2236 IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 45, NO. 12, DECEMBER 2000
Output Regulation for Linear Distributed
Parameter Systems
Christopher I. Byrnes, Fellow, IEEE, István G. Laukó, David S. Gilliam, Member, IEEE, and Victor I. Shubov
Abstract—This work extends the geometric theory of output
regulation to linear distributed parameter systems with bounded
input and output operator, in the case when the reference signal
and disturbances are generated by a finite dimensional exogenous
system. In particular, it is shown that the full state feedback and
error feedback regulator problems are solvable, under the stan-
dard assumptions of stabilizability and detectability, if and only
if a pair of regulator equations is solvable. For linear distributed
parameter systems this represents an extension of the geometric
theory of output regulation developed in [10] and [4]. We also pro-
vide simple criteria for solvability of the regulator equations based
on the eigenvalues of the exosystem and the system transfer func-
tion. Examples are given of periodic tracking, set point control,
and disturbance attenuation for parabolic systems and periodic
tracking for a damped hyperbolic system.
Index Terms—Bounded input and output operators, distributed
parameter systems, regulator problem.
I. INTRODUCTION
O
NE of the central problems in control theory is to control
a fixed plant so that its output tracks a reference signal
(and/or rejects a disturbance) produced by an external generator
or exogenous system. Generally two versions of this problem
are considered. In the first, the state feedback regulator problem,
the controller is provided with full information of the state of
the plant and exosystem. For the second, and perhaps more re-
alistic error feedback regulator problem, only the components
of the error are available for measurement. For linear finite-di-
mensional systems it has been shown by Francis [10] that the
solvability of the regulator problem is equivalent to the solv-
ability of a pair of linear matrix Sylvester equations. This in turn
can be characterized as a property of the transmission polyno-
mials of the composite system formed from the plant and the
exosystem, as was shown by Hautus [12]. Francis and Wonham
[11] have also shown that any regulator that solves the error
feedback problem has to incorporate a model of the exogenous
system generating the reference signal which is to be tracked
and/or the disturbance that must be rejected. This property is
known as the internal model principle.
Manuscript received September 18, 1998; revised September 20, 1999. Rec-
ommended by Associate Editor, I. Lasiecka. This work was supported in part
by grants from AFOSR and Texas ARP.
C. I. Byrnes is with the Department of Systems Science and Mathematics,
Washington University, St. Louis, MO 63130 USA.
I. G. Laukó is with the Center for Research in Scientific Computation, North
Carolina State University, Raleigh, NC 27695-8205 USA.
D. S. Gilliam and V. I. Shubov are with the Department of Mathematics and
Statistics, TexasTech University, Lubbock, TX 79409 USA.
Publisher Item Identifier S 0018-9286(00)10665-8.
Similar results have been established for finite dimensional
nonlinear systems in [4] in case the plant is exponentially stabi-
lizable and the exosystem has bounded trajectories that do not
trivially converge to zero. In particular, it is shown in [4] that
the solvability conditions given by Francis in the linear case
can be naturally generalized to the solvability of a pair of non-
linear equations, the regulator equations. Geometrically, these
nonlinear regulator equations express the existence of a local
manifold on which the actual and reference outputs coincide and
which can be rendered invariant using feedback.
In this paper we develop the geometric methods introduced in
[10] and [4] for solving the state and output feedback regulator
problems for infinite-dimensional linear control systems, as-
suming that the control and observation operators are bounded.
We expect to have more to say about the unbounded case in
future papers. In particular we derive the regulator equations
for a class of distributed parameters systems, obtaining an op-
erator Sylvester equation. We also obtain results characterizing
the solvability of both state and error feedback regulator prob-
lems in terms of solvability of these regulator equations. The
main difficulties that arise in developing a geometric theory for
the distributed parameter case are obvious: the phase space is
infinite dimensional; the state operator is typically unbounded
and consequently only densely defined; the error zeroing con-
trolled invariant subspace must be contained in the domain of
the state operator and the resulting regulator equations may be-
come distributed parameter equations. Section II contains a for-
mulation of the state and error feedback regulator problems and
a discussion of the basic assumptions. In Section III, we present
a motivating example of periodic tracking for a controlled heat
equation. In Section IV we present our main solvability results
in Theorems IV.1 and IV.2. In particular, we give necessary and
sufficient conditions for the solvability of both the state and
error feedback regulator problems in terms of the solvability of
a pair of linear operator equations, the regulator equations. Once
a solution of these equations is available, the appropriate state
feedback control in the case of the full information problem, or
a dynamic controller in the case of error feedback, are obtained
which provide a solution to the regulator problem.
In Section V, we recall the definitions of transmission and
invariant zeros. For certain classes of distributed parameter sys-
tems we provide a sequence of results, under various hypotheses
on the plant and exosystem, expressing necessary and sufficient
conditions for solvability of the requlator equations in terms of
nonresonance conditions involving the eigenvalues of the ex-
osystem and the plants transmission (or invariant) zeros.
Section VI contains several explicit examples of output reg-
ulation. In the first example we consider a problem of set point
0018–9286/00$10.00 © 2000 IEEE