3L fI HLff Q LV ERXQGHG LQ L 4 WKH OLPLW DV Q Â« m $OVR LL3QfQ FDQ EH PDGH DUELWUDULO\ VPDOO IRU Q VXIILFLHQWO\ ODUJH LI 3Nf FRQYHUJHV WR ]HUR 7KXV HAQf FRQYHUJHV WR ]HUR DV Q } ZKLFK LPSOLHV HNf FRQYHUJHV WR WKH WUXH V\VWHP SDUDPn HWHU PDWUL[ H ZKHQHYHU 3Nf FRQYHUJHV WR ]HUR
PAGE 40
&+$37(5 9 $'$37,9( 5(*8/$725 7KH ULQSXW PRXWSXW OLQHDU GLVFUHWHWLPH V\VWHP GHVFULEHG E\ ff ZLOO EH UHJXODWHG XVLQJ D 0,02 YHUVLRQ RI 6DPVRQ DQG )XFKVn DGDSWLYH /4 FRQWUROOHU >@ $V GLVFXVVHG LQ &KDSWHU 7ZR D QHFHVVDU\ FRQGLWLRQ IRU WKH DGDSWLYH FORVHGORRS V\VWHP WR EH JOREDOO\ VWDEOH LV D XQLIRUPO\ VWDELOL]DEOH HVWLPDWHG V\VWHP ,W ZLOO EH VKRZQ WKDW WKH SDUDPHWHU HVWLPDWRU GHVFULEHG E\ HTXDWLRQV ff ZLOO JLYH D XQLIRUPO\ VWDELOL]DEOH HVWLPDWHG V\VWHP SURYLGHG WKDW DVVXPSWLRQV $$ VXPPDUL]HG DW WKH HQG RI &KDSWHU 7KUHH DUH VDWLVILHG 8QLIRUP 6WDELOL]DELOLW\ RI WKH (VWLPDWHG 6\VWHP 8VLQJ WKH SDUDPHWHU HVWLPDWRU GHVFULEHG E\ ff LW ZLOO EH VKRZQ WKDW WKH HVWLPDWHG V\VWHP FDQ EH VWDELOL]HG XVLQJ VWDWH IHHGEDFN $V VKRZQ LQ >@ WKH VWDWHVSDFH REVHUYHU IRUP UHDOL]DWLRQ IRU WKH V\VWHP GHVFULEHG E\ f LV JLYHQ E\ [NOf )[Nf *XNf f \Nf +[Nf
PAGE 41
ZKHUH r fÂ¯ $ B$ fÂµ fÂµ * fÂµ fÂµ fÂµ fÂµ 9L , %Q fÂ¯$Q %Q + >, @ 7KH HQWULHV LQ ) DQG * GHSHQG RQ WKH XQNQRZQ V\VWHP SDUDPHWHU PDWUL[ ,Q IDFW E\ WKH GHILQLWLRQ RI H LQ f ) DQG * FDQ EH YLHZHG DV FRQWLQXRXV IXQFWLRQV RI ZKHUH UDQJHV RYHU WKH SDUDPHWHU VSDFH Q ^H @5A[P H>"QfÂµ rLO"[@` 7KLV GHSHQGDQFH FDQ EH PDGH 7 7 H[SOLFLW E\ GHQRWLQJ ) DQG * DV )Hf DQG *Hf UHVSHFWLYHO\ $VVXPSWLRQ $ LQ &KDSWHU 7KUHH LV HTXLYDOHQW WR UHTXLULQJ WKDW WKH V\VWHP GHVFULEHG E\ )f*ff LV SRLQWZLVH VWDELOL]DEOH RYHU Q 'HILQLWLRQ 7KH SDLU )Jf*ff LV SRLQWZLVH VWDELOL]DEOH RYHU IL LI IRU HDFK EHORQJLQJ WR Q WKHUH H[LVWV D IHHGEDFN PDWUL[ /Hf VXFK WKDW 8PDm)f*f/ff, Tf ZKHUH [A fÂµf LV WKH HLJHQYDOXH RI PD; PD[ fÂµf ZLWK WKH JUHDWHVW PDJQLWXGH ,W IROORZV IURP NQRZQ UHVXOWV >@ WKDW SRLQWZLVH VWDELOL]DELOLW\ RI )f *Hff RYHU Q LV HTXLYDOHQW WR WKH UDQN FRQGLWLRQ f 7KXV VWDELOL]DELOLW\ RYHU D FDQ EH FKHFNHG XVLQJ WKH SRO\QRPLDO PDWULFHV ']f DQG 1]f FRPSULVLQJ WKH V\VWHP WUDQVIHU IXQFWLRQ *LYHQ H ! OHW EH DQ H[WHQVLRQ RI WKH SDUDPHWHU VSDFH Q GHILQHG DV 4 H UQ P1[UQ fÂ± U$PLQ $PD[/ L ^H ) H> H H@` f
PAGE 42
3URSRVLWLRQ 6XSSRVH DVVXPSWLRQ $ LV VDWLVILHG VR WKDW )Hf *Hff LV SRLQWZLVH VWDELOL]DEOH RYHU Q 7KHQ WKHUH H[LVWV DQ H ! VXFK WKDW )Hf*Hff LV DOVR SRLQWZLVH VWDELOL]DEOH RYHU Q 3URRI 6XSSRVH )Hf}*Hff LV VWDELOL]DEOH RYHU IW )RU HDFK EHORQJLQJ WR IW WKHUH H[LVWV D ERXQGHG RSHQ QHLJKERUKRRG 94 RI VXFK WKDW )Hf*ff LV VWDELOL]DEOH RYHU 9 7KLV UHVXOW LV SURYHG LQ >@ E\ ILUVW VKRZLQJ WKDW WKHUH H[LVWV D IHHGEDFN PDWUL[ /Hf ZLWK HQWULHV ZKLFK DUH FRQWLQXRXV LQ VXFK WKDW _[ )f*Hf/ff_ Tf 7KH H[LVWHQFH RI WKH RSHQ QHLJKERUKRRG 9 RI WKHQ IROORZV EHFDXVH WKH HLJHQYDOXHV RI )Hf*Hf/ff DUH FRQWLQXRXV IXQFWLRQV RI 6LQFH IW LV D FRPSDFW VXEVHW RI ,nr[P DQG 8 9 LV DQ RSHQ FRYHU RI Q WKHUH H[LVWV D HIW ILQLWH VXEFRYHU 8 9 RI IW )RU N O OHW HN 9 fN )RU M O M U HYHU\ N ! Q 8 9D LVD FRPSDFW VXEVHW RI )1[P DQG HN M O M U U IW 8 94 & IW Â‘ 8 9IO HNO M O M HN M O M 6XSSRVH IW 8 9 LV QRQHPSW\ IRU DOO N ! %\ WKH SURSHUWLHV RI HN M O M QHVWHG FRPSDFW VHWV VHH > S@f WKHUH LV DW OHDVW RQH SRLQW S ZKLFK EHORQJV N ! U WR IW 8 IRU DOO N ! HN ML HM 7KHUHIRUH QHQ HN ZKLFK LPSOLHV WKDW SHQ +RZHYHU SA U 8 9 M O IRU DOO ZKLFK IRUPV DQ
PAGE 43
RSHQ FRYHU RI WL DQG KHQFH ZH KDYH D FRQWUDGLFWLRQ 7KXV U IOB f 9 PXVW EH HPSW\ IRU N M}L HM ZLOO EH SRLQWZLVH VWDELOL]DEOH RYHU N ! NM ZKLFK LPSOLHV )f*Hff Q IRU DQ\ N ! N HN /HW Nf EH GHILQHG DV WKH HVWLPDWH DW WLPH N RI WKH V\VWHP SDUDPHWHU PDWUL[ DQG OHW )Nf*Nff GHQRWH WKH WLPHYDU\LQJ HVWLPDWHG V\VWHP )Nff *Nfff 'HILQLWLRQ 7KH HVWLPDWHG V\VWHP )Nf*Nff LV XQLIRUPO\ VWDELOL]DEOH DV D WLPHYDU\LQJ V\VWHP >@ LI WKHUH H[LVWV D ERXQGHG IHHGEDFN PDWUL[ ONf VXFK WKDW )Nf*Nf/Nff LV XQLIRUPO\ DV\PSRWRWLFDOO\ VWDEOH DV D WLPHYDU\LQJ V\VWHP PDWUL[ 7KH IROORZLQJ WKHRUHP JLYHV FRQGLWLRQV XQGHU ZKLFK WKH HVWLPDWHG V\VWHP LV XQLIRUPO\ VWDELOL]DEOH 7KHRUHP $VVXPLQJ $$ DUH VDWLVILHG WKH SDUDPHWHU HVWLPDWRU GHVFULEHG E\ ff ZLOO JHQHUDWH DQ HVWLPDWHG V\VWHP )Nf*Nff ZKLFK LV XQLIRUPO\ VWDELOL]DEOH DV D WLPHYDU\LQJ V\VWHP 3URRI %\ 3URSRVLWLRQ WKHUH H[LVWV H ! VXFK WKDW )f*Hff LV SRLQWZLVH VWDELOL]DEOH IRU DOO EHORQJLQJ WR Qe 8VLQJ WKH UHVXOWV LQ .DPHQ DQG .KDUJRQHNDU >@ VLQFH LV FRPSDFW WKHUH H[LVWV D IHHGEDFN PDWUL[ /f ZLWK HQWULHV ZKLFK DUH FRQWLQXRXV LQ DQG D SRVLWLYH FRQVWDQW T VXFK WKDW _[ )Jf*Hf/Hff, T IRU DOO HIL L PD[? fÂµ H
PAGE 44
3URSHUW\ Yf LQ 7KHRUHP LPSOLHV WKDW WKHUH H[LVWV D ILQLWH 1 VXFK WKDW HNf EHORQJV WR QH IRU DOO N ! 1 ,W WKHQ IROORZV IURP WKH UHVXOW LQ .DPHQ DQG .KDUJRQHNDU JLYHQ DERYH WKDW WKHUH H[LVWV D IHHGEDFN JDLQ VHTXHQFH /Nf FRQWLQXRXV LQ VXFK WKDW T IRU DOO N ! 1 IQD; 6LQFH WKH HQWULHV LQ )Nf*Nf/Nff DUH FRQWLQXRXV IXQFWLRQV RI SURSHUW\ YLf LQ 7KHRUHP LPSOLHV >)Nf*Nf/Nf@ >)NOf*NOf/NOf@ FRQYHUJHV WR ]HUR DV N JRHV WR LQILQLW\ 2HVRHU >@ SURYHV $Nf ZLOO EH XQLIRUPO\ DV\PSWRWLFDOO\ VWDEOH DV D V\VWHP PDWUL[ LI Lf DOO HLJHQYDOXHV RI $Nf KDYH PDJQLWXGH OHVV WKDQ VRPH FRQVWDQW T IRU DOO N DQG LLf $Nf$Nf FRQYHUJHV WR ]HUR DV N JRHV WR LQILQLW\ 8VLQJ 'HVRHUnV UHVXOWV RQ VWDELOLW\ RI VORZO\ WLPH YDU\LQJ V\VWHPV )Nf*Nf/Nff LV XQLIRUPO\ DV\PSWRWLFDOO\ VWDEOH DV D V\VWHP PDWUL[ 7KXV WKH HVWLPDWHG V\VWHP )Nf*Nff LV XQLIRUPO\ VWDELOL]DEOH DV D WLPHYDU\LQJ V\VWHP 7KH DGDSWLYH FRQWURO ODZ IRU UHJXODWLQJ WKH V\VWHP GHVFULEHG E\ f LV RI WKH IRUP XNf /Nf[Nf 7KH VWDWH HVWLPDWH [Nf LV JHQHUDWHG IURP DQ DGDSWLYH REVHUYHU ZKLFK ZLOO EH GLVFXVVHG LQ WKH QH[W VHFWLRQ 7KH IHHGEDFN JDLQ PDWUL[ /Nf LV FKRVHQ WR VWDELOL]H WKH HVWLPDWHG V\VWHP )Nf*Nff 7KH H[LVWHQFH RI D VWDELOL]LQJ IHHGEDFN JDLQ /Nf LV JXDUDQWHHG E\ 7KHRUHP 6HYHUDO SRVVLEOH RSWLRQV IRU /Nf DUH JLYHQ LQ WKH VHFWLRQ IROORZLQJ WKH DGDSWLYH REVHUYHU VHFWLRQ
PAGE 45
$GDSWLYH 2EVHUYHU 7KH DGDSWLYH REVHUYHU IRU WKH HVWLPDWHG V\VWHP LV JLYHQ E\ [NOf )Nf[.f *NfXNf 0Nf\Nf\Nff f \Nf +!W Nf ZKHUH rfÂ¯$Nf , Â‘$Nf fÂµ fÂµ ?ZfÂ« %Nf fÂµ )Nf fÂµ fÂµ , B$QNf *Nf fÂµ fÂµ %QNf 0Nf $[Nf $Nf + >, 7KH DGDSWLYH REVHUYHU GHILQHG E\ f LV D GHDGEHDW REVHUYHU 7KH RXWSXW \Nf RI WKH DGDSWLYH REVHUYHU FDQ EH H[SUHVVHG LQ WKH IROORZLQJ IRUP \Nf &$M8f fÂµ fÂµ fÂµ $QNQf %Nf %QNQf@M!NOf f
PAGE 46
7KLV H[SUHVVLRQ LV HDVLO\ GHULYHG E\ ILUVW UHZULWLQJ f DV r , fÂµ fÂµ fÂµ fÂµ fÂµ fÂµ %[Nf %Nf fÂµ fÂµ XNOf $NffÂ° $Nf fÂµ fÂµ \NLf L fÂµ %QNf fÂµ BfÂ¯$QNOfB f 0XOWLSO\LQJ HTXDWLRQ f E\ + JLYHV \Nf +[Nf > @[NOf %AN'X8Of $AN'\LNOf 6XEVWLWXWLQJ LQWR WKLV HTXDWLRQ IRU [NOf XVLQJ f \LHOGV \Nf > @[Nf %NOfXNOf %NOfXNf $[NOf\NfÂ§f $Nf\Nf &RQWLQXLQJ WKH VXEVWLWXWLRQ XVLQJ f JLYHV r Q \Nf > ,@ [NQOf O % NMfXNMf M O Q , $NMf\NMf M O O % NMfXNMf O $ NMf\NMf M O ZKLFK LV HTXLYDOHQW WR f
PAGE 47
7KH HUURU EHWZHHQ WKH RXWSXW RI WKH DGDSWLYH REVHUYHU f DQG WKH RXWSXW RI WKH V\VWHP f DV GHILQHG DV HNf \Nf \Nf 7KH REVHUYHU HUURU HNf KDV WKH VDPH SURSHUWLHV DV WKH SUHGLFWLRQ HUURU HNf JLYHQ LQ 7KHRUHP WKDW LV LLHNfQ DnNfLrNOfL EnNf f ZKHUH DNf DQG EnNf DUH SRVLWLYH VFDODU VHTXHQFHV ZKLFK FRQYHUJH WR ]HUR 7KLV LV VKRZQ E\ ILUVW UHZULWLQJ HNf DV HNf \Nf >7NO+DNOf r?NOf@ >H7NOfM!DNOf m?NOf@ \Nf HNf \Nf >H7NO+DNOf L}ENOf@ HNf 7DNLQJ WKH QRUP RI ERWK VLGHV DQG XVLQJ WKH WULDQJOH LQHTXDOLW\ JLYHV LLHNfLL Q\Nf 7NOfMfDNOf LS7M!ENOffLL QHNfQ 1RWLQJ IURP ff WKDW H7Nf!DNf AANOf LV HTXDO WR >$ANOf $QNf %Nf QNf@c!NOf DQG XVLQJ H[SUHVVLRQ f JLYHV
PAGE 48
QÂ«NfLL Q&A Nf QNf@MfNf Q LLHNf Q f ZKHUH ,W IROORZV IURP SURSHUW\ YLf LQ 7KHRUHP WKDW QGNfLW FRQYHUJHV WR ]HUR DV N JRHV WR LQILQLW\ IRU L O Q $SSO\LQJ WKLV SURSHUW\ DQG SURSHUW\ LYf RI 7KHRUHP WR f JLYHV WKH GHVLUHG LQHTXDOLW\ f 3URSHUW\ f LV YHU\ LPSRUWDQW LQ WKH SURRI RI JOREDO VWDELOLW\ RI WKH DGDSWLYH FORVHGORRS V\VWHP )HHGEDFN *DLQ 6HTXHQFH 7KH IHHGEDFN JDLQ VHTXHQFH /Nf PXVW EH FKRVHQ VR WKDW )Nf *Nf/Nff LV H[SRQHQWLDOO\ VWDEOH DV D WLPHYDU\LQJ V\VWHP PDWUL[ $V VKRZQ LQ 7KHRUHP D VWDELOL]LQJ IHHGEDFN GRHV H[LVW EHFDXVH WKH HVWLPDWHG V\VWHP )Nf*Nff JHQHUDWHG IURP WKH SDUDPHWHU HVWLPDWRU ff LV XQLIRUPO\ VWDELOL]DEOH XQGHU DVVXPSWLRQV $$ 7KHUH DUH VHYHUDO RSWLRQV LQ WKH OLWHUDWXUH IRU FKRRVLQJ /Nf VRPH RI ZKLFK ZLOO EH GLVFXVVHG LQ WKLV VHFWLRQ 2QH DSSURDFK IRU FRPSXWLQJ /Nf LV WR VWDELOL]H WKH HVWLPDWHG V\VWHP )Nf*Nff SRLQWZLVH LQ WLPH XVLQJ FRQWURO ODZ VWUDWHJLHV IRU WLPHLQYDULDQW V\VWHPV VXFK DV SROH SODFHPHQW RU /4 FRQWURO 2I FRXUVH LQ WKH FDVH RI SROH SODFHPHQW WKH VWDELOL]DELOLW\ DVVXPSWLRQ RYHU WKH SDUDPHWHU VSDFH Q DVVXPSWLRQ $f PXVW EH VWUHQJWKHQHG WR UHDFKDELOLW\
PAGE 49
RYHU IO 7KLV DSSURDFK KDV EHHQ GHYHORSHG E\ D QXPEHU RI LQGLYLGXDOV VHH IRU H[DPSOH >@ $ GUDZEDFN WR WKLV DSSURDFK LV WKDW )Nf*Nff PD\ QRW EH UHDFKDEOH RU VWDELOL]DEOH DW FHUWDLQ SRLQWV LQ WLPH LH Nf PD\ EH D VLQJXODU SRLQW IRU VRPH YDOXHV RI Nf 5HFDOO WKDW 3URSRVLWLRQ DQG SURSHUW\ Yf RI 7KHRUHP RQO\ JXDUDQWHH WKDW )Nf*Nff LV VWDELOL]DEOH IRU DOO N H[FHHGLQJ VRPH ILQLWH 1 7KXV IRU WLPH SRLQWV SULRU WR 1 WKH FKRVHQ FRQWURO ODZ PD\ QRW KDYH D VROXWLRQ 7KHUHIRUH LQ WKLV DSSURDFK LW LV QHFHVVDU\ WR FKHFN ZKHWKHU RU QRW )Nf*Nff LV UHDFKDEOH RU VWDELOL]DEOH DW HDFK WLPH SRLQW N ZKLFK DGGV WR WKH QXPEHU RI RQOLQH FDOFXODWLRQV $OVR D GHFLVLRQ PXVW EH PDGH RQ KRZ WR PRGLI\ WKH FRQWURO ODZ ZKHQ VLQJXODU SRLQWV DUH HQFRXQWHUHG $ FRPPRQ FKRLFH LQ WKH OLWHUDWXUH LV WR XSGDWH WKH FRQWURO ODZ RQO\ ZKHQ )Nf*Nff LV VWDELOL]DEOH $QRWKHU DSSURDFK ILUVW LQWURGXFHG E\ .UHLVVHOPHLHU IRU FRQWLQXRXVn WLPH V\VWHPV LV WR FRPSXWH WKH IHHGEDFN JDLQ DV\PSWRWLFDOO\ >@ 7KLV DSSURDFK ZDV GHYHORSHG IRU 6,62 GLVFUHWHWLPH V\VWHPV E\ 6DPVRQ DQG )XFKV >@ XVLQJ DQ /4 FRQWURO VWUDWHJ\ DQG ODWHU H[WHQGHG WR 0,02 GLVFHWHWLPH V\VWHPV E\ 2VVPDQ DQG .DPHQ >@ ,Q >@ WKH VWDELOL]LQJ IHHGEDFN /Nf LV FRPSXWHG DW HDFK WLPH SRLQW N E\ VROYLQJ LQ UHDO WLPH RQH VWHS RI D 5LFFDWL GLIIHUHQFH HTXDWLRQ 0RUH SUHFLVHO\ OHW /Nf >*7Nf5N*Nf ,@B*7Nf5N)Nf f ZKHUH 5Nf LV WKH VROXWLRQ WR WKH 5LFFDWL GLIIHUHQFH HTXDWLRQ 5N 4 /7Nf/Nf )Nf *Nf/Nff75N)Nf *Nf/Nff f
PAGE 50
,Q f ERWK DQG WKH LQLWLDO YDOXH 5 DUH SRVLWLYH GHILQLWH V\PPHWULF PDWULFHV ,Q FRQWUDVW WR WKH ILUVW DSSURDFK ZKHUH WKH DOJHEUDLF 5LFFDWL HTXDWLRQ ZRXOG EH VROYHG IRU HDFK N WKLV DV\PSWRWLF DSSURDFK UHTXLUHV WKH VROXWLRQ RI RQO\ RQH LWHUDWLRQ RI WKH 5LFFDWL GLIIHUHQFH HTXDWLRQ DW HDFK SRLQW LQ WLPH 2EYLRXVO\ WKLV DSSURDFK RIIHUV FRQVLGHUDEOH VDYLQJV LQ RQOLQH FDOFXODWLRQV $OVR LW LV QRW QHFHVVDU\ WR FKHFN ZKHWKHU RU QRW )Nf*Nff LV VWDELOL]DEOH DW HDFK SRLQW N EHFDXVH /Nf JLYHQ E\ ff LV ZHOOGHILQHG HYHQ IRU LVRODWHG VLQJXODU SRLQWV ,Q RUGHU WR SURYH WKDW )Nf*Nf/Nff LV H[SRQHQWLDOO\ VWDEOH ZKHQ /Nf LV FRPSXWHG XVLQJ ff LW LV ILUVW QHFHVVDU\ WR VKRZ Q/Nf/NOffQ FRQYHUJHV WR ]HUR DV N } 7KHRUHP 6XSSRVH WKH SDUDPHWHU HVWLPDWHV DUH JHQHUDWHG XVLQJ WKH DOJRULWKP ff 7KHQ VXEMHFW WR DVVXPSWLRQV $$ WKH IHHGEDFN JDLQ VHTXHQFH JLYHQ E\ ff KDV WKH SURSHUW\ Q/Nf/NOfQ DV N m 3URRI 7KH SURRI LV UDWKHU OHQJWK\ DQG ZLOO WKHUHIRUH EH GLYLGHG LQWR VWHSV 7KURXJKRXW WKH SURRI 3Nf ZLOO UHIHU WR WKH VHTXHQFH JHQHUDWHG IURP WKH WLPHLQYDULDQW 5LFFDWL GLIIHUHQFH HTXDWLRQ 5'(f JLYHQ E\ 3NHf 4 )7Hf3NHf)Hf )7Hf3Nf*Hf>L *7f3NHf*f@fÂ¯ [ *7f3Nf)f 34 ! R 7KH PDWUL[ 3AHf ZLOO GHQRWH WKH VROXWLRQ WR WKH DOJHEUDLF 5LFFDWL HTXDWLRQ $5(f JLYHQ E\
PAGE 51
3Hf 4 IWfSRRfIf IWfSRRfJf>L JWfSZfJf@fÂ¯ ; *7f3RRf)f /HW IWJ GHQRWH WKH FRPSDFW SDUDPHWHU VSDFH GHILQHG LQ f VXFK WKDW )Hf*ff LV SRLQWZLVH VWDELOL]DEOH IRU DOO EHORQJLQJ WR IWJ /HW H ! EH JLYHQ ,W ZLOO ILUVW EH VKRZQ WKDW Q5A 5A ALL H IRU N VXIILFLHQWO\ ODUJH ZKHUH 5A LV FRPSXWHG XVLQJ f 6WHS 7KH VHTXHQFH RI PDWULFHV 5A JLYHQ E\ HTXDWLRQ f LV ERXQGHG 7KDW LV 5A EHORQJV WR D FRPSDFW VXEVHW ' RI A[1 7 7 $VVXPSWLRQ $ LPSOLHV ) f* kff LV GHWHFWDEOH IRU DOO EHORQJLQJ WR WKH SDUDPHWHU VSDFH IW 8VLQJ DUJXPHQWV VLPLODU WR WKRVH LQ 7KHRUHP LW IROORZV HDVLO\ WKDW )7Nf*7Nff LV XQLIRUPO\ GHWHFWDEOH $QGHUVRQ DQG 0RRUH >@ SURYH WKDW 5A DQG /Nf JLYHQ E\ ff ZLOO EH ERXQGHG LI )7Nf *ANff LV XQLIRUPO\ GHWHFWDEOH 6WHS /HW 3N#f DQG 3ef GHQRWH WKH VHTXHQFHV JHQHUDWHG IURP WKH 5'( XVLQJ LQLWLDO FRQGLWLRQV 34 DQG 34 UHVSHFWLYHO\ 'HILQLQJ $3Nf 3Af 3Af LW WKHQ IROORZV WKDW Â3NHf IWHf, 3NHf*f>*7fSHf*Hf Qn9WHff [L3NHf L3Nf*Hf>*7Hf3N Hf*Hf ,@ *7fD3NHff [ , SHf*f>*7fSf*Hf ,@n *7ff7)Hf
PAGE 52
7KLV UHODWLRQVKLS ZDV LQWURGXFHG E\ 6DPVRQ >@ LQ WKH VLQJOHLQSXW FDVH ,W FDQ DOVR EH YHULILHG LQ WKH PXOWLLQSXW FDVH EXW GXH WR WKH ODUJH DPRXQW RI DOJHEUD LQYROYHG WKH GHULYDWLRQ ZLOO QRW EH LQFOXGHG KHUH 6WHS )RU DQ\ 3J ! WKHUH H[LVWV DQ 1H3Jf VXFK WKDW Q3Xf 3 kf+ H IRU DOO HHL8 DQG IRU DOO N ! 1 6LQFH )f*ff LV VWDELOL]DEOH IRU HDFK EHORQJLQJ WR IOJ LW IROORZV IURP ZHOONQRZQ UHVXOWV WKDW 3AHf FRQYHUJHV WR 3AHf SRLQWZLVH RYHU eD ,W ZDV VKRZQ LQ *UHHQ DQG .DPHQ >@ 7KHRUHP f WKDW WKH FRQYHUJHQFH LV XQLIRUP RYHU 4J LI LV FRPSDFW 6WHS )RU DOO 3J EHORQJLQJ WR D FRPSDFW VXEVHW ' RI WKHUH H[LVWV DQ 1Hf VXFK WKDW L30Hf 3 f Q H IRU DOO HHVX Q r! H 7KH FRPSDFWQHVV RI LPSOLHV WKHUH H[LVW PDWULFHV 3r1 DQG 3J VXFK WKDW 3J 3J 3J IRU DOO 3J EHORQJLQJ WR ,W IROORZV IURP WKH UHODWLRQVKLS JLYHQ LQ VWHS WZR WKDW 3N f 3Nf 3NHf IRU DOO 34H' )URP VWHS WKHUH H[LVW LQWHJHUV 1H3Jf DQG 1H3Jf VXFK WKDW LL3Akf 3 fQ H IRU DOO N ! 1 DQG IRU DOO HIL$ . L H Q3NHf 3MHf H IRU DOO N ! 1 DQG IRU DOO 2H68
PAGE 53
7KXV OO3f 3 f H IRU DOO 4H68 IRU DOO 3IWH' . } ( 8 DQG IRU DOO N ! PD[&1A1Jf 6WHS *LYHQ D ! WKHUH H[LVWV D Df VXFK WKDW IRU DOO 3AH' DQG IRU DOO HHe8 H }3Nf fÂ¯ 3NAA ZKHQHYHU QH HQ 7KLV SURSHUW\ IROORZV HDVLO\ IURP WKH IDFW WKDW 3N#f LV D FRQWLQXRXV IXQFWLRQ RI DQG WKH VHWV ' DQG ILr DUH FRPSDFW 6WHS /HW 1 EH D IL[HG ILQLWH LQWHJHU VXFK WKDW WKH FRQGLWLRQ LQ VWHS IRXU LV VDWLVILHG 7KHUH H[LVWV D ILQLWH LQWHJHU 0MHf VXFK WKDW Q5N1 3} IRU DOO N ! ZKHUH 3 LV WKH VROXWLRQ WR WKH $5( ZKHQ WKH PDWULFHV )mf DQG *rf DUH HYDOXDWHG DW HNf ,Q RUGHU WR SURYH WKLV UHVXOW D IDPLO\ RI VHTXHQFHV Se LV GHILQHG DV IRORZV 3N 5N ARU N W 3N 4)7RWff3N)HWff)7Wff3A *Wff>, 67Wff3cM*Wff@fÂ¬ [ *7Wff3A)Wff IRU DOO N ! W
PAGE 54
,Q RWKHU ZRUGV 3N LV HTXDO WR 5N IRU DOO N XS WR WLPH W $IWHU WLPH W Se LV JHQHUDWHG IURP WKH WLPHLQYDULDQW 5LFFDWL GLIIHUHQFH HTXDWLRQ LQLWLDOL]HG E\ 5W DQG XVLQJ WKH FRQVWDQW PDWULFHV )rf DQG *}f HYDOXDWHG DW Wf )RU HYHU\ HWf EHORQJLQJ WR Q WKH VHTXHQFH Se ZLOO EH ERXQGHG IRU DOO N ,Q IDFW 3A FRQYHUJHV WR 3A ZKHQHYHU . RR WfHL8 ,W ZLOO ILUVW EH VKRZQ WKDW Q5N1 3AU H IRU N VXIILFLHQWO\ ODUJH 7KH PDWUL[ SA LV VLPSO\ WKH 1WK VWHS RI D WLPH LQYDULDQW 52( LQLWLDOL]HG DW WLPH N E\ 5N DQG XVLQJ FRQVWDQW PDWULFHV )mf DQG *}f HYDOXDWHG DW Nf 7KH LQWHUSUHWDWLRQ IRU OO5N1 fÂ¯ 3N1 EHLnQJ VQD )RU N VXIILFLHQWO\ ODUJH LV WKDW WKH WLPH YDU\LQJ 5'( JLYHQ E\ f ZLOO QRW YDU\ PXFK IURP WKH WLPHLQYDULDQW 52( RYHU D ILQLWH LQWHUYDO RI WLPH 1 DV ORQJ DV Nf GRHV QRW FKDQJH PXFK RYHU WKH WLPH LQWHUYDO 1 (TXDWLRQ f FDQ EH UHZULWWHQ DV 5NLL 4 )7Nff5NL)HNff )7HNff5NL*HNff ; >, *7Nff5N*Nff@fÂ¯ *7Nff5N)Nff =N 55 IRU DOO N DQG IRU DOO LH>1@ 7KH PDWUL[ ]NL LV JLYHQ E\ (NL )7NLff5NL)NLff 37HNLff5NM *NLff [ >, *7NLff5N*NLff@ *7NLff5N)NLff fÂ¯ )7Nff5N)Nff )7Nff5N*Nff [ >, *7Nff5N*Nff@n *7Nff5N)Nff IRU DOO N DQG IRU DOO LH>1@
PAGE 55
6LQFH 5N EHORQJV WR WKH FRPSDFW VHW ' IRU DOO N LW IROORZV IURP VWHS ILYH WKDW JLYHQ DQ\ D ! WKHUH H[LVWV Df VXFK WKDW QVAQ D ZKHQHYHU XHNLf Nf IRU LH>1@ N N /HW 4N 3N 8VLQJ WKH HTXDWLRQ LQWURGXFHG LQ VWHS WZR JLYHV 4NLO 5NW*Nff>*7H_2!5NL*HN!! Q LHNfff ; 4IFL VNff&*7nNff3NLVHNff ,@n ; *7Nff4_Nf , 5NL*Nff>67Nff5N*Nff ,@ ; *7HNfff7 ,NL 4N LH&21@ ,W IROORZV IURP WKH ERXQGHGQHVV RI HNf 5N DQG SMM IRU W VXIILFLHQWO\ ODUJHf WKDW OO)7HNff, 5NL*4Nff&*7Nff5NL*Nff ,@fÂ¯9Nfff DQG LL*HNff>*7HNff3N? *HNff ,@*7Nff . } 7KXV IRU DOO N VXFK WKDW HNfHLA ZH KDYH 9LO .On&On .*+Ln! n:r 4N r ÂH&1!
PAGE 56
&OHDUO\ WKHUH H[LVWV D FUAHf VXFK WKDW LI Q]N ,, A IRU DOO LH>1@ WKHQ Q4 N N1 H /HW 6RAf EH FKRVHQ VXFK WKDW QVNL+ 2M IRU DOO LH>1@ ZKHQHYHU QHNLf HNf Q ,W IROORZV IURP SURSHUWLHV Yf DQG YLf RI 7KHRUHP WKDW WKHUH H[LVWV D ILQLWH LQWHJHU VXFK WKDW HNfHLJ IRU DOO N ! DQG QHNLf HNfQ DAf IRU DOO LH >1@ DQG IRU DOO N ! 0A 7KHUHIRUH 4N1 fÂ°5N1 fÂ¯ SN1 H IRU N ! n9 8VLQJ WKH UHVXOWV LQ VWHS IRXU ZH REWDLQ D3N1 fÂ¯ SpQ IrU DA N ! 0Lr 7KXV r15N1 fÂ¯ 3p 5N1 fÂ¯ 3N1 3N1 fÂ¯ 3m HA IRU D N ! 0Lr ZKLFK SURYHV WKH UHVXOW
PAGE 57
6WHS 7KHUH H[LVWV D ILQLWH LQWHJHU 0JWHf VXFK WKDW Q3N SNQ H IRU DOO N ! 0 7KH UHVXOW IROORZV LPPHGLDWHO\ IURP WKH XQLIRUP FRQWLQXLW\ RI 3 fÂµf RYHU VLr DQG WKH FRQYHUJHQFH RI +Nf HNfX WR ]HUR e 6WHS 7KHUH H[LVWV D ILQLWH LQWHJHU 0Hf VXFK WKDW Q5N fÂ¬ 5N r1 H IRU D N ! $SSO\LQJ WKH WULDQJOH LQHTXDOLW\ DQG XVLQJ WKH UHVXOWV LQ VWHSV VL[ DQG VHYHQ JLYHV N N N N OO5N1O 5N1% %5N1O 3 r LS R} fÂ« 33r 5N1% H IRU DOO N ! PD[ 0A 0f 7KH UHVXOW WKHQ IROORZV HDVLO\ E\ VHWWLQJ 0 PD[0A 0f 1 6WHS ,I /Nf LV FRPSXWHG XVLQJ ff WKHQ Q/Nf /NOfQ FRQYHUJH WR ]HUR DV N JRHV WR LQILQLW\ 7KLV UHVXOW IROORZV HDVLO\ VLQFH Q5A 5A ALO Q)Nf )NOfOO DQG Q*Nf *Nf DOO FRQYHUJH WR ]HUR 8VLQJ WKH UHVXOWV RI 7KHRUHP LW LV QRZ SRVVLEOH WR SURYH WKDW )Nf*Nf/Nff LV H[SRQHQWLDOO\ VWDEOH DV D WLPHYDU\LQJ V\VWHP PDWUL[ 7KH H[SRQHQWLDO VWDELOLW\ RI )Nf*Nf/Nff LV YHU\ LPSRUWDQW IRU SURYLQJ JOREDO VWDELOLW\ RI WKH DGDSWLYH FORVHGORRS V\VWHP
PAGE 58
7KHRUHP 6XSSRVH HNf LV JHQHUDWHG XVLQJ WKH SDUDPHWHU HVWLPDWRU ff DQG WKH IHHGEDFN JDLQ VHTXHQFH /Nf LV FRPSXWHG XVLQJ f f 7KHQ VXEMHFW WR DVVXPSWLRQV $$ )Nf*Nf/Nff LV H[SRQHQWLDOO\ VWDEOH DV D WLPHYDU\LQJ V\VWHP PDWUL[ 7 7 3URRI $VVXPSWLRQ $ LPSOLHV WKDW ) #f* Hff LV GHWHFWDEOH IRU HDFK EHORQJLQJ WR Q 8VLQJ DUJXPHQWV VLPLODU WR WKRVH LQ 3URSRVLWLRQ DQG 7KHRUHP LW IROORZV WKDW )ANf*7Nff LV XQLIRUPO\ GHWHFWDEOH ,Q >@ $QGHUVRQ DQG 0RRUH SURYH WKDW LI )ANf*ANff LV XQLIRUPO\ GHWHFWDEOH WKHQ )Nf*Nf/Nf fA LV H[SRQHQWLDOO\ VWDEOH DV D WLPHYDU\LQJ V\VWHP PDWUL[ ZKHUH /Nf LV JLYHQ E\ ff 6LQFH PDWUL[ SURGXFWV GR QRW FRPPXWH LW GRHV QRW LPPHGLDWHO\ IROORZ WKDW )Nf*Nf/Nff LV H[SRQHQWLDOO\ VWDEOH +RZHYHU 6DPVRQ DQG )XFKV >@ VKRZ $Nf ZLOO EH H[SRQHQWLDOO\ VWDEOH DV D WLPHYDU\LQJ V\VWHP PDWUL[ LI $7Nf LV H[SRQHQWLDOO\ VWDEOH DQG LI LL$Nf$Nf Q FRQYHUJHV WR ]HUR DV N Â« } 6LQFH )Nf*Nf/NffA LV H[SRQHQWLDOO\ VWDEOH LW WKHQ IROORZV WKDW )Nf*Nf/Nff ZLOO DOVR EH H[SRQHQWLDOO\ VWDEOH LI X)Nf*Nf/Nff )NOf*NOf/NOffQ FRQYHUJHV WR ]HUR DV N Â« } 3URSHUW\ YLf RI 7KHRUHP LPSOLHV ERWK LL)Nf)Nf Q DQG Q*Nf*NOf Q FRQYHUJH WR ]HUR ,W ZDV VKRZQ LQ 7KHRUHP WKDW Q/Nf/NOfQ DOVR FRQYHUJHV WR ]HUR 6LQFH )Nf*Nf/Nf)NOf*NOfONOff )Nf)NOff *Nf/Nf/NOff *Nf*NOff/NOf DQG ERWK *Nf DQG /Nf DUH ERXQGHG LW IROORZV WKDW Q)Nf*Nf/Nff)NOf*NOf/NOffQ GRHV LQGHHG FRQYHUJH WR ]HUR 7KXV WKH PDWUL[ )Nf*Nf/Nff LV H[SRQHQWLDOO\ VWDEOH DV D WLPHYDU\LQJ V\VWHP PDWUL[
PAGE 59
6WDELOLW\ RI WKH $GDSWLYH &ORVHG/RRS 6\VWHP 7KH FRQWURO ODZ FKRVHQ WR UHJXODWH WKH V\VWHP LV JLYHQ E\ XNf /Nf[Nf f ZKHUH [Nf LV JHQHUDWHG IURP WKH DGDSWLYH REVHUYHU f DQG /.f LV D VWDELOL]LQJ IHHGEDFN IRU WKH HVWLPDWHG V\VWHP )Nf*Nff 7KH IROORZLQJ WKHRUHP VKRZV WKDW WKH DGDSWLYH UHJXODWRU FRQVLVWLQJ RI WKH REVHUYHU f DQG WKH FRQWURO ODZ f UHVXOWV LQ D JOREDOO\ VWDEOH FORVHGORRS V\VWHP WKDW LV IRU DQ\ LQLWLDO VWDWHV LQ WKH SODQW DQG WKH REVHUYHU WKH LQSXW XNf DQG WKH RXWSXW \Nf FRQYHUJH WR ]HUR 7KHRUHP 6XSSRVH WKDW WKH SDUDPHWHU HVWLPDWRU ff LV XVHG VR WKDW WKHUH LV D VWDELOL]LQJ IHHGEDFN /Nf IRU WKH HVWLPDWHG V\VWHP )Nf*Nff VXEMHFW WR DVVXPSWLRQV $$ 7KHQ ZLWK WKH DGDSWLYH UHJXODWRU GHILQHG E\ f DQG f WKH UHVXOWLQJ FORVHGORRS V\VWHP LV JOREDOO\ VWDEOH 3URRI 7KH SURRI LV EDVHG RQ D 0,02 H[WHQVLRQ RI WKH UHVXOWV LQ >@ /HWWLQJ eNf \Nf\Nf DQG XVLQJ HTXDWLRQV f DQG f JLYHV [NOf )Nf*Nf/Nff[Nf 0NfHNf \Nf +[Nf HNf
PAGE 60
$OVR WKH UHJUHVVLRQ YHFWRU ^!Nf GHILQHG LQ f FDQ EH ZULWWHQ LQ WKH IRUP ! Nf 6M!Nf 'Nf[Nf 9Nf ZKHUH R + &' \Q' 'Nf /Nf YNf ,SQOf fÂµ AIO2 'HILQLQJ ]NOf [NOf JLYHV ZKHUH $Nf 6 ]NOf $Nf]Nf ZNf f 'Nf )Nf*Nf/Nf ZNf YNf 0NfHNf 7KH PDWUL[ 6 LV D VWDEOH PDWUL[ VLQFH LW LV ORZHUEORFN WULDQJXODU 7KH PDWUL[ )Nf*Nf/Nf LV VWDEOH E\ FRQVWUXFWLRQ WKHUHIRUH $Nf LQ f LV H[SRQHQWLDOO\ VWDEOH DV D WLPHYDU\LQJ V\VWHP PDWUL[ ,Q DGGLWLRQ LW IROORZV HDVLO\ IURP f DQG WKH ERXQGHGQHVV RI 0Nf WKDW LLZNfQ DNfQ]NfQ ENf f ZKHUH DNf DQG ENf DUH SRVLWLYH VFDODU VHTXHQFHV ZKLFK FRQYHUJH WR ]HUR DV N fÂµ}Â‘ } 7KXV f FDQ EH YLHZHG DV DQ H[SRQHQWLDOO\ VWDEOH WLPH YDU\LQJ V\VWHP GULYHQ E\ DQ LQSXW ZNf ZKLFK FDQ JURZ QR IDVWHU WKDQ OLQHDUO\ ZLWK WKH VWDWH ]Nf )ROORZLQJ WKH SURRI LQ >@ LW ZLOO EH VKRZQ WKDW LL]Nf W FRQYHUJHV WR ]HUR /HW $N1Nf GHQRWH WKH VWDWH WUDQVLWLRQ PDWUL[ IRU f GHILQHG E\
PAGE 61
0O $N0Nf Q $NLf L n 6LQFH $Nf LV DQ H[SRQHQWLDOO\ VWDEOH WLPHYDU\LQJ V\VWHP PDWUL[ LW IROORZV WKDW Q$N0Nf, 5 } IRU DOO N0 ! f $OVR JLYHQ WKHUH H[LVWV D ILQLWH S VXFK WKDW Q$NSNfLL IRU DOO N f )RU WKH UHPDLQGHU RI WKH SURRI S ZLOO EH D IL[HG FRQVWDQW VXFK WKDW f LV VDWLVILHG 7DNLQJ WKH QRUP RI ERWK VLGHV RI f DQG XVLQJ f JLYHV Q]NOfQ }$Nf LLQ]Nf Q QZNfLL f K $ Nf G LNff Q]Nf LL ENf 1RWLQJ WKDW Q$NfQ DNf DQG ENf DUH DOO ERXQGHG DQG S LV IL[HG LW IROORZV IURP HTXDWLRQ f WKDW WKHUH H[LVW ILQLWH FRQVWDQWV 5 DQG 5J VXFK WKDW ]NLf 5AW]Nf 5J IRU DOO LH >S@ DQG IRU DOO N f 6ROYLQJ HTXDWLRQ f UHFXUVLYHO\ JLYHV
PAGE 62
3 ]NSf $NSNf]Nf $NSNMOfZNMf M 7DNLQJ WKH QRUP RI ERWK VLGHV DQG XVLQJ f UHVXOWV LQ 3 Q]NSfQ LL$NSNfQ Q]Nf Q e Q$NSNMOf Q QZNMfQ M 3fÂ¯ Q$NSNfQ WL]Nf LW O Q$NSNMOf QDNMf Q]NMf ENMff M ,W IROORZV IURP HTXDWLRQV f f DQG f WKDW S 3 B ,]NSf LL LL]Nf LL PD[O5M O DNMf5LL]Nf ,, 5af ENMf IRU M Â Â£ DOO N f 6LQFH ,Nf FRQYHUJHV WR ]HUR WKHUH H[LVWV D ILQLWH 1 VXFK WKDW DNf S5JGLD[LO 5Af IRU DOO N ! 1 7KHUHIRUH IRU DOO N ! 1 S 3 B L]NSfQ UQ]NfLL PD[O5c_f O DNMf5 ENMf RU HTXLYDOHQWO\ M L]NSfLL UQ]NfQ FNf f ZKHUH U DQG FNf FRQYHUJHV WR ]HUR DV N Â« m ,Q RUGHU WR VKRZ f LPSOLHV X]NfQ FRQYHUJHV WR ]HUR OHW H! EH JLYHQ 6ROYLQJ f UHFXUVLYHO\ DQG XVLQJ f JLYHV
PAGE 63
Q]NLQSf LL UQQ]NLfLL O UQfÂ¯rfÂ¯AFNLMSf M Q Q]NLQSf LL UQ5Q]Nf LL 5f e UQ FNLMSf Â G M IRU DOO N ! 1 LH>S@ f 6LQFH FNf FRQYHUJHV WR ]HUR WKHUH H[LVWV D ILQLWH ! 1 VXFK WKDW FNf IfÂ§ IRU DOO N ! 1 $SSO\LQJ WKLV LQHTXDOLW\ WR f JLYHV +]1ALQSf OO UQ5+]1Af ,W 5Jf H LH>S@ $OVR VLQFH U WKHUH H[LVWV D ILQLWH 1 VXFK WKDW U ,5K]1AfLL 5f IRU D Q ! 9 7KXV +]1ALQSf H IRU DOO Q ! 1 LH>S@ ZKLFK LPSOLHV Q]Nf H IRU DOO N ! 1S %\ WKH GHILQLWLRQ RI ]Nf ERWK WKH V\VWHP LQSXW XNf /Nf[Nf DQG WKH V\VWHP RXWSXW \Nf PXVW FRQYHUJH WR ]HUR 7KHUHIRUH WKH DGDSWLYH
PAGE 64
FORVHGORRS V\VWHP LV JOREDOO\ VWDEOH XVLQJ WKH DGDSWLYH UHJXODWRU GHILQHG E\ f DQG f
PAGE 65
&+$37(5 9, $33/,&$7,21 72 75$&.,1* $1' ',6785%$1&( 5((&7,21 7KH DGDSWLYH FRQWUROOHU GHULYHG LQ &KDSWHU 9 FDQ EH DSSOLHG WR WKH SUREOHP RI WUDFNLQJ ZLWK GLVWXUEDQFH UHMHFWLRQ XVLQJ WKH LQWHUQDO PRGHO SULQFLSOH 7KH LQWHUQDO PRGHO SULQFLSOH KDV EHHQ GLVFXVVHG E\ D QXPEHU RI LQGLYLGXDOV VHH IRU H[DPSOH >@ &RQVLGHU WKH ULQSXW PRXWSXW SODQW GHVFULEHG E\ ZKHUH ) [NOf )[Nf *XNf 'YNf \Nf +[Nf f fÂµ fÂ§ n$L , % fÂ¯$" , fÂµ fÂµ E fÂµ fÂµ fÂµ * fÂµ fÂµ fÂµ 9L , fÂµ %Q $Q %Q + >, @ 7KH YHFWRU YNf FRQVLVWV RI WKH H[RJHQRXV GLVWXUEDQFH VLJQDOV $V LQ SUHYLRXV FKDSWHUV LW LV DVVXPHG WKDW VRPH RU DOO RI WKH HQWULHV LQ WKH PDWULFHV $M DQG %M DUH XQNQRZQ DQG ZLOO EH HVWLPDWHG XVLQJ WKH SDUDPHWHU HVWLPDWRU GHVFULEHG E\ ff /HWWLQJ UNf GHQRWH DQ PYHFWRU UHIHUHQFH VLJQDO WKH REMHFWLYH LV WR GHVLJQ DQ DGDSWLYH FRQWUROOHU ZKLFK VWDELOL]HV WKH V\VWHP GHVFULEHG E\ f DQG IRUFHV WKH WUDFNLQJ HUURU HNf \NfUNf WR FRQYHUJH WR ]HUR DV N !fÂµ }
PAGE 66
$V PHQWLRQHG WKH FRQWUROOHU GHVLJQ ZLOO EH EDVHG RQ WKH LQWHUQDO PRGHO SULQFLSOH $VVXPH WKDW WKH GLVWXUEDQFH YHFWRU YNf DQG WKH UHIHUHQFH VLJQDO UNf ERWK VDWLVI\ WKH GLIIHUHQFH HTXDWLRQ ]NTf r L=NTOf mANOf DA]Nf f /HW U EH GHILQHG DV fÂ¯DR aDO fÂµfÂµfÂµ aDTO 7KH LQWHUQDO PRGHO FDQ WKHQ EH UHDOL]HG DV [FNOf $F[FNf %FHNf ZKHUH $F EORFN GLDJRQDO ^U U U` PWXSOH fÂµ %F EORFN GLDJRQDO ^7 7 7` ZLWK W PWXSOH f 7KH IROORZLQJ WKHRUHP JLYHV WKH FRQGLWLRQV XQGHU ZKLFK LW LV SRVVLEOH WR GHVLJQ DQ DGDSWLYH FRQWUROOHU ZKLFK ZLOO VWDELOL]H WKH SODQW GHVFULEHG
PAGE 67
E\ f DQG ZLOO GULYH WKH WUDFNLQJ HUURU HNf WR ]HUR DV\PSWRWLFDOO\ 7KHRUHP 6XSSRVH DVVXPSWLRQV $$ DUH VDWLVILHG DQG LQ DGGLWLRQ UDQN ],)Hf *Hf %&+ ],$F QPT IRU DOO kHIW DQG IRU DOO ] _]_ ! ZKHUH Q ^H II7[P H >RnQA[@ /HW [ Nf EH JHQHUDWHG IURP WKH LM LM LM HUURU GULYHQ V\VWHP GHVFULEHG LQ f DQG [Nf EH JHQHUDWHG IURP WKH DGDSWLYH REVHUYHU JLYHQ E\ f 7KHQ WKHUH H[LVWV D FRQWURO ODZ XNf /MeNf/;FNf ZKLFK ZKHQ DSSOLHG WR WKH SODQW f UHVXOWV LQ D VWDEOH FORVHGORRS V\VWHP DQG GULYHV WKH WUDFNLQJ HUURU HNf \Nf UNf WR ]HUR 3URRI 8VLQJ HTXDWLRQV f DQG f DQG VHWWLQJ YNf DQG UNf HTXDO WR ]HUR ZH REWDLQ [NOf [FNOf )Nf %&+ [Nf [FNf *Nf XNf f ZKHUH HNf +[Nf \Nf $VVXPLQJ WKH JLYHQ UDQN FRQGLWLRQ KROGV LW IROORZV IURP SURSHUW\ Yf RI 7KHRUHP WKDW WKH V\VWHP GHVFULEHG E\ f LV VWDELOL]DEOH 7KH VWDELOL]LQJ FRQWURO ODZ LV RI WKH IRUP XNf /ANANf /[FNf f $ )Nf $ *nNI /HW )Nf B%&+ $FB DQG *Nf $V LQ WKH FDVH RI WKH DGDSWLYH UHJXODWRU GfÃ¼
PAGE 68
VWDELOL]LQJ IHHGEDFN JDLQ /Nf >/ANf /"Nf@ FDQ EH FRPSXWHG IURP RQH VWHS RI WKH 5LFFDWL GLIIHUHQFH HTXDWLRQ /Nf >*7Nf5N*Nf ,@ *7Nf5N)Nf f 5N 4 /7Nf/Nf )NfNf/Nff75N)Nf*Nf/Nff $JDLQ 54 DQG 4 DUH DUELWUDU\ SRVLWLYH GHILQLWH V\PPHWULF PDWULFHV ,Q RUGHU WR XVH WKH H[LVWLQJ UHVXOWV RQ WUDFNLQJ DQG GLVWXUEDQFH UHMHFWLRQ WKH FORVHGORRS V\VWHP FRQVLVWLQJ RI f DQG WKH FRQWURO ODZ f PXVW FRQYHUJH WR D WLPHLQYDULDQW V\VWHP 3URSHUWLHV Yf DQG YLLf RI 7KHRUHP LPSO\ WKDW WKH SDLU )Nf*Nff FRQYHUJHV WR D VWDELOL]DEOH SDLU ) * f 8VLQJ WKH UHVXOWV RI 6DPVRQ &/2@ /Nf FRQYHUJHV WR / WKH RSWLPDO /4 IHHGEDFN IRU WKH SDLU ) * f JLYHQ E\ p / >*75 * ,@ *7 5 ) &2 ZKHUH WKH PDWUL[ 5fÂ± LV WKH VROXWLRQ WR WKH $5( IRU ) * f 0RUH p SUHFLVHO\ 5 4 /7/ ) * / f75 ) * / f $SSO\LQJ WKH FRQWURO ODZ f WR WKH V\VWHP GHVFULEHG E\ f UHVXOWV LQ DQ H[SRQHQWLDOO\ VWDEOH V\VWHP GULYHQ E\ HNf )URP WKH UHVXOWV LQ &KDSWHU )LYH WKH REVHUYHU HUURU HNf FRQYHUJHV WR ]HUR $OVR LW IROORZV IURP SURSHUW\ LLf RI 7KHRUHP WKDW 0Nf LV ERXQGHG 7KXV WKH FORVHGORRS V\VWHP FRQVLVWLQJ RI f DQG WKH
PAGE 69
FRQWURO ODZ XNf /M;Nf/[FNf LV H[SRQHQWLDOO\ VWDEOH DQG FRQYHUJHV WR D WLPHLQYDULDQW V\VWHP )LQDOO\ IURP WKH UHVXOWV LQ &KHQ >@ LI WKH FRQWURO ODZ f LV DSSOLHG WR WKH SODQW f WKH UHVXOWLQJ FORVHGORRS V\VWHP LV VWLOO VWDEOH QRW LQFOXGLQJ WKH H[RJHQRXV GLVWXUEDQFH YNff DQG WKH V\VWHP RXWSXW \Nf FRQYHUJHV WR WKH UHIHUHQFH VLJQDO UNf DV N rÂ‘ m 7KH DGDSWLYH FRQWUROOHU GHVFULEHG E\ f DQG ff FDXVHV WKH SODQW f WR UHMHFW GHWHUPLQLVWLF GLVWXUEDQFHV DQG WR WUDFN D JLYHQ UHIHUHQFH VLJQDO UNf 6LPXODWLRQV RI WKH DGDSWLYH FRQWUROOHU IRU YDULRXV W\SHV RI H[WHUQDO GLVWXUEDQFH DQG UHIHUHQFH VLJQDOV DUH LQFOXGHG LQ &KDSWHU 9,,
PAGE 70
&+$37(5 9,, 6,08/$7,216 7KLV FKDSWHU FRQWDLQV VHYHUDO VLPXODWLRQV RI ERWK WKH DGDSWLYH UHJXODWRU GHULYHG LQ &KDSWHU 9 DQG WKH DGDSWLYH FRQWUROOHU GLVFXVVHG LQ &KDSWHU 6L[ DSSOLHG WR 6,62 DV ZHOO DV 0,02 GLVFUHWHWLPH V\VWHPV ([DPSOH &RQVLGHU WKH QRQPLQLPXP SKDVH GLVFUHWHWLPH V\VWHP GHVFULEHG E\ WKH IROORZLQJ GLIIHUHQFH HTXDWLRQ \Nf \NOf\NfXNOfXNf 7KH DGDSWLYH UHJXODWRU GHILQHG E\ WKH REVHUYHU f DQG WKH FRQWURO ODZ f ZLWK /Nf FRPSXWHG XVLQJ ff ZDV LPSOHPHQWHG 7KUHH GLIIHUHQW DOJRULWKPV ZHUH XVHG WR HVWLPDWH WKH V\VWHP SDUDPHWHUV WKH HVWLPDWLRQ VFKHPH GHVFULEHG E\ ff WKH UHFXUVLYH OHDVWVTXDUHV DOJRULWKP GHILQHG E\ f DQG WKH HVWLPDWLRQ VFKHPH JLYHQ E\ f f ZLWK GDWD QRUPDOL]DWLRQ LH PD[O +M!Nf Qf IRU DOO Nf ,Q DOO WKUHH FDVHV WKH LQLWLDO FRYDULDQFH 3f ZDV FKRVHQ WR EH , DQG WKH LQLWLDO VWDWH RI WKH SODQW ZDV > @7 $OVR 4 DQG 54 LQ f ZHUH FKRVHQ WR EH , 7KH SDUDPHWHU UDQJHV LQLWLDO SDUDPHWHU HVWLPDWHV DQG VWHDG\ VWDWH HVWLPDWHV IRU DOO WKUHH HVWLPDWLRQ VFKHPHV DUH GLVSOD\HG LQ 7DEOH $V VHHQ IURP 7DEOH WKH HVWLPDWLRQ VFKHPH GHVFULEHG E\ ff IRUFHV WKH HVWLPDWHV RI WKH SDUDPHWHUV WR FRQYHUJH WR WKH JLYHQ UDQJHV ZKHUHDV UHFXUVLYH OHDVW VTXDUHV GRHV
PAGE 71
QRW )LJXUHV FRQWDLQ SORWV RI WKH RXWSXW UHVSRQVH FRQWURO LQSXW DQG HVWLPDWHG SDUDPHWHUV IRU HDFK RI WKH WKUHH FDVHV $ FRPSDULVRQ RI )LJXUHV DQG VKRZ WKH HVWLPDWLRQ DOJRULWKP GHILQHG E\ ff KDV D VOLJKWO\ EHWWHU WUDQVLHQW UHVSRQVH WKDQ OHDVW VTXDUHV )LJXUH LOOXVWUDWHV WKH GHWULPHQWDO HIIHFW RI WKH GDWD QRUPDOL]DWLRQ RQ WKH WUDQVLHQW UHVSRQVH RI WKH V\VWHP $V SUHYLRXVO\ PHQWLRQHG H LQ f VKRXOG EH FKRVHQ YHU\ VPDOO WR PLQLPL]H WKH HIIHFW RI GDWD QRUPDOL]DWLRQ RQ WKH V\VWHP WUDQVLHQW UHVSRQVH 7DEOH '$7$ DO aD EO E $FWXDO 3DUDPHWHUV ,QLWLDO (VWLPDWHV 3DUDPHWHU 5DQJHV > @ > @ > @ > @ (VWLPDWLRQ 6FKHPH ff 5HFXUVLYH /HDVW 6TXDUHV (VWLPDWLRQ 6FKHPH ff ZLWK PD[OQHNfQf
PAGE 72
\Nf XNf
PAGE 73
)LJXUH FRQWLQXHG (VWLPDWLRQ 6FKHPH ff
PAGE 74
\Nf 8Nf L $ n9 Ef Âœ4 JJfÂ§ IW 97 )LJXUH 5HFXUVLYH /HDVW 6TXDUHV
PAGE 75
)LJXUH FRQWLQXHG 5HFXUVLYH /HDVW 6TXDUHV
PAGE 76
\Nf ORn VW WLW QL9n Nf Df )LJXUH (VWLPDWLRQ 6FKHPH ff ZLWK QN PD[ __I!Nf__ f
PAGE 77
([DPSOH &RQVLGHU WKH QRQPLQLPXP SKDVH VWDELOL]DEOH EXW QRW UHDFKDEOH GLVFUHWHWLPH V\VWHP GHVFULEHG E\ \Nf \NOf\Nf\NfXNOfXNfXNf 7KH DGDSWLYH FRQWUROOHU GHILQHG E\ WKH DGDSWLYH REVHUYHU f DQG WKH FRQWURO ODZ f ZLWK /Nf FRPSXWHG IURP f ZDV LPSOHPHQWHG WR IRUFH WKH RXWSXW WR WUDFN WKH UHIHUHQFH VLJQDO UNf 1R GLVWXUEDQFH ZDV LQWURGXFHG LQ WKLV H[DPSOH 7KH LQWHUQDO PRGHO ZDV JLYHQ E\ [FNOf [FNf HNf ZLWK HNf \Nf 7KH SDUDPHWHUV ZHUH HVWLPDWHG XVLQJ ff 7KH LQLWLDO FRQGLWLRQV ZHUH FKRVHQ WR EH 3f , [f > @7 [Ff DQG 5 , 7KH LQLWLDO SDUDPHWHU HVWLPDWHV SDUDPHWHU UDQJHV DQG VWHDG\ VWDWH YDOXHV RI WKH HVWLPDWHV DUH GLVSOD\HG LQ 7DEOH 3ORWV RI WKH RXWSXW UHVSRQVH FRQWURO LQSXW DQG SDUDPHWHU HVWLPDWHV DUH VKRZQ LQ )LJXUH (TXDWLRQ f LQGLFDWHV WKDW WKH LQLWLDO FRYDULDQFH PDWUL[ 3f VKRXOG VDWLVI\ 3f, :KHQ WKLV H[DPSOH ZDV UXQ DJDLQ ZLWK 3f , WKH V\VWHP UDSLGO\ ZHQW XQVWDEOH $W WLPH N WKH RXWSXW ZDV \Nf [Op DQG WKH FRQWURO LQSXW ZDV XNf [OA 7KH SDUDPHWHU HVWLPDWHV DW N ZHUH DNf DNf DNf EMNf ENf DQG ENf 7KXV FKRRVLQJ 3f VPDOOHU WKDQ LV QRW PHUHO\ D WHFKQLFDOLW\ LQ WKH SURRI RI 7KHRUHP EXW LV D QHFHVVDU\ FRQGLWLRQ IRU SUHVHUYLQJ VWDELOLW\ RI WKH DGDSWLYH FORVHGORRS V\VWHP
PAGE 78
7DEOH '$7$ DO D D EO E E $FWXDO 3DUDPHWHUV ,QLWLDO (VWLPDWHV fÂµ 3DUDPHWHU 5DQJHV >@ >@ >@ >@ >@ >@ (VWLPDWLRQ 6FKHPH fÂ§ffÂ§fÂ§f L
PAGE 79
\Nf 8Nf
PAGE 80
)LJXUH FRQWLQXHG (VWLPDWLRQ 6FKHPH ff
PAGE 81
([DPSOH &RQVLGHU WKH GLVFUHWHWLPH V\VWHP GHVFULEHG E\ \Nf \NOf\NfXNOfXNf 7KH DGDSWLYH FRQWUROOHU GHILQHG E\ WKH REVHUYHU f DQG WKH FRQWURO ODZ f ZLWK /Nf FRPSXWHG IURP f ZDV LPSOHPHQWHG WR IRUFH WKH RXWSXW WR WUDFN WKH UHIHUHQFH VLJQDO UNf $ VWHS GLVWXUEDQFH GHILQHG E\ 2I RI . . ! ZDV LQWURGXFHG 7KH LQWHUQDO PRGHO ZDV JLYHQ E\ [FNOf [FNf HNf ZLWK HNf \Nf 7KH SDUDPHWHUV ZHUH HVWLPDWHG XVLQJ WKH DOJRULWKP GHILQHG E\ ff 7KH LQLWLDO FRQGLWLRQV ZHUH FKRVHQ WR EH 3f , [f >O RI [Ff 5 , 7KH LQLWLDO SDUDPHWHU HVWLPDWHV SDUDPHWHU UDQJHV DQG VWHDG\ VWDWH HVWLPDWHG YDOXHV DUH GLVSOD\HG LQ 7DEOH 7KH RXWSXW UHVSRQVH FRQWURO LQSXW DQG SDUDPHWHU HVWLPDWHV DUH VKRZQ LQ )LJXUH )RU FRPSDULVRQ SXUSRVHV SORWV RI WKH RXWSXW UHVSRQVHV DQG FRQWURO LQSXW ZKHQ WKH SDUDPHWHUV DUH NQRZQ H[DFWO\ DUH VKRZQ LQ )LJXUH $V H[SHFWHG WKH WUDQVLHQW UHVSRQVH LV FRQVLGHUDEO\ ZRUVH ZKHQ WKH V\VWHP SDUDPHWHUV PXVW EH HVWLPDWHG +RZHYHU RQFH WKH V\VWHP LV LQ VWHDG\ VWDWH WKH DGDSWLYH V\VWHP VHHPV WR UHVSRQG HTXDOO\ ZHOO WR D VXGGHQ VWHS GLVWXUEDQFH DV WKH V\VWHP XVLQJ H[DFW SDUDPHWHUV
PAGE 82
7DEOH '$7$ DO D EO E $FWXDO 3DUDPHWHUV ,QLWLDO (VWLPDWHV n 3DUDPHWHU 5DQJHV > @ ,fÂ§ } &2 > @ > LR@ (VWLPDWLRQ 6FKHPH fÂ§ffÂ§fÂ§f
PAGE 83
\IO2 )LJXUH (VWLPDWLRQ 6FKHPH ff
PAGE 84
L fÂ§8 ! fÂµ m fÂ§ ER WRR QR Gf We^Nf EMNf Z )LJXUH FRQWLQXHG (VWLPDWLRQ 6FKHPH ff
PAGE 85
\IO2 XNf )LJXUH ([DFW 3DUDPHWHUV
PAGE 86
([DPSOH &RQVLGHU WKH QRQPLQLPXP SKDVH GLVFUHWHWLPH V\VWHP GHVFULEHG E\ \Nf \NOf\NfXNOfXNf 7KH DGDSWLYH FRQWUROOHU GHILQHG E\ WKH REVHUYHU f DQG WKH FRQWURO ODZ f ZLWK /Nf FRPSXWHG XVLQJ f ZDV LPSOHPHQWHG WR IRUFH WKH V\VWHP RXWSXW WR WUDFN WKH UHIHUHQFH VLJQDO UNf DQG UHMHFW WKH VLQXVRLGDO GLVWXUEDQFH YNf VLQ LUNf 7KH LQWHUQDO PRGHO ZDV JLYHQ E\ [FN' R Rn R [H OfÂ§r HNf ZLWK HNf \Nf 7KH V\VWHP SDUDPHWHUV ZHUH HVWLPDWHG XVLQJ WKH DOJRULWKP ff 7KH LQLWLDO FRQGLWLRQV ZHUH FKRVHQ WR EH 3f , [f > 2I [Ff > @A DQG 5 , 7KH LQLWLDO SDUDPHWHU HVWLPDWHV SDUDPHWHU UDQJHV DQG VWHDG\ VWDWH HVWLPDWHG YDOXHV DUH GLVSOD\HG LQ 7DEOH )LJXUH LQFOXGHV SORWV RI WKH V\VWHP RXWSXW UHVSRQVH FRQWURO LQSXW DQG HVWLPDWHG V\VWHP SDUDPHWHUV )RU FRPSDULVRQ SXUSRVHV ERWK WKH V\VWHP RXWSXW DQG FRQWURO LQSXW DUH VKRZQ LQ )LJXUH IRU WKH FDVH ZKHQ WKH SDUDPHWHUV DUH NQRZQ H[DFWO\
PAGE 87
7DEOH '$7$ DO D EO E $FWXDO 3DUDPHWHUV ,QLWLDO (VWLPDWHV 3DUDPHWHU 5DQJHV > @ fÂ§ R fÂµ r R fÂµ &0 > @ > @ (VWLPDWLRQ 6FKHPH ff
PAGE 88
\Nf , , nL , ,W ? O L Nf Df )LJXUH (VWLPDWLRQ 6FKHPH ff
PAGE 89
)LJXUH FRQWLQXHG (VWLPDWLRQ 6FKHPH ff
PAGE 90
\Nf n n n n 722 72 Df XNf )LJXUH ([DFW 3DUDPHWHUV
PAGE 91
([DPSOH &RQVLGHU WKH VLQJOH LQSXW WZR RXWSXW GLVFUHWHWLPH V\VWHP GHVFULEHG E\ \ANf O D r Nfn n r \Nf O r Nf E XNOf ZKHUH D DQG E DUH XQNQRZQ SDUDPHWHUV $V GLVFXVVHG LQ &KDSWHU 7KUHH WKH UHTXLUHG FRPSXWDWLRQV IRU SDUDPHWHU HVWLPDWLRQ FDQ EH UHGXFHG E\ VHSDUDWLQJ WKH NQRZQ SDUDPHWHUV IURP WKH XQNQRZQ SDUDPHWHUV 7KLV LV DFFRPSOLVKHG E\ UHZULWLQJ WKH V\VWHP HTXDWLRQV DV \Nf H7!DNOf LS7! Nf ZLWK 7 D n ADNLf U+ &0 /!! E XNOf W AENOf \N'n \NLf XNOf 7KH DGDSWLYH UHJXODWRU GHILQHG E\ WKH REVHUYHU f DQG WKH FRQWURO ODZ f ZLWK /Nf FRPSXWHG XVLQJ ff ZDV LPSOHPHQWHG 7KH V\VWHP SDUDPHWHU PDWUL[ ZDV HVWLPDWHG XVLQJ WKH DOJRULWKP ff 7KH LQLWLDO FRQGLWLRQV ZHUH FKRVHQ WR EH 3f , [f >O @7 DQG 54 , 7KH LQLWLDO HVWLPDWHV SDUDPHWHU UDQJHV DQG VWHDG\ VWDWH HVWLPDWHV DUH JLYHQ LQ 7DEOH 3ORWV RI WKH RXWSXW UHVSRQVH FRQWURO LQSXW DQG SDUDPHWHU HVWLPDWHV DUH VKRZQ LQ )LJXUH
PAGE 92
7DEOH '$7$ D E $FWXDO 3DUDPHWHUV ,QLWLDO (VWLPDWHV 3DUDPHWHU 5DQJHV > @ fÂ§ R fÂµ &0 fÂµY R fÂµ , , (VWLPDWLRQ 6FKHPH ff
PAGE 93
\[Nf \Nf
PAGE 94
XNf )LJXUH FRQWLQXHG (VWLPDWLRQ 6FKHPH ff
PAGE 95
([DPSOH &RQVLGHU WKH PXOWLLQSXW PXOWLRXWSXW GLVFUHWHWLPH V\VWHP GHVFULEHG E\ 9ONfn nO D r Nfr E X[ Nf \Nf Nf EL E B X Nf 7KH RUGHU RI UHTXLUHG FRPSXWDWLRQV IRU SDUDPHWHU HVWLPDWLRQ LV UHGXFHG E\ UHZULWLQJ WKH V\VWHP LQ WKH IRUP \Nf } H7I!DNOf L_}7}E Nf ZLWK 7 ! f _!D7NOf >\NOf XNOf XNOf@ !E7Nf &\[NOf \Nf XANf@ 7KH DGDSWLYH UHJXODWRU GHILQHG E\ WKH REVHUYHU f DQG WKH FRQWURO ODZ f ZLWK /Nf FRPSXWHG XVLQJ ff ZDV DSSOLHG WR WKH V\VWHP 7KH DOJRULWKP JLYHQ E\ ff ZDV XVHG WR HVWLPDWH WKH V\VWHP SDUDPHWHU PDWUL[ H 7KH LQLWLDO FRQGLWLRQV IRU WKH VLPXODWLRQ ZHUH 3f , [f >O @A DQG 54 , 7KH LQLWLDO SDUDPHWHU HVWLPDWHV SDUDPHWHU UDQJHV DQG VWHDG\ VWDWH HVWLPDWHV DUH GLVSOD\HG LQ 7DEOH )LJXUH FRQWDLQV SORWV RI WKH RXWSXW UHVSRQVH FRQWURO LQSXW DQG SDUDPHWHU HVWLPDWHV
PAGE 96
7DEOH '$7$ D E E E $FWXDO 3DUDPHWHUV ,QLWLDO (VWLPDWHV 3DUDPHWHU 5DQJHV ,fÂ§ fÂµ R $ LQ fÂµ R L L > @ > @ > f (VWLPDWLRQ 6FKHPH fÂ§ffÂ§fÂ§f
PAGE 97
\[Nf \ )LJXUH (VWLPDWLRQ 6FKHPH ff
PAGE 98
XNf $ $ }r 9 / nM _ Nf L g Ff XNf Â‘ r L? Mr rÂ‘ LU _.LW \fÂ§ U WfÂ§/ fÂ« LZ / fÂ§OL0LPLP r fÂ« r L } 9 9Â Nf fÂ¬ Gf )LJXUH FRQWLQXHG (VWLPDWLRQ 6FKHPH ff
PAGE 99
E D Nf ;(f Hf 8P;I If ENf ENf )LJXUH FRQWLQXHG (VWLPDWLRQ 6FKHPH ff
PAGE 100
&+$37(5 9,,, ',6&866,21 $ JOREDOO\ VWDEOH DGDSWLYH /4 FRQWUROOHU ZKLFK GRHV QRW UHTXLUH SHUVLVWHQW H[FLWDWLRQ ZDV LQWURGXFHG IRU PXOWLLQSXW PXOWLRXWSXW OLQHDU GLVFUHWHWLPH V\VWHPV 7KH DVVXPSWLRQV PDGH RQ WKH SODQW ZHUH f DQ XSSHU ERXQG RQ V\VWHP RUGHU LV NQRZQ f WKH XQNQRZQ V\VWHP SDUDPHWHUV EHORQJ WR NQRZQ ERXQGHG LQWHUYDOV DQG f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
PAGE 101
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f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
PAGE 102
DORQH EXW RQ WKH LGHQWLILFDWLRQ HUURU QRUPDOL]HG E\ WKH VLJQDOV LQ WKH DGDSWLYH V\VWHP ,Q RUGHU WR SURYH JOREDO VWDELOLW\ RI WKH DGDSWLYH V\VWHP .UHLVVHOPHLHU DVVXPHV WKDW WKH XQNQRZQ V\VWHP SDUDPHWHUV H EHORQJ WR NQRZQ ERXQGHG LQWHUYDOV >k7 QU"D[O +H DOVR DVVXPHV WKDW WKH PRGHOOLQJ HUURU PHHWV D FHUWDLQ JURZWK FULWHULRQ ZKLFK ZLOO QRW EH GLVFXVVHG KHUH ,W ZLOO VXIILFH WR PHQWLRQ WKDW .UHLVVHOPHLHU JLYHV VRPH JRRG H[DPSOHV RI W\SHV RI SODQW XQFHUWDLQWLHV ZKLFK PHHW KLV FULWHULRQ %\ LQFRUSRUDWLQJ WKH UHODWLYH GHDG ]RQH LQWR KLV LGHQWLILFDWLRQ VFKHPH .UHLVVHOPHLHU LV DEOH WR SURYH JOREDO VWDELOLW\ RI DQ LQGLUHFW DGDSWLYH FRQWUROOHU IRU 6,62 FRQWLQXRXVWLPH V\VWHPV ZLWK PRGHOOLQJ HUURUV $OWKRXJK .UHLVVHOPHLHUn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f DGGLWLRQ RI D SRVLWLYH WHUP WR WKH FRYDULDQFH 3Nf f D ILQLWH GDWD ZLQGRZ DQG f H[SRQHQWLDO GDWD ZHLJKWLQJ ,Q WKH ILQLWH GDWD ZLQGRZ WKH ROG GDWD DUH GLVFDUGHG SHULRGLFDOO\ E\ VRPH FKRVHQ SDWWHUQ ,Q H[SRQHQWLDO GDWD ZHLJKWLQJ D IRUJHWWLQJ IDFWRU LV LQWURGXFHG LQWR WKH FRYDULDQFH HTXDWLRQ ZKLFK KDV WKH HIIHFW RI ZHLJKWLQJ WKH PRVW UHFHQW GDWD KHDYLHU WKDQ ROGHU GDWD $OO WKUHH PHWKRGV SUHVHUYH WKH IDVW LQLWLDO FRQYHUJHQFH RI OHDVW VTXDUHV ZKLOH SUHYHQWLQJ WKH FRYDULDQFH IURP FRQYHUJLQJ WR ]HUR ZKLFK ZRXOG FDXVH
PAGE 103
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
PAGE 104
5()(5(1&(6 >@ .( 1DUHQGUD DQG /6 9DODYDQL 6WDEOH $GDSWLYH &RQWUROOHU 'HVLJQ'LUHFW &RQWURO ,((( 7UDQV $XWRPDW &RQWU 9RO $& 1R SS $XJ >@ .6 1DUHQGUD <+ /LQ DQG /6 9DODYDQL 6WDEOH $GDSWLYH &RQWUROOHU 'HVLJQ 3DUW ,, 3URRI RI 6WDELOLW\ ,((( 7UDQV $XWRPDW &RQWU 9RO $& 1R SS XQH >@ *& *RRGZLQ 35DPDGJH DQG 3( &DLQHV 'LVFUHWH7LPH 0XOWL YDULDEOH $GDSWLYH &RQWURO ,((( 7UDQV $XWRPDW &RQWU 9RO $& 1R SS XQH >@ *& *RRGZLQ DQG .6 6LQ fÂ¯$GDSWLYH &RQWURO RI 1RQPLQLPXP 3KDVH 6\VWHPV ,((( 7UDQV $XWRPDW &RQWU 9RO $& 1R SS $SULO >@ *& *RRGZLQ '+LOO DQG 0 3DODQLVZDPL $ 3HUVSHFWLYH RQ &RQYHUJHQFH RI $GDSWLYH &RQWURO $OJRULWKPV $XWRPÂ£WLFD 9RO SS >@ *& *RRGZLQ DQG .6 6LQ $GDSWLYH )LOWHULQJ 3UHGLFWLRQ DQG &RQWURO 3UHQWLFH +DOO 1HZ @ + (OOLRW 5 &ULVWL DQG 0 'DV *OREDO 6WDELOLW\ RI $GDSWLYH 3ROH 3ODFHPHQW $OJRULWKPV ,((( 7UDQV $XWRPDW &RQWU 9RO $& 1R SS $SULO >@ $QGHUVRQ DQG 50 RKQVWRQH *OREDO $GDSWLYH 3ROH 3RVLWLRQLQJ ,((( 7UDQV $XWRPDW &RQWU 9RO $& 1R SS DQ >@ * .UHLVVHOPHLHU 2Q $GDSWLYH 6WDWH 5HJXODWLRQ ,((( 7UDQV $XWRPDW &RQWU 9RO $& 1R SS )HE >@ & 6DPVRQ 7KHVH GH 'RFWHXU,QJHQLHXU /DERUDWRLUH Gn$XWRPDWLTXH GH n,5,6$ &DPSXH GH %HDXOLHX 5HQQHV &HGH[ >@ & 6DPVRQ $Q DGDSWLYH /4 &RQWUROOHU IRU 1RQPLPLQXP 3KDVH 6\VWHPV ,QW &RQWURO 9RO SS >@ & 6DPVRQ DQG )XFKV 'LVFUHWH $GDSWLYH 5HJXODWLRQ RI 1RW 1HFHVVDULO\ 0LQLPXP3KDVH 6\VWHPV 3URF ,(( 9RO 3W ' 1R SS 0D\
PAGE 105
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Â£WLFD 9RO SS >@ * .UHLVVHOPHLHU $Q $SSURDFK WR 6WDEOH ,QGLUHFW $GDSWLYH &RQWURO $XWRPÂ£WLFD 9RO SS XO\ >@ 33 .KDUJRQHNDU DQG (' 6RQWDJ 2Q WKH 5HODWLRQ %HWZHHQ 6WDEOH 0DWUL[ )UDFWLRQ )DFWRUL]DWLRQV DQG 5HJXODEOH 5HDOL]DWLRQV RI /LQHDU 6\VWHPV RYHU 5LQJV ,((( 7UDQV $XWRPDWLF &RQWURO 9RO $& SS >@ (: .DPHQ DQG 33 .KDUJRQHNDU 2Q WKH &RQWURO RI /LQHDU 6\VWHPV :KRVH &RHIILFLHQWV $UH )XQFWLRQV RI 3DUDPHWHUV ,((( 7UDQV $XWRPDW &RQWU 9RO $& 1R SS DQ >@ : 5XGLQ 3ULQFLSOHV RI 0DWKHPDWLFDO $QDO\VLV 0F*UDZ+LOO 1HZ
PAGE 106
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Â£WLFD 9RO SS >@ *& *RRGZLQ DQG (. 7HRK $GDSWLYH &RQWURO RI D &ODVV RI /LQHDU 7LPH9DU\LQJ 6\VWHPV ,)$& :RUNVKRS RQ $GDSWLYH 6\VWHPV LQ &RQWURO DQG 6LJQDO 3URFHVVLQJ 6DQ )UDQFLVFR XQH
PAGE 107
%,2*5$3+,&$/ 6.(7&+ 7KH DXWKRU ZDV ERUQ LQ &LQFLQQDWL 2KLR RQ XO\ 6KH UHFHLYHG D %6(( DQG D 06(( GHJUHH IURP WKH *HRUJLD ,QVWLWXWH RI 7HFKQRORJ\ LQ 0DUFK DQG 'HFHPEHU UHVSHFWLYHO\ 6KH H[SHFWV WR UHFHLYH D 3K' LQ HOHFWULFDO HQJLQHHULQJ IURP WKH 8QLYHUVLW\ RI )ORULGD LQ $XJXVW
PAGE 108
, FHUWLI\ WKDW , KDYH UHDG WKLV VWXG\ DQG WKDW LQ P\ RSLQLRQ LW FRQIRUPV WR DFFHSWDEOH VWDQGDUGV RI VFKRODUO\ SUHVHQWDWLRQ DQG LV IXOO\ DGHTXDWH LQ VFRSH DQG TXDOLW\ DV D GLVVHUWDWLRQ IRU WKH GHJUHH RI 'RFWRU RI 3KLORVRSK\ fÂµ OO I ( : .DPHQ &KDLUPDQ 3URIHVVRU RI (OHFWULFDO (QJLQHHULQJ , FHUWLI\ WKDW , KDYH UHDG WKLV VWXG\ DQG WKDW LQ P\ RSLQLRQ LW FRQIRUPV WR DFFHSWDEOH VWDQGDUGV RI VFKRODUO\ SUHVHQWDWLRQ DQG LV IXOO\ DGHTXDWH LQ VFRSH DQG TXDOLW\ DV D GLVVHUWDWLRQ IRU WKH GHJUHH RI 'RFWRU RI 3KLORVRSK\ 7 ( %XOORFN 3URIHVVRU RI (OHFWULFDO (QJLQHÂ«UL QJ , FHUWLI\ WKDW , KDYH UHDG WKLV VWXG\ DQG WKDW LQ P\ RSLQLRQ LW FRQIRUPV WR DFFHSWDEOH VWDQGDUGV RI VFKRODUO\ SUHVHQWDWLRQ DQG LV IXOO\ DGHTXDWH LQ VFRSH DQG TXDOLW\ DV D GLVVHUWDWLRQ IRU WKH GHJUHH RI 'RFWRU RI 3KLORVRSK\ B 3URIHVVRU RI ,QGXVWULDO DQG 6\VWHPV (QJLQHHULQJ , FHUWLI\ WKDW , KDYH UHDG WKLV VWXG\ DQG WKDW LQ P\ RSLQLRQ LW FRQIRUPV WR DFFHSWDEOH VWDQGDUGV RI VFKRODUO\ SUHVHQWDWLRQ DQG LV IXOO\ DGHTXDWH LQ VFRSH DQG TXDOLW\ DV D GLVVHUWDWLRQ IRU WKH GHJUHH RI 'RFWRU RI 3KLORVRSK\ 3 = 3HHEOHV n 3URIHVVRU RI (OHFWULFDO (QJLQHHULQJ
PAGE 109
, FHUWLI\ WKDW , KDYH UHDG WKLV VWXG\ DQG WKDW LQ P\ RSLQLRQ LW FRQIRUPV WR DFFHSWDEOH VWDQGDUGV RI VFKRODUO\ SUHVHQWDWLRQ DQG LV IXOO\ DGHTXDWH LQ VFRSH DQG TXDOLW\ DV D GLVVHUWDWLRQ IRU WKH GHJUHH RI 'RFWRU RI 3KLORVRSK\ I Wn`nFa! rÂ‘ 6 6YRURQRV 3URIHVVRU RI &KHPLFDO (QJLQHHULQJ 7KLV GLVVHUWDWLRQ ZDV VXEPLWWHG WR WKH *UDGXDWH )DFXOW\ RI WKH &ROOHJH RI (QJLQHHULQJ DQG WR WKH *UDGXDWH 6FKRRO DQG ZDV DFFHSWHG DV SDUWLDO IXOILOOPHQW RI WKH UHTXLUHPHQWV IRU WKH GHJUHH RI 'RFWRU RI 3KLORVRSK\ $XJXVW 'HDQ &ROOHJH RI (QJLQHHULQJ 'HDQ *UDGXDWH 6FKRRO
CHAPTER V
ADAPTIVE REGULATOR
The rinput moutput linear discretetime system described by
(31)(33) will be regulated using a MIMO version of Samson and Fuchs'
adaptive LQ controller [12]. As discussed in Chapter Two, a necessary
condition for the adaptive closedloop system to be globally stable is a
uniformly stabilizable estimated system. It will be shown that the
parameter estimator described by equations (42)(45) will give a
uniformly stabilizable estimated system provided that assumptions A1A3
summarized at the end of Chapter Three are satisfied.
Uniform Stabilizability of the Estimated System
Using the parameter estimator described by (42)(45), it will be
shown that the estimated system can be stabilized using state
feedback. As shown in [6], the statespace observer form realization
for the system described by (31) is given by
x(k+l) = Fx(k) + Gu(k) (51)
y(k) = Hx(k)
36
21
Theorem 4.1: The estimation algorithm (42) (45) has the following
properties
i. P(k) converges to a positive semidefinite matrix P^ < 21;
ii.  9..(k) < M < for all k and i = 1,2,...,N, j = l,2,...,m;
J
m
ie(k)
2
nkl
4>^(kl) P( k1) <(>a(kl)
+ 0 as k + ;
iv. iie(k)n < a(k)a(k1)n + b(k), where a(k) and b(k) are positive
scalar sequences which converge to zero;
v. f(e(k)) + 0 as k + , which implies that j(k) converges to the
set [9,I'1n>0^x] for i = 1,2 N, j = 1,2,...,m;
' J \J
VI .
0. .(k) 0..(kp) j * 0 as k for any integer p,
' J J
i 1 y2y t yNy j ^92^ ^
vii. If 0.. 0^" or 0^x for every i and j, then 0..(k) converges
' J J U ^ J
j, fjroi n .max,
to a point in [0.. ,0.. ]*
* j j
Proof: Property (i) is a wellknown property first proved by Samson
[10] with ti^.i equal to one. The proof is included here for the sake of
completeness. Equation (43) implies that 0 < x^P(k)x
all xe RN, and for all k > 0. Therefore, for each x eEN, the se
quence x^P(k)x is monotone decreasing and must converge. Denoting
P.jj(k) as the ij^L element of P(k), it follows that
(e. + eJ.)TP(k)(ei + e^.) elp(k)ei eTp(k)e..
PijOO =
(47)
68
y(k)
u(k)
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
l/l f
E. W. Kamen, Chairman
Professor of Electrical Engineering
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
T. E. Bullock
Professor of Electrical Engineri
ng
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
_
Professor of Industrial and Systems
Engineering
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
P. Z. Peebles '
Professor of Electrical Engineering
59
nz(k+i+np) ii < rnnz(k+i)ii + l rn*^c(k+i+jp)
j=0
n1 .
nz(k+i+np) ii < rn(R9nz(k) ii + R) + Â£ rn Jc(k+i+jp)
d j=0
for all k > N, ie[0,p]. (519)
Since c(k) converges to zero, there exists a finite > N such that
c(k) < f for all k > N,.
1
Applying this inequality to (519) gives
Hz(N^+i+np) ll < rn(R2Hz(N^) It + Rg) + e/2, ie[0,p].
Also, since r < 1 there exists a finite N2 such that
r < I(R2hz(N^)ii + R3) for a11 n > V
Thus
Hz(N^+i+np) 11 < e for all n > N2, ie[0,p]
which implies nz(k) 11 < e for all k > + N2p.
By the definition of z(k), both the system input u(k) = L(k)x(k) and
the system output y(k) must converge to zero. Therefore the adaptive
27
where a(k) and b(k) are positive scalar sequences which converge to
zero. To prove (vi), first subtract e^kl) from both sides of equa
tion (48). Pramultiplying both sides of the resulting equation by its
transpose and noting that e(k) = /eT(kl)<Â¡> (k1) gives
Tr[(e(k)9(kl))T(0(k)9(kl))] = Tr[(e(k) '9(kl))T(0(k) ?(kl))>
e(k)J(kl)P(kl)2* (kl)eT(k)
y p ~ q ~ ~ ~
(Vl + ^(kl)P(kl)^a(kl))2 _
+2Tr
t P(kl)2* (kl)J(kl)9(kl)
fT( 9(kl)) 2 2
Vl + a(k1)p(kl)*a(,c_1)
(413)
+ Tr[fT(e(kl))P(kl)2f(e(kl))].
Also
I(kl)P(kl)2* (k1) a(kl)P(klH (k1)
3 3 < CW>()3 TTT
n2_! + ^(kl)P(kl)*a(kl)
nkl++aik"1,p(k'1)*a(k1)
< X P(0)
max '
(419)
Combining (418) and (419) gives
TrC(0(k)e(kl))1(0(k)e(kl))3 < CxmaxP(0)]
ie(k) ii'
Vl+Vk1)P(k_1)+a(k1)
+ 2Tr
T P(kl)2 (kl)/(kl)0(kl)
fT(0(kl))^ a a
nkl + a (kDP(kl)^(kl)
+ *2ax (P(O))Tr[fT(0(kl))f(0(kl))]. (420)
77
Example 7.3: Consider the discretetime system described by
y(k) = 2y(kl)5y(k2)u(kl)0.5u(k2)
The adaptive controller defined by the observer (53) and the control
law (65) with L(k) computed from (66) was implemented to force the
output to track the reference signal r(k)=5. A step disturbance defined
by
Of
of
K < 40
K > 40
was introduced. The internal model was given by
xc(k+l) = xc(k) + e(k)
with e(k)=y(k)5. The parameters were estimated using the algorithm
defined by (42)(45). The initial conditions were chosen to be:
P(0)=I, x(0)=[l of, xc(0)=0, R0=I. The initial parameter estimates,
parameter ranges, and steady state estimated values are displayed in
Table 73. The output response, control input, and parameter estimates
are shown in Figure 75. For comparison purposes, plots of the output
responses and control input when the parameters are known exactly are
shown in Figure 76. As expected, the transient response is
considerably worse when the system parameters must be estimated.
However, once the system is in steady state, the adaptive system seems
to respond equally well to a sudden step disturbance as the system using
exact parameters.
35
ne(n)n < iP(n) 11 nP"1(0)'e(0) n + nP(n) u n l P1(i+l)P(i )f (e(i)) n.
i =0
nr 1
It was shown previously that ii l P (i+l)P(i )f (e(i)) n is bounded in
i=Q
the limit as n  . Also iiP(n)n can be made arbitrarily small for n
sufficiently large if P(k) converges to zero. Thus, e^n) converges to
zero as n which implies e(k) converges to the true system param
eter matrix e whenever P(k) converges to zero.
67
not. Figures 7173 contain plots of the output response, control
input, and estimated parameters for each of the three cases. A
comparison of Figures 71 and 72 show the estimation algorithm defined
by (42)(45) has a slightly better transient response than least
squares. Figure 73 illustrates the detrimental effect of the data
normalization on the transient response of the system. As previously
mentioned, e in (45) should be chosen very small to minimize the effect
of data normalization on the system transient response.
Table 71
DATA
"al
~a2
bl
b2
Actual Parameters
2.0
0.99
0.5
3.0
Initial Estimates
1.95
1.0
0.75
6.0
Parameter Ranges
[1.9, 2.0]
[2.0, 4.0]
[0.5, 2.0]
[2.5, 9.5]
Estimation Scheme
(42)(45)
2.00
1.00
0.530
3.04
Recursive Least
Squares
2.11
1.05
0.594
3.19
Estimation Scheme
(42)(45) with
= max(l,ne(k)n)
2.00
0.991
0.571
3.14
38
Proposition 5.1: Suppose assumption A3 is satisfied so that
(F(e) ,G(e)) is pointwise stabilizable over n. Then there exists an
e > 0 such that (F(e),G(e)) is also pointwise stabilizable over n .
Proof: Suppose (F(e)G(e)) is stabilizable over ft. For each 9
belonging to ft, there exists a bounded open neighborhood VQ of 9 such
that (F(e),G(9)) is stabilizable over V0. This result is proved in
[25], by first showing that there exists a feedback matrix L(e) with
entries which are continuous in 9 such that
x (F(9)G(e)L(0)) < q(9) < 1. The existence of the open
neighborhood V0 of 9 then follows because the eigenvalues of
(F(e)G(e)L(9)) are continuous functions of 9. Since ft is a compact
subset of I'*xm and U V0 is an open cover of n, there exists a
9 eft
finite subcover U V of ft. For k=l,2,..., let ek = ( V2 )k. For
j=l 9j
r
every k > 1, n U Va isa compact subset of FNxm and
ek j=l 9j
r r
ft U VQ C ft U Vfl
ek+l j=l 9j ek j=l 9j
Suppose ft U V. is nonempty for all k > 1. By the properties of
ek j=l 9j
nested compact sets (see [26, p.38]) there is at least one point p which
belongs
k > 1
r
to ft U for all k > 1.
ek ji ej
Therefore, nen
ek
which implies that pen.
However, p^
r
U V
j=l
for all
which forms an
40
Property (v) in Theorem 4.1 implies that there exists a finite N such
that e(k) belongs to ne for all k > N. It then follows from the
result in Kamen and Khargonekar given above that there exists a feedback
gain sequence L(k) continuous in 9 such that
< q < 1 for all k > N.
fnaX
Since the entries in (F(k)G(k)L(k)) are continuous functions of 9,
property (vi) in Theorem 4.1 implies [F(k)G(k)L(k)]
[F(kl)G(kl)L(kl)] converges to zero as k goes to infinity. Oesoer
[28] proves A(k) will be uniformly asymptotically stable as a system
matrix if (i) all eigenvalues of A(k) have magnitude less than some
constant q < 1 for all k and (ii) A(k)A(k1) converges to zero as k
goes to infinity. Using Desoer's results on stability of slowly time
varying systems, (F(k)G(k)L(k)) is uniformly asymptotically stable as a
system matrix. Thus the estimated system (F(k),G(k)) is uniformly
stabilizable as a timevarying system.
The adaptive control law for regulating the system described by
(31) is of the form u(k) = L(k)x(k). The state estimate x(k) is
generated from an adaptive observer which will be discussed in the next
section. The feedback gain matrix L(k) is chosen to stabilize the
estimated system (F(k),G(k)). The existence of a stabilizing feedback
gain L(k) is guaranteed by Theorem 5.1. Several possible options for
L(k) are given in the section following the adaptive observer section.
CHAPTER VI
APPLICATION TO TRACKING AND DISTURBANCE REJECTION
The adaptive controller derived in Chapter V can be applied to the
problem of tracking with disturbance rejection using the internal model
principle. The internal model principle has been discussed by a number
of individuals, see for example [20].
Consider the rinput moutput plant described by
where
F =
x(k+l) = Fx(k) + Gu(k) + Dv(k)
y(k) = Hx(k)
(61)
'Ai I
B1
A?
2 I
b2
G =
Vi I
Bn1
An 0
. Bn .
H = [I 0 ... 0].
The vector v(k) consists of the exogenous disturbance signals. As in
previous chapters, it is assumed that some or all of the entries in the
matrices Aj and Bj are unknown and will be estimated using the parameter
estimator described by (42)(45). Letting r(k) denote an mvector
reference signal, the objective is to design an adaptive controller
which stabilizes the system described by (61) and forces the tracking
error e(k) = y(k)r(k) to converge to zero as k > .
61
15
ft {0s IR
Nxm
V
r imn inax,..
L0.. ,0.. j}.
The concept of compactness (i.e., closed and bounded) is relative to the
matrix norm previously defined.
The only additional assumption made on the system is a pointwise
stabilizability assumption over ft. Specifically, the system described
by (33) is assumed to be stabilizable for each 0eft. This assumption
can be verified a priori because it depends only on the known set ft, not
on the parameter estimates. The stabilizability condition can be tested
using the transfer function matrix of the system. Let D(z"*) and N(z"*)
denote the polynomial matrices defined by
OU1) = 1+1 A. (z~J), NU"1) = l B.(zj).
j=l J j=l J
The coefficients of D(z"l) and N(z"^) depend on the system parameters
and therefore can be viewed as functions of 0. This dependancy is made
explicit by denoting Diz1) and N(z1) as D(z1,0) and N(zl,e)
respectively. The transfer function matrix of the system described by
(31) is then given by
W(z1,0) = D1(z"1,8)N(z1,9).
The stability assumption is that the system described by (31) with
transfer function matrix W(z1, 0) can be stabilized by dynamic output
feedback for all 0eft. From well known results [24], this assumption
is equivalent to the following rank condition:
31
also bounded for all k in the case when the determinant of P(k) con
verges to zero. Therefore, there exists a finite v such that
nP*(k+l)P(k) ii < v < for all k. Applying this result to the in
equality (421) gives
1 a/ ml
9i (m)e. .(n)  < nP(m)P(n) n nP"i(0)9(0)ii + 2v l nf(e(i))n
J i=n
n1
+ nP(m)P(n)nv Â£ nf(e(i))a. (423)
i=0
In order to complete the convergence proof, it must be shown that
l ilf(e(k)) n < L < . It follows from equation (413) that
k=0
V(k) < V(kl) 
iie(k)
Vi + +Â¡(kl)P(kD*a(kl)
 2Tr[e (kl)f(e(kl))]
+ xmax[P(kl)P"1(k)P(kl)]Tr[fT(0(kl))f(9(kl))]. (424)
Now assume < 9.. <0^X for i = 1, ..., N and j = 1, ..., m (i.e.,
0i is an interior point of [el1!11, 0n!'?x]). Using the definition of
'J ij ij
f(0(kl)) in (44), if fij(0(kl)) 0 then eij(kl) > q > 0 where q
n IT13X mirin a i,
= [(j ij. 'ij,] N.
J
j=l, ..., m. Also from the definition of f(e(kl)), it must be true
that the sign of e^kl) is equal to the sign of fin(9(kl)) for i=l,
ij
..., N and j=l, ..., m. These facts when combined with equation (424),
imply
V(k) < V(kl) 
N m
2 y 2q l Â£ fii (0(kl))
 ,ln' 1' i=l j=l J
ne(k) ii'
nkl+ i>;(kDP(kl)
+ ^xCPknP'^kPkmrC^eUUntetk1))].
Abstract of Dissertation Presented to the Graduate School
of The University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
NEW RESULTS IN INDIRECT ADAPTIVE CONTROL
By
KATHLEEN A. K. OSSMAN
August 1986
Chairman: Dr. Edward W. Kamen
Major Department: Electrical Engineering
An adaptive regulator which does not require a persistently
exciting input is derived for multiinput multioutput linear discrete
time systems. The assumptions made on the unknown plant are (1) an
upper bound on the system order is known, (2) the system parameters
belong to known bounded intervals, and (3) the plant is stabilizable for
all possible values of the unknown system parameters ranging over the
known intervals. A recursive parameter estimator is described which
forces the estimates of the system parameters to converge to the known
intervals asymptotically. Using this parameter estimator, an adaptive
LQ regulator is developed which results in a globally stable adaptive
closedloop system in the sense that the system inputs and outputs
converge to zero asymptotically. The results are then extended to the
case of tracking and/or deterministic disturbance rejection using the
internal model principle. Simulations illustrating the performance of
the adaptive controller for several discretetime systems are included.
iv
56
Also, the regression vector <{>(k) defined in (32) can be written in
the form
4>( k) = S(j>(k1) + D(k)x(k) + V(k)
where
"o ... 0 0 ... 0
H
1
CD
1
ynD
0
0
0 ... 0 0 ... 0
, D(k) =
L(k)
, v(k) =
0
0
Ip(nl) :
0
0
0

flO 1
Defining z(k+l) =
x(k+l)
gives
where
A(k) =
S
0
z(k+l) = A(k)z(k) + w(k)
(511)
D(k)
F(k)G(k)L(k)
w(k) =
v(k)
M(k)e(k)
The matrix S is a stable matrix since it is lowerblock triangular. The
matrix F(k)G(k)L(k) is stable by construction; therefore, A(k) in
(511) is exponentially stable as a timevarying system matrix. In
addition, it follows easily from (56) and the boundedness of M(k) that
iiw(k)n < a(k)nz(k)n + b(k)
(512)
where a(k) and b(k) are positive scalar sequences which converge to zero
as k . Thus, (511) can be viewed as an exponentially stable time
varying system driven by an input w(k) which can grow no faster than
linearly with the state, z(k). Following the proof in [12], it will be
shown that iiz(k) t converges to zero. Let A(k+N,k) denote the state
transition matrix for (511) defined by
42
This expression is easily derived by first rewriting (53) as
*00 =
0 I
+
Bx(k1)
B2(k1)
u(kl) +
A1(k1)
A2(k1)
y(ki).
i
0 0
Bn(k1)
_An(kl)_
(55)
Multiplying equation (55) by H gives
y(k) = Hx(k) =[010 ... 0]x(kl) + B^kDuUl) A^kDyikl)
Substituting into this equation for x(kl) using (55) yields
y(k) =[001 ... 0]x(k2) + B1(kl)u(kl) + B2(kl)u(k2)
 Ax(kl)y(k1) A2(k2)y(k2)
Continuing the substitution using (55) gives
* n1
y(k) = [0 ... 0 I] x(kn+l) + l B. (kj)u(kj)
j=l J
n1
 I A.(kj)y(kj)
j=l J
= l B (kj)u(kj) l A (kj)y(kj)
j=l J=1
which is equivalent to (54).
REFERENCES
[1] K.E. Narendra and L.S. Valavani, "Stable Adaptive Controller
DesignDirect Control," IEEE Trans. Automat. Contr., Vol. AC23,
No. 4, pp. 570583, Aug. 1978.
[2] K.S. Narendra, Y.H. Lin, and L.S. Valavani, "Stable Adaptive
Controller Design, Part II: Proof of Stability," IEEE Trans.
Automat. Contr., Vol. AC25, No. 3, pp. 440448, June 1980.
[3] G.C. Goodwin, P.J. Ramadge, and P.E. Caines, "DiscreteTime Multi
variable Adaptive Control," IEEE Trans. Automat. Contr., Vol. AC
25, No. 3, pp. 449456, June 1980.
[4] G.C. Goodwin and K.S. Sin, Adaptive Control of Nonminimum Phase
Systems," IEEE Trans. Automat. Contr., Vol. AC26, No. 2, pp. 478
483, April 1981.
[5] G.C. Goodwin, D.J. Hill, and M. Palaniswami, "A Perspective on
Convergence of Adaptive Control Algorithms," Automtica, Vol. 20,
pp. 519531, 1984.
[6] G.C. Goodwin and K.S. Sin, Adaptive Filtering, Prediction and
Control, Prentice Hall, New York, 1984.
[7] H. Elliot, R. Cristi, and M. Das, "Global Stability of Adaptive
Pole Placement Algorithms," IEEE Trans. Automat. Contr., Vol. AC
30, No. 4, pp. 348356, April 1985.
[8] 8.0.0. Anderson and R.M. Johnstone, "Global Adaptive Pole
Positioning," IEEE Trans. Automat. Contr., Vol. AC30, No. 1, pp.
1122, Jan. 1985.
[9] G. Kreisselmeier, "On Adaptive State Regulation," IEEE Trans.
Automat. Contr., Vol. AC27, No. 1, pp. 316, Feb. 1982.
[10] C. Samson, These de DocteurIngenieur, Laboratoire d'Automatique
de 1'IRISA, Campue de Beaulieu, 35042 Rennes Cedex.
[11] C. Samson, "An adaptive LQ Controller for Nonmiminum Phase
Systems," Int. J. Control, Vol. 35, pp. 128, 1982.
[12] C. Samson and J.J. Fuchs, "Discrete Adaptive Regulation of Not
Necessarily MinimumPhase Systems," Proc. IEE, Vol. 128, Pt. D,
No. 3, pp. 102108, May 1981.
100
63
by (61) and will drive the tracking error e(k) to zero asymptotically.
Theorem 6.1: Suppose assumptions A1A3 are satisfied and in addition
rank
zIF(e) 0 G(e)
BCH zIAc 0
n+mq for all eft and
for all z: z > 1
where n = {0e ffTxm: 0. .e [o'!1!n,0^x] Let x (k) be generated from the
ij ij ij
error driven system described in (63) and x(k) be generated from the
adaptive observer given by (53). Then there exists a control law u(k)
= LjÂ£(k)L2Xc(k) which when applied to the plant (61) results in a
stable closedloop system and drives the tracking error e(k) = y(k) 
r(k) to zero.
Proof: Using equations (53) and (63) and setting v(k) and r(k) equal
to zero, we obtain
x(k+l)
xc(k+l)
F(k)
BCH
0
x(k)
xc(k)
G(k)
0
u(k) +
(64)
where e(k) = Hx(k) y(k). Assuming the given rank condition holds, it
follows from property (v) of Theorem 4.1 that the system described by
(64) is stabilizable. The stabilizing control law is of the form
u(k) = L^k^k) L2xc(k).
(65)
A
F(k)
0
A
G('kf
Let F(k) =
_BCH
Ac_
and G(k) =
0
As in the case of
the
adaptive regulator d
6
The key properties of the leastsquares scheme used for proving global
stability of an indirect adaptive controller are
i.ne(k)n
ii.ii9(kje(kl) h > 0 as k +
iii.lim =Â¡= 6 =0
k+ 1 + <(> (kl)P(kl)<(>(k1)
where e(k) = y(k) e^(kl)(kl).
The error e(k) is referred to as the prediction error since it is simply
the difference between the actual output at time k and the predicted
output at time k using the most recent parameter estimate. These pro
perties do not depend on the type of control input chosen or on the
boundedness of the system input and output. The leastsquares algorithm
also has the property that the parameter estimates converge, although
not necessarily to the true values. This property is not included among
the key properties because it is not necessary for stability proofs.
An example of indirect adaptive control in the SISO case is Samson
and Fuchs' LQ controller discussed in [12] and briefly outlined here. A
statespace observer realization of the system described by (21) is
given by
x(k+l) = Fx(k) + Gu(k) (24)
y(k) = Hx(k)
43
The error between the output of the adaptive observer (53) and the
output of the system (31) as defined as
e(k) y(k) y(k).
The observer error e(k) has the same properties as the prediction error
e(k) given in Theorem 4.1; that is,
iie(k)n < a'(k)i*(kl)i + b'(k) (56)
where a1(k) and b'(k) are positive scalar sequences which converge to
zero. This is shown by first rewriting e(k) as
e(k) = y(k) [0T(klHa(kl) + *\(kl)]
+ [eT(kl)a(kl) + .\(kl)] y(k)
e(k) = y(k) [eT(klHa(kl) + /ib(kl)] e(k).
Taking the norm of both sides and using the triangle inequality gives
iie(k)ii < ny(k) (9T(kl)b(kl))ii + ne(k)n
Noting from (31)(33) that eT(k1)4>a(k1) + ^^(kl) is equal to
[A^(kl) ... An(k1) B1(k1) ... 3n(k1)]<Â¡>(kl) and using expression
(54) gives
87
Example 7.5: Consider the single input two output discretetime system
described by
y^k)"
l
a
*1
(k1)'
'3 *
y2(k)
l
2
*2
(k1)
b
u(kl)
where a and b are unknown parameters. As discussed in Chapter Three,
the required computations for parameter estimation can be reduced by
separating the known parameters from the unknown parameters. This is
accomplished by rewriting the system equations as
y(k) = eT4>a(kl) + ipT4>5 (k1)
with
T
0
a 0'
.a(ki) 
rH
1
CM
L>>
,/ =
.0 b.
u(kl).
t
0 3
2 0
b(kl)
y1(kD'
y2(ki)
u(kl)
The adaptive regulator defined by the observer (53) and the control law
(510) with L(k) computed using (58)(59) was implemented. The system
parameter matrix 9 was estimated using the algorithm (42)(45). The
initial conditions were chosen to be: P(0)=I, x(0)=[l 1]T, and RQ=I.
The initial estimates, parameter ranges, and steady state estimates are
given in Table 75. Plots of the output response, control input, and
parameter estimates are shown in Figure 79.
Internet Distribution Consent Agreement
In reference to the following dissertatirrr
AUTHOR: Ossman, Kathleen
TITLE: New Results.in Indirect Adaptive Control (record number: 985894)
PUBLICATION DATE! 1986
, as copyright holder for the
aforementioned dissertationrherebygrantspycific add limited archive and distribution rights to
the Board of Trustees ofthgaHowoSity of Floridaiand its agents. I authorize the University of
Florida to digitize and distribute the dissertation described above for nonprofit, educational
purposes via the Internet or successive technologies.
This is a nonexclusive grant of permissions for specific offline and online uses for an
indefinite term. Offline uses shall be limited to those specifically allowed by "Fair Use" as
prescribed by the terms of United States copyright legislation (cf, Title 17, U.S. Code) as well as
to the maintenance and preservation of a digital archive copy. Digitization allows the University
of Florida to generate image and textbased versions as appropriate and to provide and enhance
access using search software.
This grant of permissions prohibits use of the digitized versions for commercial use or profit.
Personal information blurred
'/'o/*2ooÂ£/
Date of Signature
Please print, sign and return to:
Cathleen Martyniak
UF Dissertation Project
Preservation Department
University of Florida Libraries
P.O. Box 117007
Gainesville, FL 326117007
5/28/2008
79
yflO
Figure 75
Estimation Scheme (42)(45)
99
estimation to cease. Most likely, these modifications could be used in
the estimation scheme introduced in Chapter Four since it is similar to
the leastsquares algorithm. Thus far, the only results on stability of
adaptive controllers applied to timevarying systems require precise
assumptions on the parameter variations, see for example Goodwin and
Teoh [37].
Due to the difficulty of rigorous analysis, the transient response
of systems using adaptive controllers has received very little
attention. Much of the literature on adaptive control contains only
algorithms not simulations. The simulations included in Chapter Seven
show wide variations in the system output response and control input
initially. Once the identification scheme begins to converge, the
system response becomes wellbehaved.
There are many open questions in the area of adaptive control. As
indicated in the discussion above, it may be possible to modify the
adaptive LQ controller developed in this dissertation for timevarying
systems or for systems with unmodelled dynamics.
76
Figure 74 continued
Estimation Scheme (42)(45)
14
y(k) =
y2(ki)
eT =
+a(ici) = y2(ki)
1 0 3
1 2 1
1 0 3
1 2 1
+(kl)
b(kl) = (k1).
Again, consider the linear discretetime system described by
(33). It is assumed that each component e^. of the unknown system
parameter matrix 9 belongs to a known bounded interval [0i!*1.n, a1??*].
This assumption is reasonable in those applications where some a priori
information is available on the system. Clearly, the unknown system
parameter matrix 9 belongs to a subspace of ]R^xm which is defined by
the known bounded intervals. In order to use the necessary concepts of
boundedness, compactness and convergence, a suitable norm must be
defined on jRNxm. The vector norm which will be used throughout this
dissertation is the Euclidean norm defined by
IX l = (x x)
V,
The matrix norm which is induced by the vector Euclidean norm is given
by
T ^2
* CWM M)]
where ^max^1^ is the ei9envalue of MTM with the greatest
magnitude. By the above assumption on the entries of 0, the unknown
system parameter matrix 0 belongs to a known compact subspace n of ]RNxm
given by
78
Table 73
DATA
"al
"a2
bl
b2
Actual Parameters
2.0
5.0
1.0
0.5
Initial Estimates
0.5 '
6.0
1.0
0.0
Parameter Ranges
[3, 2]
I1
1
CO
1
1 1
[1.1, 0.9]
[1, i.o]
Estimation Scheme
(42)(45)
1.64
4.3
0.90
0.478
73
Example 7.2: Consider the nonminimum phase stabilizable but not
reachable discretetime system described by
y(k) = 2.6y(kl)2.13y(k2)+0.54y(k3)+u(kl)+1.5u(k2)u(k3).
The adaptive controller defined by the adaptive observer (53) and the
control law (65) with L(k) computed from (66) was implemented to force
the output to track the reference signal r(k) =10. No disturbance was
introduced in this example. The internal model was given by
xc(k+l) = xc(k) + e(k)
with e(k) = y(k)10. The parameters were estimated using (42)(45).
The initial conditions were chosen to be: P(0) = I, x(0) = [0 1 0]T,
xc(0)=0, and R0=I. The initial parameter estimates, parameter ranges,
and steady state values of the estimates are displayed in Table 72.
Plots of the output response, control input, and parameter estimates are
shown in Figure 74.
Equation (44) indicates that the initial "covariance" matrix P(0)
should satisfy 0
the system rapidly went unstable. At time k=20, the output was
y(k)=6.5xl0 and the control input was u(k)=2.6xl01^. The parameter
estimates at k=20 were a1(k)=451, a2(k)=943, a3(k)=484, bj(k)=1.0,
b2(k)=447, and b2(k)=898. Thus, choosing P(0) smaller than 21 is not
merely a technicality in the proof of Theorem 4.1, but is a necessary
condition for preserving stability of the adaptive closedloop system.
25
Since P(k) converges, equation (43) implies that the matrix
P(kl)a(kl)<}>g(kl)P{kl)
Vi + k1}
converges to zero. For k sufficiently large, n. 1 = max (1, n<(> (kl) n)
Kl a
which implies that
P(klHa(kl)^(kl)P(kl) P(klHa(kl)<(,J(kl)P(kl)
nkl + ia(k1)P(k1)<(,a(k"1) \l + xmaxP(0),+a(k"1)Z
P(kl) (kl)
> a a > o
Vl^1 +
P(klU (kl)
It then follows that the matrix ^ converges
Vl
to zero since it is bounded above by a matrix which is converging to
zero. Thus by equation (416), the matrix 21 P(kl)P"l(k)P(kl) also
converges to 21 PM in the case when the determinant of P(k) converges
to zero. Therefore, the matrix 21 P(kl)P"^(k)P(kl) is positive
definite for k sufficiently large since the matrix converges to the
positive definite matrix 21 P^. By (415), this implies that V(k) is
a monotone decreasing sequence for k sufficiently large and must there
fore converge. It then follows that 9. .(k) is bounded (property (ii))
' J
by noting that
75
y(k)
U(k)
24
Combining equations (413) and (414) gives
V(k) < V(kl) 
iie(k) ii^
nkl + Â¡.(kl)P(KlHa(kl)
(415)
TrCfT(e(kl))C2IP(kl)P1(k)P(kl)]f(0(kl))].
If the matrix 21 P(kl)P"*(k)P(kl) is positive definite for k suffi
ciently large, then V(k) will be a monotone decreasing sequence which
must converge since V(k) is bounded below by zero. In order to show
that the matrix 21 P(kl)P1(k)P(kl) is positive definite for k
sufficiently large, first assume that the determinant of P(k) does not
converge to zero. In this case, P"^(k) converges to p~* which implies
00
that 21 P(kl)P"l(k)P(kl) converges to 21 P^. The matrix 21 P^
is positive definite by property (i) in Theorem 4.1. Mow suppose that
the determinant of P(k) does converge to zero. Using (43) and the
matrix inversion lemma gives
^(kiH^kD
P1(k) = p^kl) 
2
11 k1
Multiplying both sides of this equation on the left and right by P(kl)
and adding 21 to both sides give
21 P(kl)P1(k)P(kl) = 2IP(k1)
P(klHa(klHÂ¡(kl)P(kl)
2
n
k1
(416)
102
[27] B.D.O. Anderson and J.B. Moore, "Detectability and Stabilizability
of TimeVarying DiscreteTime Linear Systems," SIAM J. Control,
Vol. 19, No. 1, pp. 2032, Jan. 1981.
[28] C.A. Desoer, "Slowly Varing Discrete System xt+1 = Atxt,"
Electron. Lett., No. 6, pp. 339340, 1970.
[29] G. Kreisselmeier, "Adaptive Control via Adaptive Observation and
Asymptotic Feedback Matrix Synthesis," IEEE Trans. Automat.
Contr., AC25, No. 4, pp. 717722, Aug. 1980.
[30] K.A. Ossman and E.W. Kamen, "An Adaptive LQ Controller for MIMO
Linear DiscreteTime Systems," in Proc. 24th IEEE Conf. Decision
Contr., Ft. Lauderdale, FL, pp. 843844, 1985.
[31] K.A. Ossman and E.W. Kamen, "A Parameter Estimator with
Convergence to PreSpecified Intervals for DiscreteTime Systems,"
in Proc. Amer. Contr. Conf., Seattle, Wash., 1986.
[32] K.A. Ossman and E.W. Kamen, "Adaptive Regulation of MIMO Linear
DiscreteTime Systems Without Requiring a Persistent Excitation,"
submitted to IEEE Trans. Automat. Contr., Jan. 1986.
[33] W.L. Green and E.W. Kamen, "Stabilizability of Linear Systems Over
a Commutative Normed Algebra with Applications to Spatially
Distributed and ParameterDependent Systems," J. Contr. and
Optimization, pp. 118, Jan. 1985.
[34] G. Kreisselmeier and K.S. Narendra, "Stable Model Reference
Adaptive Control in the Presence of Bounded Disturbances," IEEE
Trans. Automat. Contr. Vol. AC27, No. 6, pp. 11691168, Dec.
1982.
[35] B.B. Peterson and K.S. Narendra, "Bounded Error Adaptive Control,"
IEEE Trans. Automat. Contr., Vol., AC27, No. 6, pp. 11611168,
Dec. 1982.
[36] C. Samson, "Stability Analysis of Adaptively Controlled Systems
Subject to Bounded Disturbances," Automtica, Vol. 19, pp. 8186,
1983.
[37] G.C. Goodwin and E.K. Teoh, "Adaptive Control of a Class of Linear
TimeVarying Systems," IFAC Workshop on Adaptive Systems in
Control and Signal Processing, San Francisco, June 1983.
39
open cover of ti and hence we have a contradiction. Thus,
r
fl_ 1) V. must be empty for
k ji ej
will be pointwise stabilizable over
k > kj which implies (F(9),G(e))
n for any k > k..
ek 1
Let 9(k) be defined as the estimate at time k of the system
parameter matrix 9, and let (F(k),G(k)) denote the timevarying
estimated system (F(9(k)) ,G(9(k))).
Definition: The estimated system (F(k),G(k)) is uniformly stabilizable
as a timevarying system [27], if there exists a bounded feedback matrix
l(k) such that (F(k)G(k)L(k)) is uniformly asympototically stable as a
timevarying system matrix.
The following theorem gives conditions under which the estimated system
is uniformly stabilizable.
Theorem 5.1: Assuming A1A3 are satisfied, the parameter estimator
described by (42)(45) will generate an estimated system (F(k),G(k))
which is uniformly stabilizable as a timevarying system.
Proof: By Proposition 5.1, there exists e > 0 such that (F(9),G(e)) is
pointwise stabilizable for all 9 belonging to nÂ£. Using the results in
Kamen and Khargonekar [25], since is compact, there exists a feedback
matrix L(9) with entries which are continuous in 9 and a positive
constant q < 1 such that
x (F(g)G(e)L(e))I < q < 1 for all 9efi
i max\ e
51
Since Rk belongs to the compact set D for all k, it follows from step
five that given any a > 0 there exists 6(a) such that
ns^n < a whenever ue(k+i) 0(k) 11 < 5 for ie[0,N].
k k
Let Qk = Pk# Using the equation introduced in step two gives
Qk+i+l = Rkt1G(0(k))[GT(e(O>Rk+iG> + n ie(k)))
X (Qfc+i + s(9(k))CGT'(8(k))Pki,s(e(k)) + I]'1
X GT(8(k))Q
X GT(e(k)))T + Ik+i; Qk = 0, ieCO.N].
It follows from the boundedness of e(k), Rk, and pjj (for t
sufficiently large) that
llFT(e(k))(I Rk+iG(Q(k))CGT(8(k))Rk+iG(0(k)) + I]V(0(k))) 11 < 
and iiG(e(k))[GT(e(k))PkJ\ G(e(k)) + I]"1GT(8(k)) 1 < K2 < .
Thus for all k such that e(k)ei2^, we have
"Vi+l1 < Kl2('Cl' + K2,GHi'2> + 'W* Qk =  eC0N:>
19
1 when the determinant of P(k) > e
nk_1 = / where s is any small positive number. (45)
max (1, ii<> (k1) ii) otherwise
k a
The parameter estimation algorithm described by (42) (45)
differs from the recursive leastsquares estimator in three ways.
First, the initial covariance matrix P(0) in (43) must be less than 21,
whereas the leastsquares algorithm allows for any positive definite
initial covariance matrix. This condition is needed to ensure that the
parameter estimator described by (42) (45) possesses the properties
listed in Theorem 4.1 which are necessary for proving global stability
of the adaptive regulator discussed in Chapter V.
The second difference is the data normalization introduced through
the term ^ defined in (45). In the leastsquares algorithm,
nk i is simply equal to one. Again, the data normalization is neces
sary for proving that the parameter estimator has the desirable proper
ties given in Theorem 4.1. Simulations have shown that if e is not
suitably small, the normalization has a detrimental effect on the
transient response of the system. This effect can be minimized by
choosing e to be very small in (45). Simulations which show the effect
of the data normalization and the choice of e on system transient re
sponse are discussed in Chapter VII.
The major difference between the estimator given by (42) (45)
and the recursive leastsquares estimator is the addition of a "correc
tion term" P(kl)f(e(kl)) in (42). This term forces the estimates
8.jj(k) of the components of the system parameter matrix 0 to converge to
the sets Ce?lnT^XU. This property, when combined with the stabiliz
ability assumption over n discussed in Chapter III, eliminates the need
CHAPTER II
BACKGROUND
In indirect adaptive control, the system parameters are estimated
in real time using available input/output information. The control law
is then computed for the estimated system using the most recent param
eter estimates. A wellknown problem in proving global stability of
indirect adaptive control algorithms is the estimated system may not be
controllable or even stabilizable for certain parameter estimates which
are referred to as singular points. Global stability of the indirect
adaptive controller can only be ensured if there is some finite point in
time after which the parameter estimates are not arbitrarily close to a
singular point. An example of a parameter estimator and an adaptive
regulator will further illustrate the problem of singular points in
indirect adaptive control.
Consider the SISO discretetime system described by the input/out
put difference equation:
q
P
(21)
In (21), y(k) is the system output and u(k) is the control input. It
is assumed that an upper bound, n, on q and p is known but some or all
of the system parameters aj and bj are unknown. The system described by
(21) can be rewritten in the following form which is convenient for
parameter estimation:
y(k) = eT(kl)
4
20
for a persistently exciting input. It is important to note that the
algorithm is not a projection algorithm. That is, 0.Â¡j(k) does not
belong to the set [0l.n1.n,et!?x] for every k but does converge to the set
T J 1J
in the limit as k  .
The idea for the correction term came from Kreisselmeier [23]. In
fact, f(e(kl)) is defined exactly the same as in [23]; however, the
parameter estimator described here for MIMO discretetime systems is
quite different from the adaptive observer Kreisselmeier uses for SISO
continuoustime systems.
The following theorem shows the estimator described by (42) 
(45) has many of the same properties as recursive leastsquares. In
addition, the estimates 0.jj(k) converge to the known intervals
[9?]n9^x]. The prediction error e(k) which appears in the theorem
is defined as
e(k) = y(k) eT(kl)a(kl) /^(kl).
26
V(k) Tr(eT(k)p1(k)'e(k)) > [A^P'V)] Tr(eT(k)'e(k))
T N m ,
v(k) > V2 Tr(T(k>r(k)) = V2 ^ ^ Ce1j(k) 0,j].
Now solving (415) recursively gives the relationship
q ie(j)i
0 < V(q) < V(0) .I 2 Y77
j=l njl +
(417)
 I Tr[fT(0(jl))[2I P(jl)P"1(j)P(jl)]f(e(jl))].
j=l
For sufficiently large k, the terms within the summations are nonnega
tive and must converge to zero since V(k) is nonnegative. Properties
(iii) and (v) then follow easily. Property (iv) follows immediately
from (iii) by first noting that
iie(k)
(nki+ *I
< b(k)
where b(k) is a positive scalar sequence which converges to zero,
Multiplying both sides of the inequality by
2
(nÂ£_i + a(kl)P(kl)a(kl)) and noting that < 1 + iia(k1) ii
gives
ne(k) < b(k)(n^ + Â¡(kl)P(kl)*a(kl))
< b(k)(Vl + UmaxP(><l))1/2 *+a(kl)i)
< a(k)
71
Figure 72 continued
Recursive Least Squares
10
In order to prove global stability, Samson and Fuchs must assume
the estimated system (F(k),G(k)) is uniformly stabilizable. Unfortu
nately, this stabilizability condition cannot be checked a priori which
means global stability of the adaptive regulator cannot be guaranteed.
If the leastsquares estimator is used, the assumption that (F(k),G(k))
is uniformly stabilizable is equivalent to requiring that is
stabilizable where F = lim F(k) and G = lim G(k).
oo KM 00 KH
As previously mentioned, the problem of avoiding singular limit
points is common to all indirect adaptive controllers. Until the recent
work of LozanoLeal and Goodwin [21], Oe Larminat [22] and Kreisselmeier
[23], the only way to avoid a singular limit point was to require a
persistently exciting input which would force the parameter estimates to
converge to the true system parameters. This approach has been devel
oped by a number of individuals, see for example [49]. The idea is to
use an external input with many different frequencies which enriches the
input/output information allowing perfect identification of the system
parameters. As discussed in [1516], indirect adaptive controllers
which require a persistently exciting input are not robust. Since the
adaptive closedloop system is inherently timevarying and nonlinear,
any uncertainty in the plant could counteract the exciting input. The
problem of choosing a persistently exciting input in the presence of
unmodeled plant dynamics has not been completely resolved.
LozanoLeal and Goodwin [21] developed an estimation scheme which
gives nonsingular parameter estimates at each point in time and in the
limit for SISO linear discretetime systems. They modify (when neces
sary) the parameter estimates generated from the leastsquares estima
tion algorithm with data normalization. This modification allows them
xml record header identifier oai:www.uflib.ufl.edu.ufdc:UF0008242500001datestamp 20090216setSpec [UFDC_OAI_SET]metadata oai_dc:dc xmlns:oai_dc http:www.openarchives.orgOAI2.0oai_dc xmlns:dc http:purl.orgdcelements1.1 xmlns:xsi http:www.w3.org2001XMLSchemainstance xsi:schemaLocation http:www.openarchives.orgOAI2.0oai_dc.xsd dc:title New results in indirect adaptive controldc:creator Ossman, Kathleen A. K.dc:publisher Kathleen A. K. Ossmandc:date 1986dc:type Bookdc:identifier http://www.uflib.ufl.edu/ufdc/?b=UF00082425&v=0000117647544 (oclc)000985894 (alephbibnum)dc:source University of Floridadc:language English
44
n(k)ii < nC^ (k) ... J2n(k)]
(57)
where
It follows from property (vi) in Theorem 4.1 that nd,.(k)it converges to
zero as k goes to infinity for i=l 2n. Applying this property and
property (iv) of Theorem 4.1 to (57) gives the desired inequality
(56). Property (56) is very important in the proof of global
stability of the adaptive closedloop system.
Feedback Gain Sequence
The feedback gain sequence L(k) must be chosen so that (F(k)
G(k)L(k)) is exponentially stable as a timevarying system matrix. As
shown in Theorem 5.1, a stabilizing feedback does exist because the
estimated system (F(k),G(k)) generated from the parameter estimator
(42)(45) is uniformly stabilizable under assumptions A1A3. There
are several options in the literature for choosing L(k), some of which
will be discussed in this section.
One approach for computing L(k) is to stabilize the estimated
system (F(k),G(k)) pointwise in time using control law strategies for
timeinvariant systems such as pole placement or LQ control. Of course,
in the case of pole placement, the stabilizability assumption over the
parameter space n (assumption A3) must be strengthened to reachability
5
where
(22)
T(kl) = [y(kl) ... y(kn) u(kl) ... u(kn)].
The vector 0 consists of all the system parameters and (kl) is a
regression vector of past inputs and outputs. Throughout the following
discussion, n 11 will designate the vector Euclidean norm defined by
T ^2
11x11 = (x x) .
Of all the parameter estimation schemes discussed in [6], the
leastsquares algorithm has the fastest convergence rate and is most
easily modified to handle output disturbances and slowly timevarying
systems. The algorithm results from minimizing the quadratic cost
function:
k=l
The cost function consists of the sum of the squares of the prediction
errors plus an additional term which takes initial conditions into
account. The matrix P(0) can be interpreted as a measure of confidence
in the initial parameter estimate 0(0). Designating 0(k) as the esti
mate of the system parameters at time k, the leastsquares algorithm is
described by
Cy(k) eT(kl)(t>(kl)]
P(k) = P(ki) P(kl)4>(kl)/(kl)P(kll.
1 + (kl)
; P(0) = PT(0) > 0.
(23)
69
Figure 71 continued
Estimation Scheme (42)(45)
13
y(k) 0T<>a(kl) + /^(kl)
(33)
The Nxm matrix e contains all of the unknown entries in P while the
matrix tp contains only those entries of P which are known a priori. The
vectors (k1) and 4>. (k1) are regression vectors whose components
come from (kl). Clearly, the decomposition described in (33) is not
unique for multiinput multioutput systems. In order to minimize the
required computations for parameter estimation, the size of (k1)
St
should be made as small as possible. An example will serve to
illustrate these concepts.
Example 1: Consider the linear discretetime system with two outputs
and a single input described by
y(k) =
yi(k)
1 a
'y^kD
+
3
y?(k)
1 2
y2(ki)
1
u(kl)
The parameter a is assumed to be unknown. The system can be rewritten
in the form
y(k) = PT <(>( k1)
where
' 1 a 3'
y^ki)'
4>(kl) =
y2(kD
1 2 1
u(kl)_
A decomposition for this system which minimizes the required
computations for parameter estimation is given by
47
PJe) = Q + ft(0)poo(8)f(0) ft(0)poo(0)g(0)[i + gt(0)pw(0)g(0)]1
X GT(0)Poo(0)F(0).
Let ftg denote the compact parameter space defined in (52) such that
(F(e),G(0)) is pointwise stabilizable for all 0 belonging to ftg. Let
e > 0 be given. It will first be shown that nR^ R^ ^ii < e for k
sufficiently large where R^ is computed using (59).
Step 1: The sequence of matrices R^ given by equation (59) is
bounded. That is, R^ belongs to a compact subset D of ^xN.
T T
Assumption A3 implies (F (8),G ()) is detectable for all 0 belonging
to the parameter space ft. Using arguments similar to those in Theorem
5.1, it follows easily that (FT(k),GT(k)) is uniformly detectable.
Anderson and Moore [27], prove that R^ and L(k) given by (58)(59)
will be bounded if (FT(k) ,G^(k)) is uniformly detectable.
1 2
Step 2: Let Pk(@) and PÂ£(0) denote the sequences generated from the
. 1 2
RDE using initial conditions PQ and PQ respectively. Defining
APk(0) = P^(0) P^(0), it then follows that
Pk+1(e) ft(e)
x(iPk(e) + iPk(0)G(e)[GT(e)Pk (e)G(e) + I]"1 GT(8)aPk(e))
x (I pJ(e)G(9)[GT(8)pJ(6)G(e) + I]'1 GT(0))TF(e).
97
Many individuals, including Kreisselmeier and Narendra [34],
Peterson and Narendra [35], Goodwin and Sin [6], and Samson [36],
consider the problem of maintaining stability of adaptive controllers in
the presence of bounded disturbances. A dead zone is introduced into
either the identification scheme or the adaptive law to "shut off" the
adaptive algorithm whenever the magnitude of the identification error
falls below a certain bound related to the upper bound on the
disturbance. Global stability of the adaptive system can be ensured
only if an upper bound on the disturbance is known.
Narendra and Annaswamy [17] and Anderson [18] propose using an
external input to ensure persistent excitation of the adaptive system.
Robustness will then follow as a consequence of exponential stability of
the closedloop system. However, there is currently no method for
ensuring persistent excitation in the presence of unmodeled plant
dynamics unless the disturbance due to plant uncertainty is bounded.
Ioannou and Kokotovic [19] introduce a amodification (i.e., the
addition of a linear feedback term) to the adaptive law for a model
reference adaptive controller. As in the case of the persistent
excitation approach, stability can be guaranteed only if the disturbance
due to plant uncertainty are bounded.
All of the approaches to robustness discussed so far require the
disturbance to be bounded. A disturbance due to unmodelled plant
dynamics is internally generated and therefore dependant on the system
input and output. Thus, boundedness of the disturbance can only be
assumed if stability of the adaptive closedloop system has already been
established. Kreisselmeier [15] introduces a relative dead zone into
the estimation scheme which depends not on the identification error
53
Step 7: There exists a finite integer Mgte) such that
nPk+1 pkn < e/3 for all k > M0.
The result follows immediately from the uniform continuity of
P () over si* and the convergence of H8(k+1) e(k)u to zero.
00 Â£
Step 8; There exists a finite integer M(e) such that
nRk Rk 1 < e for a11 k >
Applying the triangle inequality and using the results in steps six and
seven gives
k+1 k+1 k k
llRk+N+l Rk+NB < BRk+N+l P + i,p o PJ + "P* Rk+NB < e
for all k > max (M^, M2).
The result then follows easily by setting M = max(M^, M2) + N + 1.
Step 9: If L(k) is computed using (58)(59), then nL(k) L(kl)n
converge to zero as k goes to infinity. This result follows easily
since nR^ R^ ^il, nF(k) F(kl)ll, and nG(k) G(k1) 11 all converge
to zero.
Using the results of Theorem 5.2, it is now possible to prove that
(F(k)G(k)L(k)) is exponentially stable as a timevarying system
matrix. The exponential stability of (F(k)G(k)L(k)) is very important
for proving global stability of the adaptive closedloop system.
11
to prove global stability of an adaptive pole placement scheme without
requiring a persistently exciting input. It is, however, possible for
the parameter estimates to converge to a point that is near a singular
point, in which case the controller gains may be large. Also, a MIMO
version of the algorithm is not available at the present time.
De Larminat [22] has also proposed a parameter estimation scheme
that does not yield singular points. He assumes a priori knowledge of a
space G which contains the actual system parameters and is devoid of
singular points. He then introduces a modification to the estimates
obtained from the standard leastsquares algorithm. The modification is
only required for a finite period of time and produces estimates which
belong to G for all time. Although his modification prevents conver
gence near a singular point, De Larminat's recursive procedure is not as
explicit as the one proposed by LozanoLeal and Goodwin [21].
As in the case of De Larminat, Kreisselmeier [23] assumes prior
information on the system parameters is available. Specifically, he
assumes that the components 9. of the system parameter vector 9 lie
within a known bounded interval [emln, eTax]. He then modifies the
identification scheme for SISO continuoustime systems to force the
parameter estimate of e.. to converge to the set [e?1 n,em^x] for every
i. Kreisselmeier also assumes that for each e with 9.e [e1?1 n,er[,ax]
there is a feedback control system with a prescribed degree of stabil
ity. This condition can be checked a priori since it does not involve
the parameter estimates. With these assumptions, Kreisselmeier proves
global stability of the SISO continuoustime adaptive controller.
54
Theorem 5.3: Suppose e(k) is generated using the parameter estimator
(42)(45) and the feedback gain sequence L(k) is computed using (58)
(59). Then subject to assumptions A1A3, (F(k)G(k)L(k)) is
exponentially stable as a timevarying system matrix.
T T
Proof: Assumption A3 implies that (F (@),G (e)) is detectable for
each 9 belonging to n. Using arguments similar to those in
Proposition 5.1 and Theorem 5.1, it follows that (F^(k),GT(k)) is
uniformly detectable. In [27], Anderson and Moore prove that if
(F^(k),G^(k)) is uniformly detectable then (F(k)G(k)L(k) )^ is
exponentially stable as a timevarying system matrix where L(k) is given
by (58)(59). Since matrix products do not commute, it does not
immediately follow that (F(k)G(k)L(k)) is exponentially stable.
However, Samson and Fuchs [12] show A(k) will be exponentially stable as
a timevarying system matrix if AT(k) is exponentially stable and if
iiA(k)A(k1) n converges to zero as k  . Since (F(k)G(k)L(k))^ is
exponentially stable, it then follows that (F(k)G(k)L(k)) will also be
exponentially stable if u(F(k)G(k)L(k)) (F(kl)G(kl)L(kl))n
converges to zero as k  . Property (vi) of Theorem 4.1 implies both
iiF(k)F(k1) n and nG(k)G(kl) n converge to zero. It was shown in
Theorem 5.2 that nL(k)L(kl)n also converges to zero. Since
F(k)G(k)L(k)(F(kl)G(kl)l(kl)) =
(F(k)F(kl)) G(k)(L(k)L(kl)) (G(k)(G(kl))L(kl)
and both G(k) and L(k) are bounded, it follows that
n(F(k)G(k)L(k))(F(kl)G(kl)L(kl))n does indeed converge to
zero. Thus the matrix (F(k)G(k)L(k)) is exponentially stable as a
timevarying system matrix.
94
u1(k)
: A A
!.*.! . 1 .
V L
'j 10 20
30
40
50
60
70 (k)
1 i
: Â¥
(c)
u2(k)
10:
5
* i\
1j* * ir Kit y r tL iw 1 ,L liMimim 1 1 1 i 
V V0 20 30 40 50 60 70 (k)
5 ;
10
(d)
Figure 710 continued
Estimation Scheme (42)(45)
83
Table 74
DATA
"al
"a2
bl
b2
Actual Parameters
2
0.99
0.5
3.0
Initial Estimates
1.95
1.0
0.75
5.0
Parameter Ranges
[1.9, 2.0]
11
o
*
o
CM
1
1 1
[0.5, 2.0]
[2.5, 7.5]
Estimation Scheme
(42)(45)
1.91
0.920
0.406
2.83
52
Clearly, there exists a cr^e) such that if nzk+. II < ^ for all ie[0,N]
then
nQ
k
k+N
11 < e/6.
Let S(o^) be chosen such that
nsk+iH < Oj for all ie[0,N] whenever ne(k+i) e(k) n < 6.
It follows from properties (v) and (vi) of Theorem 4.1 that there exists
a finite integer such that
e(k)ei2g for all k >
and ne(k+i) e(k)n < 6(a^) for all ie [0,N] and for all k > M^.
Therefore
"Qk+N" = Rk+N pk+N11 < e/6 for 311 k > 'V
Using the results in step four, we obtain
aPk+N pn < fr a^ k > Mi*
Thus
Rk+N P11 < 11Rk+N Pk+N11 + 11Pk+N P11 < e^3 for a11 k > Mi*
which proves the result.
65
control law u(k) = LjX(k)L2xc(k) is exponentially stable and converges
to a timeinvariant system. Finally, from the results in Chen [20], if
the control law (65) is applied to the plant (61), the resulting
closedloop system is still stable (not including the exogenous
disturbance v(k)) and the system output y(k) converges to the reference
signal r(k) as k * .
The adaptive controller described by (63) and (65)(66) causes
the plant (61) to reject deterministic disturbances and to track a
given reference signal r(k). Simulations of the adaptive controller for
various types of external disturbance and reference signals are included
in Chapter VII.
86
y(k)
40
30
20
10
10
20
30
40
20 30 40 50 60 TOO TO
(a)
u(k)
Figure 78
Exact Parameters
NEW RESULTS IN INDIRECT ADAPTIVE CONTROL
By
KATHLEEN A. K. OSSMAN
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN
PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
1986
93
yx(k)
y200
Figure 710
Estimation Scheme (42)(45)
CHAPTER I
INTRODUCTION
Adaptive control, the problem of controlling a system whose
parameters are unknown prior to or changing during system operation, has
been a major research topic during the past three decades. Although
originally intended for timevarying or nonlinear systems, most of the
stability results on adaptive controllers have been limited to linear
timeinvariant systems. The majority of adaptive controllers can be
classified as either direct or indirect.
In direct adaptive control, no attempt is made to estimate the
unknown system parameters. Instead, the controller parameters are
updated directly using real time input/output information. One specific
example of direct adaptive control is model reference adaptive control
(MRAC) where the unknown plant is forced to behave asymptotically like
some prechosen reference model. Several individuals have developed
this approach, see for example [13]. The assumptions needed to prove
global stability of model reference adaptive controllers are (1) the
unknown plant is minimum phase, (2) the relative degree of the plant is
known and (3) the sign of the plant gain is known.
Indirect adaptive controllers are applicable to nonminimum phase as
well as minimum phase systems. In indirect adaptive control, the
unknown system parameters are estimated in real time using available
input/output information then the control law is computed using the most
recent parameter estimates. A number of individuals have introduced
indirect adaptive controllers, see for example [412]. The assumptions
1
CHAPTER IV
PARAMETER ESTIMATION
Consider the rinput moutput linear discretetime system described
by
y(k) = 9%a(kl) + /^(kl)
(41)
As previously discussed in Chapter III, it is assumed that each compo
nent 8.. of the unknown system parameter matrix 9 belongs to a known
bounded interval [9*1n, 9II,^X].
L ij 5 ij J
The parameter estimation algorithm is given by
P(klU (k1)
9(k) = 9(k1) P(kl)f(9(kl)) += = ^ x (42)
\ 1 + ^(kl)P(kl)* (k1)
[yT(k) <(>J(kl)^^(kl)e(kl)]
P(k) = P(kl) 
P(klHa(kl)^(kl)P(kl)
\l + Â¡(klWkD^kl)'
0 < P(0) = PT(0) < 21,
(43)
'ij(k
max
ij
when 9. .(k1) > 9^x
fij(kl) = ^
utki)
 0I!'in
IJ
when 9  (k1) < g?!0
* \J V
(44)
0
when 9. .(k1) e[9^n, 9x]
18
CHAPTER VIII
DISCUSSION
A globally stable adaptive LQ controller which does not require
persistent excitation was introduced for multiinput multioutput linear
discretetime systems. The assumptions made on the plant were (1) an
upper bound on system order is known, (2) the unknown system parameters
belong to known bounded intervals, and (3) the plant is stabilizable for
all values of the unknown parameters ranging over the known bounded
intervals. When applied to the unknown plant using this parameter
estimator, the adaptive LQ controller ensures the system inputs and
outputs will remain bounded, and forces the output to track a given
reference signal in the presence of a deterministic external
disturbance. Some remaining considerations include robustness,
application to timevarying or nonlinear systems, and analysis of
transient response.
In many control applications, the order of the model will be lower
than that of the plant. As mentioned in the introduction, Rohrs et al.
[1314] demonstrated that most adaptive controllers could go unstable if
the order of the system was underestimated, even if the modelling errors
were small. Since knowledge of an upper bound on system order is one of
the assumptions made on the adaptive controller presented in this
dissertation, it is likely that this controller will also go unstable
for systems with unmodelled dynamics. There have been several
approaches in the literature towards development of robust adaptive
controllers, some of which will be briefly discussed here.
96
84
y(k)
40
30
20
10
10
20
30
40
I I
'i 11
I It
\ .10 l i 20
30 40 50 60 70 (k)
(a)
Figure 77
Estimation Scheme (42)(45)
74
Table 72
DATA
"al
a2
a3
bl
J
b2
b3
Actual Parameters
2.6
2.13
0.54
1.0
1.5
1.0
Initial Estimates
2.65
1.0
0.5
2.0
1.6
1.5
Parameter Ranges
[2.5,2.8]
[5,3]
[2,1]
[4,8]
[1.5,1.7]
[3,0]
Estimation Scheme
(42)(45)
2.5
1.92
0.429
0.999
i
!
1.60
0.795
9
An example of a parameter estimator possessing these properties is the
leastsquares estimator discussed previously. Clearly, properties (i)
and (ii) listed above are equivalent to properties (i) and (ii) given
for the leastsquares algorithm. It is not quite as obvious that Samson
and Fuchs' property (iii) is equivalent to property (iii) of the least
squares algorithm. The proof is given in Chapter IV.
The second assumption for proving global stability is that the
estimated system (F(k),G(k)) must be uniformly stabilizable.
Definition: The system (F(k),G(k)) is uniformly stabilizable if there
exists an integer r > 1, a constant q and a uniformly bounded sequence
L(k) such that
t+r1
l H (F(k)G(k)L(k))I
k=t
where
t+r1
n (F(k)G(k)L(k)) = (F(t+r) G(t+r)L(t+r))x
k=t
(F(t+r1) G(t+rl)L(t+rl)) ... (F(t) G(t)L(t)).
Theorem: Assuming the system (F,G) is stabilizable, the parameter
estimator possesses the three required properties, and the estimated
system (F(k),G(k)) is uniformly stabilizable, the control law u(k) =
L(k)x(k) described by (25)(27) will result in a globally stable
closedloop system in the sense that the system input and output
converge to zero for any initial states in the plant and observer
(Samson and Fuchs [12]).
46
In (59), both 0 and the initial value R0 are positive definite
symmetric matrices. In contrast to the first approach, where the
algebraic Riccati equation would be solved for each k, this asymptotic
approach requires the solution of only one iteration of the Riccati
difference equation at each point in time. Obviously, this approach
offers considerable savings in online calculations. Also, it is not
necessary to check whether or not (F(k),G(k)) is stabilizable at each
point k because L(k) given by (58)(59) is welldefined even for
isolated singular points. In order to prove that (F(k)G(k)L(k)) is
exponentially stable when L(k) is computed using (58)(59), it is
first necessary to show n(L(k)L(kl))n converges to zero as k + <.
Theorem 5,2: Suppose the parameter estimates are generated using the
algorithm (42)(45). Then subject to assumptions A1A3, the feedback
gain sequence given by (58)(59) has the property
nL(k)L(kl)n + 0 as k + .
Proof: The proof is rather lengthy and will therefore be divided into
steps. Throughout the proof, Pk(9) will refer to the sequence generated
from the timeinvariant Riccati difference equation (RDE) given by
Pk+1(e) = Q + FT(e)Pk(e)F(e) FT(e)Pk(9)G(e)[i + GT(9)Pk(e)G(0)]1
x GT(9)Pk(0)F(9), PQ > o.
The matrix P^e) will denote the solution to the algebraic Riccati
equation (ARE) given by
ACKNOWLEDGEMENTS
I would like to acknowledge the following people for their
contributions to the completion of this dissertation. Special thanks go
to my advisor, Dr. Edward W. Kamen, for his numerous helpful suggestions
and creative ideas. Thanks also go to the members of my supervisory
committee: Dr. T. E. Bullock, Dr. D. W. Hearn, Dr. P. Z. Peebles, and
Dr. S. Svoronos.
This work was supported in part by the U. S. Army Research Office,
Research Triangle Park, N.C., under Contract No. DAAG2984K0081.
ii
33
Also, there exists a finite N such that the matrix
[2IP(kl)P"*(k)P(kl)] is positive definite for k > N because
[2IP(kl)P"l(k)P(kl)] converges to the positive definite matrix 21
PM. Since
00
00
0 < l xmin[2IP(kl)P"1(k)P(k)P(kl)]Tr[fT(e(kl))f(0(kl))]
k=N min
00
< I Tr[fT(0(kl))[2IP(kl)P1(k)P(kl)]f(0(kl))] <
k=N
00
it follows that l Tr[f^(0(kl))f(0(kl))] < . Therefore, the right
k=l
hand side of the inequality (425) is bounded in the limit as M ap
proaches infinity. This implies the existence of a finite L such that
00
l Ilf(0(k1)) ii < L < . Returning to the inequality (423), let
k=l
e > 0 be given. The sequence 0^j(k) will be a Cauchy sequence if
10^j(m)0Â¡j(n)  is less than e for m and n sufficiently large. Since
P(k) converges to P^, there exists N^(e) such that
nP(m)P(n)n < for m,n > N^(e)
where Â£l = (1/3) min [e/(nP"1(O)0(O) ii e/(vL)].
Also, there exists N2U) such that
m1
l nf(e(i))n < e/6v for m,n > N?(e).
i=n
CHAPTER III
SYSTEM DEFINITIONS AND ASSUMPTIONS
The system to be regulated is the multiinput multioutput linear
discretetime system described by
yOO ? A.y(kj) + ? B,u(kj) (31)
3=1 J 3=1 J
In (31), y(k) is the mxl output vector and u(k) is the rxl control
input vector. It is assumed that an upper bound, n, on p and q is known
but all or some of the entries in the matrices Aj and Bj are unknown.
The system described by (31) can be rewritten in the following form
convenient for parameter estimation:
y(k) = PT$(kl) (32)
where
PT = CA]_ ... An ... Bn]
4>T(k1) = [yT(kl) ... yT(kn) uT(kl) ... uT(kn)]
The n(m+r)xm matrix P consists of all the system parameters and the
n(m+r)xl vector (kl) is a regression vector of past inputs and
outputs. In certain applications, some of the entries in the system
matrix P will be known a priori. Since it is not necessary to estimate
known parameters, a scheme for separating the known parameters from the
unknown parameters is advantageous. This is accomplished by rewriting
(32) in the following form:
12
22
where
O
O
1
O
 ith position.
0
Each of the terms on the right hand side of the equation (47) must
converge which implies Pjj(k) converges for every i and j. Noting that
P(0) < 21, it then follows that P(k) must converge to a positive semi
definite matrix P^ < 21. To prove (ii), let 'e(k) = 0(k)e. Rewriting
(42) in terms of e(k) gives
p(klHa(kl)^(kl)e(kl)
ei(k) = 'e(kl) 
P(kl)f(e(kl)). (48)
\l + Â¡(kDP(kl)*a(kl)
Multiplying both sides of the equation (43) on the right by P1(kl)
gives
P(k)P1(k1) = I 
P(kl)
(49)
\l + 4(k~1)P{k'1K(k'1)
Combining (48) and (49) yields
e(k) = P(k)P1(kl)e(kl) p(kl)f(e(kl))
(410)
81
yflO
u(k)
Figure 76
Exact Parameters
32
Rearranging terms in this inequality results in
II c \ Pit / u gq
Vl + ^(kl)P(kl)+a(kl)
< V(kl) V(k) + X^C^DP'1
N m
I l
i=l j=l
fij(0(kD)
(k)P(kl)]Tr[fT(0(kl))f(9(kl))].
Summing all the terms in this inequality from k equals one to M and
N m
noting that nf(e(kl))n < l l f.Â¡_.(e(kl)) gives
1=1 j=l J
M
0 < I
ne(k) ii*
M
2 t + 2q l llf(e(kl))
k=1 nkl + *;(k1)P(k1)*a(c1) k=1
M
(425)
< V(0) V(M) + l AmaxCP(kl)P1(k)P(kl)]TrCfT(0(kl))f(9(kl))].
k=l
If the right hand side of the inequality (425) is bounded in the limit
as M approaches infinity, then the norm of f(e(k1)) will be summable.
The term V(M) is bounded since V(k) converges. Also
AmaxCPikDP^OOPUl)]
is bounded since the matrix
P(kl)P1(k)P(kl)
converges to P^. It remains to be shown that
l Tr[fT(e(kl))f(e(kl))] < ..
k=l
Equation (417) implies
l Tr[fT(0(kl))[2IP(kl)P'1(k)P(kl)]f(e(kl))] <
k=l
80
8
6
4
2
2
4
6
8
. i
2U 40"
 > 1 .
60 bo too no
(d)
tÂ£{k)
bj(k)
w
Figure 75 continued
Estimation Scheme (42)(45)
60
closedloop system is globally stable using the adaptive regulator
defined by (53) and (510).
28
By properties (iii) and (v), the first and third terms on the right side
of the inequality (420) converge to zero. The second term on the right
side of the inequality will also converge to zero by property (v) pro
vided that the matrix
P(kl)2*a(kl)^(kl)
Vl + *Â¡(kl)P(klHa(kl)
is bounded. This matrix is shown to be bounded by first noting that
0 <
P2(kl)a(kl)^(kl)
\l + +Â¡(kl)P(klHa(kl)
2 UklH'ikl)
P(k1) a a
iia(kl)ii<
'k1
iia(kl)
y+ AminP(kl)
,2 min
Clearly, the numerator on the right hand side of the inequality is
bounded above. The denominator is bounded below by one if the deter
minant of P(k) converges to zero, and by x (P ) if the determinant
min o
of P(k) does not converge to zero. Thus, all three terms on the right
hand side of the inequality (420) converge to zero. Since
Tr[(e(k) e(kl))T(0(k) e(kl))] = J J [e..(k) e^kl)]2,
1=1 j=l u ,J
it follows that 0. .(k) e..(k1) must converge to zero which proves
J 1 J
property (vi) in the case where p =1. The proof for arbitrary p fol
lows from the Schwarz inequality. To prove (vii), equation (410) is
first solved recursively giving
e(n) = ?(n)pl(0)Q(0) P(n)
l P"1(i+l)P(i)f(0(i))
i=0
72
y(k)
20;
15
lo
st
5
10
15
20
tit
4 nJJ
iV'20
30
40
50
60
70 (k)
(a)
Figure 73
Estimation Scheme (42)(45)
with nk= max(1, (k) )
89
yx(k)
y2(k)
88
Table 75
DATA
a
b
Actual Parameters
4.0
1.0
Initial Estimates
6.0
1.0
Parameter Ranges
[2.0, 10.0]
11
o
CM
v
o
I
1 I
Estimation Scheme
(42)(45)
3.981
1.001
34
It then follows from (423) that e.^(m) 0^(n) < s for m, n >
max (Nj,^). Combining this with property (v), e^j(k) converges to a
point in [9.. ,0.. ] for i=l, .... N and j=l, ..., m.
I J 1 J
As discussed in Chapter II, properties (ii)(iv) and (vi) are
standard properties required for any parameter estimator used in in
direct adaptive control. Property (v) forces the parameters to converge
to a subset 0 of ]R^xm which contains no singular points. This addi
tional property will ensure that the estimated system is uniformly
stabilizable thus eliminating the need for a persistently exciting
input. Property (vii), the convergence of 0Â¡j(k) to a point in
[9?]n9^X] is not required for proving global stability of the
adaptive closedloop system. Property (vii) will be used in Chapter VI
for tracking and disturbance rejection based on the internal model
principle. It is interesting that unlike the case of least squares
where e(k) converges to 0 + PooP"1(O)'0(O) as k  , there is no closed
form expression for the limit of 0(k) as k > when using the param
eter estimator described by (42)(45). However, if P(k) converges to
zero, then e(k) converges to the true system parameter matrix e for both
the parameter estimator discussed in this chapter and the recursive
leastsquares estimator. This is easily shown using the recursive
solution of (410) which is given by
?(n) P(n>p1{0)'e(0) P(n) J P'1 (1+l)P(l)f(e(1)).
i=Q
Taking the norm of both sides and using the properties of the induced
matrix norm, it follows that
2'
used to prove global stability of indirect adaptive controllers are (1)
an upper bound on system order is known, (2) the estimated system is
uniformly stabilizable, and (3) the parameter estimator possesses
certain key properties. Since the adaptive controller developed in this
dissertation can be classified as indirect, each of these assumptions
will be discussed in more detail.
The importance of the first assumption was illustrated in [1314]
by Rohrs and associates who investigated the behavior of adaptive
controllers in the presence of unmodeled dynamics. Rohrs demonstrated
through computer simulations that all existing adaptive controllers
could go unstable if the order of the system was underestimated. In
response to Rohrs' findings, several individuals [1519] have achieved
promising results in the development of robust adaptive controllers.
Verification of the second assumption, which requires the existence
of a stabilizing feedback control law for the estimated system, is
difficult because it depends on the parameter estimates which are
generated in real time. Unless something more can be said about where
the parameter estimates are going, the assumption cannot be verified a
priori and, consequently, global stability cannot be ensured. The first
approach towards satisfying this assumption was the use of a
persistently exciting external input which allows perfect identification
of the plant. This approach along with more recent results which do not
require persistent excitation are discussed in more detail in Chapter
II.
The third assumption requires that the parameter estimator possess
certain key properties which are explained in detail in Chapter II. It
will suffice to mention the existence of several parameter estimation
55
Stability of the Adaptive ClosedLoop System
The control law chosen to regulate the system is given by
u(k) = L(k)x(k) (510)
where x(k) is generated from the adaptive observer (53) and L(K) is a
stabilizing feedback for the estimated system (F(k),G(k)). The
following theorem shows that the adaptive regulator consisting of the
observer (53) and the control law (510) results in a globally stable
closedloop system; that is, for any initial states in the plant and the
observer, the input u(k) and the output y(k) converge to zero.
Theorem 5.4: Suppose that the parameter estimator (42)(45) is used
so that there is a stabilizing feedback L(k) for the estimated system
(F(k),G(k)) subject to assumptions A1A3. Then with the adaptive
regulator defined by (53) and (510), the resulting closedloop system
is globally stable.
Proof: The proof is based on a MIMO extension of the results in [12].
Letting Â£(k) = y(k)y(k) and using equations (53) and (510) gives
x(k+l) = (F(k)G(k)L(k))x(k) M(k)e(k)
y(k) = Hx(k) e(k).
48
This relationship was introduced by Samson [10] in the singleinput
case. It can also be verified in the multiinput case, but due to the
large amount of algebra involved the derivation will not be included
here.
Step 3: For any Pg > 0 there exists an N(e,Pg) such that
nPu(0) P ()H < e/6 for all eeiU and for all k > N.
Since (F(0),G(0)) is stabilizable for each 0 belonging to flg, it follows
from wellknown results that P^(e) converges to P^e) pointwise over
Â£1a. It was shown in Green and Kamen ([33], Theorem 1) that the
convergence is uniform over Qg if is compact.
Step 4: For all Pg belonging to a compact subset D of there
exists an N(e) such that
iPM(e) P (0) n < e/6 for all eesu.
n <*> e
The compactness of 0 implies there exist matrices P and Pg such that
Pg < Pg < Pg for all Pg belonging to 0. It follows from the
relationship given in step two that
Pk (9) < Pk(9) < Pk(e) for all PQeD.
...
From step 3, there exist integers N1(e,Pg) and N2(e,Pg) such that
iiP1^() P (0)n < e/6 for all k > N. and for all 0efiA;
K 00 i e
nPk(e) Pje) 11 < e/6 for all k > N2 and for all OeSU.
62
As mentioned, the controller design will be based on the internal
model principle. Assume that the disturbance vector v(k) and the
reference signal r(k) both satisfy the difference equation:
z(k+q) + <* iZ(k+ql) + ... + ^(k+l) + a^z(k) = 0. (62)
Let r be defined as
0 1
ao ~al ~aql
The internal model can then be realized as
xc(k+l) = Acxc(k) + Bce(k)
where Ac = block diagonal {r, r, ..., r}
mtuple
0
0
1
Bc = block diagonal {T, T, ..., T} with t =
mtuple
(63)
The following theorem gives the conditions under which it is possible to
design an adaptive controller which will stabilize the plant described
16
rank [D(z,e) N(z,0)] = m for z < 1 and all 0efl.
where m is the number of system outputs and
n n .
D(z,e) = I + l A.zJ, N(z,0) = l B.zJ.
j1 J j=l J
As previously mentioned, this stabilizability condition can be checked
since it does not depend on the parameter estimates. An example will
illustrate the test.
Example 2: Consider the system described in Example 1.
[D(z,e) N(z,e)] =
1z
z
az 3z
l2z z
In order to satisfy the stabilizability assumption, the rank of
[D(z,0) N(z,0)] must equal 2 for z < 1 and 0en. The determinants of
the 2x2 submatrices of [D(z,0) N(z,0)] are (2+a)z2 3z+l, z((6a)z3),
z(l4z). All three minors are zero only when z = V4 and a = 6. Hence
the rank condition is violated only when z = V4 and a = 6. Therefore,
the stabilizability assumption will be satisfied for any interval which
does not contain the point {6}.
The amount of effort involved in checking the rank condition in
(34) will of course depend on the number of unknown parameters. It may
be possible to use a rootfinding algorithm for polynomials in several
variables (i.e., z and 0ij) to check the rank of [D(z,0) N(z,e)] for all
0eG. It should also be noted that by overestimating the max (p,q) in
(31), the system cannot be controllable, but may be stabilizable. In
17
this case, adaptive pole placement would not be possible, but a
stabilizing feedback could be computed using the Riccati approach
discussed in Chapter II.
To summarize, the assumptions on the rinput moutput linear
discretetime system described by (31) (33) are
Al:
A2:
A3:
An upper bound, n, on p and q in (31) is known;
the components eij of the unknown system parameter matrix e belong
to known bounded intervals [e^Q^];
the system described by (33) is stabilizable for all e belonging
to n where n = {ee RNxm: e [e^.n ,9^x]}.
* J J J
49
Thus,
llP.(0) P (9)11 < e/6 for all QeSU, for all PfteD,
K E U
and for all k > maxCN^.Ng).
Step 5: Given a > 0, there exists a 6(a) such that for all P^eD and
for all e,0eÂ£U
e
Pk+1(9) Pk+1^^11 < 0 whenever ne en < 6
This property follows easily from the fact that Pk+1(@) is a continuous
function of 8 and the sets D and fi* are compact.
Step 6: Let N be a fixed finite integer such that the condition in step
four is satisfied. There exists a finite integer Mj(e) such that
nRk+N P11 < for all k >
1/
where P is the solution to the ARE when the matrices F() and G(*)
00
are evaluated at e(k).
In order to prove this result, a family of sequences pÂ£ is defined as
fol1ows
Pk = Rk ^or k < t.
Pk+1 = Q+FT(o(t))PkF(e(t:))FT(0(t))P^ G(0(t))[I + ST(0(t))PÂ¡jG(0(t))]1
x GT(9(t))P^F(9(t)) for all k > t.
23
Now define the Lyapunov functional
V(k) = TrCeT(k)P1(k)^(k)) > 0,
where "Tr" denotes the trace operation.
Inserting equation (410) into the expression for V(k) gives
V(k) = Tr [0'T(kl)P"1(kl)/e(k)] Tr[fT(e(kl))P(kl)P1(k)e(k)].
(411)
Inserting the expression (48) for e(k) in the first term on the right
hand side of (411) yields
V(k) = V(kl) 5 TSSUilJ
Vl + +a(k1)p(k1)*a(k1)
Tr [0T(kl)f(e(kl))]
Tr[fT(e(kl))P(kl)P1(k)e'(k)]. (412)
Equation (410) for e(k) is then inserted into the last term on the
right hand side of (412) giving
V(k) = V(kl) 
ne(k) ii2
Vl + Â¡(kl)P(kl)+a(kl)
2Tr[0T(kl)f(0(kl))]
(413)
It follows from the definition of f(0(kl)) in equation (44) that
Tr[fT(0(kl))ff(kl)] > Tr[fT(9(kl))f(0(kl))].
(414)
70
y(k)
U(k)
20
15
10
5
5
10
15
20
:
1
1
i A
: "'V
20
30
40
(b)
Q
gg
ft VT
Figure 72
Recursive Least Squares
98
alone but on the identification error normalized by the signals in the
adaptive system. In order to prove global stability of the adaptive
system, Kreisselmeier assumes that the unknown system parameters e.
belong to known bounded intervals ["T11 n,9r?axl. He also assumes that
the modelling error meets a certain growth criterion which will not be
discussed here. It will suffice to mention that Kreisselmeier gives
some good examples of types of plant uncertainties which meet his
criterion. By incorporating the relative dead zone into his
identification scheme, Kreisselmeier is able to prove global stability
of an indirect adaptive controller for SISO continuoustime systems with
modelling errors. Although Kreisselmeier's identification scheme
differs from the parameter estimator introduced in Chapter Four, it may
be possible to introduce a relative dead zone into the estimation scheme
in order to develop a robust adaptive controller for MIMO discretetime
systems.
Adaptive control was originally intended for timevarying and/or
nonlinear systems; however, most of the results have been restricted to
linear timeinvariant systems. Goodwin and Sin [6] make several
suggestions on modifying the leastsquares algorithm for timevarying
systems: (1) addition of a positive term to the covariance P(k), (2) a
finite data window, and (3) exponential data weighting. In the finite
data window, the old data are discarded periodically by some chosen
pattern. In exponential data weighting, a forgetting factor is
introduced into the covariance equation which has the effect of
weighting the most recent data heavier than older data. All three
methods preserve the fast initial convergence of least squares while
preventing the covariance from converging to zero which would cause
CHAPTER VII
SIMULATIONS
This chapter contains several simulations of both the adaptive
regulator derived in Chapter V and the adaptive controller discussed in
Chapter Six applied to SISO as well as MIMO discretetime systems.
Example 7.1: Consider "the nonminimum phase discretetime system
described by the following difference equation
y(k) = 2y(kl)0.99y(k2)+0.5u(kl)+3u(k2).
The adaptive regulator defined by the observer (53) and the control law
(510) with L(k) computed using (58)(59) was implemented. Three
different algorithms were used to estimate the system parameters: the
estimation scheme described by (42)(45), the recursive leastsquares
algorithm defined by (23), and the estimation scheme given by (42)
(45) with data normalization (i.e., = max(l, H(j>(k1) n) for all
k). In all three cases, the initial "covariance" P(0) was chosen to be
I and the initial state of the plant was [1 0]T. Also, Q and RQ in
(59) were chosen to be I. The parameter ranges, initial parameter
estimates and steady state estimates for all three estimation schemes
are displayed in Table 71. As seen from Table 71, the estimation
scheme described by (42)(45) forces the estimates of the parameters
to converge to the given ranges; whereas, recursive least squares does
66

