[13] M. Van Nieuwstadt, M. Rathinam, and R. M. Murray, Differential atness and absolute equivalence of nonlinear control systems, SIAM J.
Control Optim., vol. 36, no. 4, pp. 12251239, 1998.
[14] A. Teel and L. Praly, Global stabilizability and observability imply
semi-global stabilizability by output feedback, Syst. Control Lett., vol.
22, pp. 313325, 1994.
[15] F. Esfandiari and H. Khalil, Output feedback stabilization of fully linearizable systems, Int. J. Control, vol. 56, no. 5, pp. 10071037, 1992.
[16] A. Atassi and H. Khalil, A separation principle for the stabilization of
a class of nonlinear systems, IEEE Trans. Autom. Control, vol. 44, pp.
16721687, Sep. 1999.
[17] L. Consolini and M. Maggiore, Robust output feedback tracking with
a matching condition, presented at the 42nd Conf. Decision Control,
Maui, HI, Dec. 2003.
[18] M. Vidyasagar, On the stabilization of nonlinear systems using state
detection, IEEE Trans. Autom. Control, vol. AC-25, pp. 504509, Mar.
1980.
I. INTRODUCTION
A systematic framework for tackling the robust output regulation
problem for general uncertain nonlinear systems was given in [9],
which was evolved from the approach in [1] which deals with the
local robust output regulation problem by output feedback control.
This framework addresses the robust output regulation problem in
two steps. First, convert the robust output regulation problem for a
given system into a robust stabilization problem of an augmented
system. The augmented system consists of the given system and a
class of appropriately dened dynamic systems called internal model
candidates. Second, solve the robust stabilization problem of the
augmented system. The accomplishment of the rst step depends on
the existence of what is called the steady-state generator in [9] which
is a special internal model candidate. A linear steady-state generator
always exists if the solution of the regulator equations is a polynomial in the exogenous signal. However, the polynomial assumption
is restrictive as it essentially limits the nonlinear systems to be of
Manuscript received May 7, 2003; revised June 27, 2004. Recommended by
Associate Editor Hua Wang. This work was supported in part by the Hong Kong
Research Grant Council under Grant CUHK 4168/03E and by the National Natural Science Foundation of China under Grant 60374038.
The authors are with the Department of Automation and Computer-Aided
Engineering, The Chinese University of Hong Kong, Shatin, N.T., Hong Kong
(e-mail: zychen@acae.cuhk.edu.hk; jhuang@acae.cuhk.edu.hk).
Digital Object Identier 10.1109/TAC.2004.841125
v_ = A1 v v(0) = v0
(2)
where A1 is a matrix with all eigenvalues simple with zero real parts.
It is assumed that all the functions in the setup are sufciently smooth
and D1 (0; 0) = 0, D2 (0; 0) = 0, and q (0; 0) = 0.
Briey, the global robust output regulation problem aims to design a
state or output feedback control law such that, for all uncertain parameter w 2 W <N and all exogenous signals v (t) 2 V <q with
W and V being any known compact sets of the respective Euclidian
spaces, the solution of the closed-loop system starting from any initial state of the plant and the controller exists and is bounded, and the
tracking error approaches 0 asymptotically.
To study the solvability conditions for this problem, let us rst state
some standard assumptions as follows.
A1)
@z(v; w) A v = F (w)z(v; w)
1
@v
+G (q(v; w); v; w) q(v; w) + D1 (v; w):
(3)
All the eigenvalues of F (w) have negative real parts for all
w 2 <N .
A3) b(w) > 0 for all w 2 <N .
Remark 2.1: Under assumptions A1) and A3), let
A2)
y(v; w) = q(v; w)
41 (v; w) = 1 @y(v; w) A1 v 0 H(w)z(v; w)
b(w)
@v
0K (y(v; w); v; w) y(v; w) 0 D2 (v; w))
(4)
pairwise
coprime
polynomials
with r1 ; . . . ; rI being the degrees
of their minimal zeroing polynomials P1 (s); . . . ; PI (s),
and sufciently smooth function 01 : <r +111+r 7! <
vanishing at the origin such that
41 (v; w)) = 01 1 (v; w); _ 1 (v; w); . . . ; 1(r 01) (v; w);
. . . ; I (v; w); _ I (v; w); . . . ; (r 01) (v; w)
I
(5)
and
for i = 1; . . . ; I;
(7)
(8)
There
Then, it can be veried that the regulator equations associated with system (1) admit a global solution given by
col(z(v; w); y(v; w); 4(v; w)), and u(v; w), where
prime.
A4)
and
118
(6)
(9)
_ = (; ; u)
(10)
(11)
A system of the form (10) with property (11) is called an internal model
candidate with output col(; u) [9]. The given plant (1) and the internal
model candidate (10) constitute an augmented system with the state
variables ( , z , y , ). Performing on the augmented system the following coordinate and input transformation:
(12)
puts the augmented system in new coordinates and input to the following form:
_ = fa (; u; v; w)
(13)
119
_ = (; ; u) = M + N 1 0 1 () + 9T 01 :
Z_ = f0 (Z; x1 ; )
x_ = f (Z; x1 ; . . . ; x ; ) + ^b ()x +1 ; i = 1; . . . ; r
i
i i
(16)
(17)
(18)
(19)
~ = 0 Nb01 (w)e
where
i i
(20)
(21)
(22)
f1 (Z; x1 ; )
= H (w)z + K~ (e; v; w)e
+ b(w) 1 ~ + b01 (w)Ne + 0 1 ()
f2 (Z; x1 ; x2 ; )
1 ( )
= 0 1 + @@
N 1
f +1 (Z; x1 ; . . . ; x +1 ; )
= 0 @@() N 1 0 ;
i = 2 ; . . . ; r 0 1:
x_ 1 = f1 (x1 ; x2 ; u; (t))
x_ 2 = f2 (x2 ; u; (t)) ;
t t0 0
(; 4; u) = M + N 41 0 1 () + 9T 01
= M + N 9T 01 = T 8T 01 = ():
(15)
f0 (Z; x1 ; )
= col M ~ 0 N 1[2] (~ + d) 0 1[2] (d) + (z ; e; v; w);
Then
where
(14)
Thus, (14) is an internal model candidate of the system (1) with output
col(; u). It will be seen later that this internal model candidate is such
that the augmented system is, under some other conditions, stabilizable
in the sense described in Lemma 3.1. Thus, in the rest of this note, we
will call (14) an internal model of the system (1) with output col(; u).
Remark 3.1: It is noted that in the special case where 41 (v; w) is a
polynomial in v , 1 ( ) = 9T 01 . The internal model (14) becomes
_ = M + N1 . This linear internal model was proposed in [16].
With the specic internal model (14), the augmented system (13)
takes the following form:
(23)
(24)
<
2<
, and x (t)
2L
, the solution
1[t ;t]
i)
L1 +
kx1 (t)k max f1 (kx1 (t0 )k ; t 0 t0 ) ;
1 x2[ ] ); 1 ( u[ ]
for some class KL function 1 , and some class K functions 1
and 1 , independent of .
ii) Subsystem (24) is RISS with x2 as state, u as input, i.e., for all
t t0 0, x2 (t0 ) 2 < , and u(t) 2 L1
kx2 (t)k max 2 (kx2 (t0 )k ; t 0 t0 ) ; 2 ( u[ ] )
for some class KL function 2 and some class K function 2 ,
independent of .
Then, system (23) and (24) is RISS with x(t) = col(x1 (t); x2 (t)) as
state and u as input, i.e., for all t t0 0, x(t0 ) 2 < + , and
u(t) 2 L1
kx(t)k max f(kx(t0 )k ; t 0 t0 ) ; u[ ] )
for some class KL function and class K function , independent of
. Moreover, it sufces to choose
(25)
(s) = max f21 (s); 21 2 (s); 22 (s)g :
n
t ;t
t ;t
t ;t
t ;t
120
(29)
Since 1e , z1 and 2e are C 1 functions, it is always possible to choose
a C 1 function satisfying (29). The proof is completed.
Remark 3.2: From [2], [9], [12], and [14], a controller that stabilizing the lower triangular system (22) can be explicitly constructed as
follows:
i = 1; . . . ; r 0 1
0 () 0 (~x );
i
Remark 3.3: Since M is Hurwitz, there exists a symmetric positivedenite matrix P such that
P M + M P = 0I:
T
Using Lemmas 3.1 and 3.2, we are ready to obtain the main result of
this note.
Theorem 3.1: Under assumptions A1)A4), assume the following.
A5) System (21) is RISS viewing ~ as state and col(
z ; e) as input
with a C 1 gain function. Then, the global robust output regulation problem for system (1) and exosystem (2) is solvable.
Proof: By Lemma 3.1, it sufces to show that under the assumptions A2) and A5), the upper subsystem of (22) is RISS viewing Z as
state and x1 as input with a C 1 gain function. This can be done by utilizing Lemma 3.2.
First, under Assumption A5), we have
; 1 e[0
e
;t]
(26)
;t]
z
~ v; w)e
(27)
z_ = F (w)z + G(e;
1
is RISS with a C gain function viewing z as state and e as input.
By assumption A2), there exists a symmetric positivedenite matrix
;t
(30)
(31)
for all ~, d. This assumption is to restrict the growth of the nonlinear
part of the function 1 (1). Moreover, the inequality (31) is satised in
[2]
some meaningful cases. First, if 1 is linear, 1 (x) 0. Thus, (31)
[2]
holds automatically. Next, note that (31) holds when j1 (~
+ d) 0
[2]
1 (d)j ((1 0 R)=2kPN k)k~k. Thus, (31) holds if 1[2] is glob[2]
ally Lipschitz, i.e., j1 (~
+ d) 0 1[2] (d)j Lk~k for some positive
number L, and the Lipschitz constant L satises L < (1=2kPN k).
Example 3.1: Consider the following system:
z_ = 0 z + 0:2v1 y + D1 (v; w)
y_ = z 0 0:1v2 y 0 sin2 (0:1wy) + 101 + D2 (v; w)
_1 = 0 1 + u; e = y 0 10v1
problem for this system. To be specic, we will only consider the case
where V = fv jv12 + v22 1g and W = [01; 1].
Solution: It can be easily veried that the solution of the regulator
equations is
121
any nonsingular matrix. Since 9 = [01 0], the pair f9; 8g is observable. Thus, the generator is linearly observable.
So far, we have veried the satisfaction of assumptions A1)A4). To
obtain an internal model that makes the satisfaction of assumption A5,
choose M = diagf01; 02g and N = [0:2 0:4]T , which are control-
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
1[2] (~
+ d) 0 1[2](d)
(1 0R)~T ~:
It can be easily veried that the previous inequality holds for 0 <
R < 0:72. Thus by Remark 3.3, assumption A5) is satised. Therefore, by Theorem 3.1, the global robust output regulation problem for
this system is solvable.
As a matter of fact, a controller that solves the global robust output
regulation problem for this system can be given as follows:
u = 2 () 0 1190~
x2
x~2 = 1 0 1 () + 17:3~
x1
x~1 = e
_ = M + N 1 0 1 () + 9T 01 :
The details of the synthesis are omitted due to the space limit.
IV. CONCLUDING REMARKS
[15]
[16]
[17]
[18]
, Nonlinear Output Regulation: Theory and Applications. Philadelphia, PA: SIAM, 2004.
, Remarks on robust output regulation problem for nonlinear systems, IEEE Trans. Autom. Control, vol. 46, pp. 20282031, Nov. 2001.
J. Huang and Z. Chen, A general framework for tackling the output regulation problem, IEEE Trans. Autom. Control, vol. 49, pp. 22032218,
Dec. 2004.
J. Huang and C.-F. Lin, On a robust nonlinear servomechanism
problem, IEEE Trans. Autom. Control, vol. 39, no. 7, pp. 15101513,
Jul. 1994.
A. Isidori and C. I. Byrnes, Output regulation of nonlinear systems,
IEEE Trans. Autom. Control, vol. 35, pp. 131140, Jan. 1990.
Z. P. Jiang and I. Mareels, A small-gain control method for nonlinear
cascaded systems with dynamic uncertainties, IEEE Trans. Autom.
Control, vol. 42, pp. 292308, Feb. 1997.
Z. P. Jiang, A. R. Teel, and L. Praly, Small-gain theorem for ISS systems
and applications, Math. Control, Signals Syst., vol. 7, pp. 95120, 1994.
W. Lin and Q. Gong, A remark on partial-state feedback stabilization
of cascade systems using small gain theorem, IEEE Trans. Autom. Control, vol. 48, no. 3, pp. 497500, Mar. 2003.
R. Marino and P. Tomei, Global adaptive output feedback control of
nonlinear systems, part I: linear parameterization, IEEE Trans. Autom.
Control, vol. 38, pp. 1730, Jan. 1993.
V. O. Nikiforov, Adaptive nonlinear tracking with complete compensation of unknown disturbances, Eur. J. Control, vol. 4, pp. 132139,
1998.
A. Serrari and A. Isidori, Global robust output regulation for a class of
nonlinear systems, Syst. Control Lett., vol. 39, pp. 133139, 2000.
E. D. Sontag and Y. Wang, New characterizations of input-to-state stability, IEEE Trans. Autom. Control, vol. 41, no. 9, pp. 12831294, Sep.
1996.
AbstractThe feasibility problem is studied of achieving a specied formation among a group of autonomous unicycles by local distributed control. The directed graph dened by the information ow plays a key role.
It is proved that formation stabilization to a point is feasible if and only if
the sensor digraph has a globally reachable node. A similar result is given
for formation stabilization to a line and to more general geometric arrangements.
Index TermsDistributed control, multiagent system, nonholonomic
mobile robots.
ACKNOWLEDGMENT
The authors would like to thank the anonymous reviewers for the
many constructive comments about the rst version of this note.
I. INTRODUCTION
REFERENCES
[1] C. I. Byrnes, F. Delli Priscoli, A. Isidori, and W. Kang, Structurally
stable output regulation of nonlinear systems, Automatica, vol. 33, pp.
369385, 1997.
[2] Z. Chen and J. Huang, Global robust stabilization of cascaded polynomial systems, Syst. Control Lett., vol. 47, pp. 445453, 2002.
[3]
, Nonlinear internal model and the robust output regulation
problem, in Proc. 2003 Amer. Control Conf., Denver, CO, June 2003,
pp. 15601565.
[4]
, Dissipativity, stabilization, and regulation of cascade-connected
systems, IEEE Trans. Automat. Contr., vol. 49, pp. 635650, May 2004.
[5]
, A variation of the small gain theorem, presented at the 8th Int.
Conf. Control, Automation, Robotics and Vision, Kunming, China, Dec.
2004.
[6] J. Huang, Asymptotic tracking and disturbance rejection in uncertain nonlinear systems, IEEE Trans. Autom. Control, vol. 40, pp.
11181122, June 1995.