Beruflich Dokumente
Kultur Dokumente
set of controllable modes and the set of uncontrollable modes do not Asymptotic Behavior of Nonlinear Networked Control
intersect. Some examples illustrate the result. Systems
Manuscript received April 5, 1999; revised December 23, 1999 and August
16, 2000. Recommended by Associate Editor L. Dai. This work was supported
in part by the United States Army Research Offic under Grant DAAG55-98-D-
0002, and in part by the National Science Foundation under Grant ECS-97-
02717.
G. C. Walsh is with the Department of Mechanical Engineering, University
of Maryland, College Park, MD 20742 USA (e-mail: gwalsh@eng.umd.edu).
O. Beldiman is with Mitusbishi Electric and Electronics, Inc., Durham, NC
27613 USA (e-mail: beldiman@msai.mea.com).
L. G. Bushnell is with the Department of Electrical Engineering, Univer-
sity of Washington, Seattle, WA 98195-2500 USA (e-mail: bushnell@ee.wash-
ington.edu).
Publisher Item Identifier S 0018-9286(01)06614-4.
sources may transmit its data, and so how the transmission opportuni-
ties are scheduled is critical.
Two scheduling algorithms are considered. The simplest is a static
scheduler. The transmission order is decided in advance and this order
is repeated indefinitely. This common method would typically be im-
plemented by polling or by a token ring. The other method uses bit-wise
arbitration to dynamically schedule the use of the network at run time.
This is the Try-Once-Discard (TOD) protocol introduced in [7]. Some
portion of a nodes identifier (the CAN-II identifier is 29 bit long) is
used to transmit a local evaluation of the importance of the locally held
Fig. 1. Configuration of networked control system. data. The information source or smart sensor that consider its data the
most important wins use of the network and all losing data packets are
and let y^(t) denote the input to the controller. This is the vector of discarded.
the most recently transmitted plant output values, and has the same Lemma 1 (Scheduler Error Bound) [7]: Consider a static or a dy-
dimension as y (t). At each transmission time a component(s) of y^(t) namic network scheduler starting at time t0 , visiting each node at least
is set to the corresponding component(s) of y (t). The controller is once, with integer periodicity p, maximum allowable transfer interval
described by the following equations: x_ c (t) = fc (t; xc (t); y^(t)), , maximum growth in error e(t) in seconds strictly bounded by
up (t) = gc (t; xc (t)). We define the network induced error as 2 (0; 1). Then, for any time t t0 + p , the error is bounded
e(t) = y(t) 0 y^(t). Between two successive transmission times we as ke(t)k < p(p 0 1)=2.
have e_ (t) well defined as long as gp is continuously differentiable. Proof: For the static scheduler, because the transmission is
Note that the network may also be used to transmit the control signal p-cyclic, every node will transmit at least one time in p seconds, and
u(t) by augmenting the error vector. the bound follows.
We now combine the controller and plant states into the vector x(t), For the dynamic scheduler, the error of the channels not transmitted
and we relabel the vector fields to obtain the following dynamics for is bounded by the error of the channel that did transmit. Consequently,
the closed-loop system in between two successive transmission times if any channel repeats, it establishes a bound for the remaining chan-
x_ (t) = f (t; x(t); e(t)) nels. If no channel repeats, then the error bound is similar to the static
scheduler because in p seconds all channels will have transmitted at
e_ (t) = g (t; x(t); e(t)) : (1) least one time. Let t1 ; . . . ; tp be the last p transmission times (t1 the
At each transmission time the vector e has a discontinuous jump, most recent) and i1 ; . . . ; ip be the channels that were transmitted, re-
namely the component corresponding to the transmitting output sensor spectively. Let im be the first channel that is repeated. If m = p then
is set to zero. We assume that the functions f and g are globally the lemma is proved, since the error on channel ik at time t1 is bounded
Lipschitz with associated Lipschitz constants kf and kg . We then by k .
Assume m < p. Let 1 l < m be the index of the time when
channel im was last transmitted. Then ei (tl ) (m 0 l) . Also, since
define
z (t) = (x(t); e(t))0 at time tl channel im was transmitted it must have had the highest error,
z_ (t) = h (t; z (t)) so all the other channels have error bounded by (m 0 1) . The channels
ik with 1 k l have the usual bound k . Summing up and taking
and let kh denote the global Lipschitz constant of the function h =
the worst case will give the bound from the lemma statement.
(f; g)0 .
Our approach is to model the error e(t) as a perturbation on the
system. If e(t) 0, then y^(t) = y (t) and from the control design III. STABILITY OF NONLINEAR NETWORKED CONTROL SYSTEMS
we have a Lyapunov function V (t; x) and constants c1 , c2 , c3 and c4 In this section, we present the main result of this note: that the sta-
greater than zero with bility property of a networked system is preserved if the network is fast
c1 kxk2 V (t; x) c2 kxk2 (2) enough. There are no assumptions about the initial conditions of the
error e(t), and so only BellmanGronwall can bound the growth of
@V @V
+ f (t; x; 0) 0c3 kxk2 (3) z = (x; e)0 during the first p seconds. At t = p , however, Lemma
@t @x 1 provides a bound on ke(t)k. We then prove that for t p , the error
@V remains bounded and kx(t)k decreases exponentially to a ball propor-
@x
c4 kxk: tional to the bound on ke(t)k. The previous steps are repeated to show
There exist many models of network traffic. We assume packets can overall exponential decay.
arrive at any time but the amount of time between transmissions obeys Theorem 1: Consider a NCS whose continuous dynamics are de-
a deadline, . Even a simple Poisson distribution allows for the possi- scribed by equation (1) with the control law designed so that equations
bility of arbitrary long delays between transmissions, making the sta- (2)(4) are satisfied, with p nodes of sensors operating under TOD or
bility analysis impossible. In the following derivations we assume that static scheduling, a maximum allowable transfer interval (MATI) that
there is a bound for the delays between transmission. We will call this satisfies < minf(ln(2)=kh p), (S=8), (Sc3 =32kf c4 )((c1 =c2 ))3=2 g
deadline the maximum allowable transfer interval (MATI). where S = (( c2 =c1 +1)kh pi=1 j )01 . Then, the networked system
is exponentially stable.
II. PERTURBATION BOUNDS Proof: We first expand the bounds imposed on for easy refer-
ence. Note that some of the bounds are used for several inequalities.
The following stability results are based on ideas from perturbation
theory [1]. We consider the error e(t) as a vanishing perturbation and ln(2)
we compute worst case bounds for it as a function of a bound on
<
kh p
) ek p
<2 (4)
its derivative, that is, ke_ (t)k < for t in some fixed interval of time. S
At any given transmission opportunity, only one of the p information <
8
) < S (5)
IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 46, NO. 7, JULY 2001 1095
) 2S < 14 (6) where we used (9) and (2) to get the bound on kxk and (10) for the
bound on kek.
= The growth of error in any interval included in [t0 + p; t^] is
< Sc3 c1 ) kf c 2S < c
3 2
32kf c4 c2 4 3 (7) bounded by kh (2=S + 2 c2 =c1 )kz (t0 )k, which is less than , ac-
cording to (5). Applying Lemma 1 on [t^ 0 p; t^], we get
4kf c4 c2
S c1 e(t^) < 2 kz (t0 )k
S
) c1
2: 1
(8)
so we obtain the contradiction. Therefore
Let z (t0 ) be the initial condition at the initial time t0 . For any t in ke(t)k < 2S kz(t )k ; 0 8 t t + p 0 (14)
[t0 ; t0 + p ], using BellmanGronwall and (4) we have:
=
@V +
@V (f (t; x; 0) + f (t; x; e) 0 f (t; x; 0)) :
kz(t0 + kt1 ) k (3=4)k kz(t )k : 0
@t @x Thus, the system is exponentially stable.
Using the Lyapunov bounds (2) and (3) and the Lipschitz condition for Remark. [Local Exponential Stability]: We can prove that local ex-
the function f , we get ponential stability is preserved with a networked architecture, but the
V_ (x) 0c3 kxk2 + c4 kxkkf kek (13) estimate of the region of attraction is correspondingly reduced. In par-
ticular, if the Lyapunov function, its associated constraints and the Lip-
schitz constants kf , kg and kh are defined for kxk < c5 with c5 > 0,
and due to (12) and (10)
By the choice of [see (7)], V_ (t;^ x(t^)) < 0. This means that V is With the network present, during the first p seconds, the proof pro-
decreasing locally around t^ so it cannot cross the bound. The only pos- vides only a BellmanGronwall bound on the growth and hence kz k
sible case is, consequently, when (10) does not hold meaning that may as much as double. The proof guarantees this is the maximum that
2
kzk will attain; however this limits the region of attraction to half the
e(t^) =
S
kz(t )k : 0 limit attained without the network to fz : kz k < (c5 =2) c1 =c2 g.
Fig. 4. Single link: stable ( = 0:01) (solid line) and unstable ( = 0:05)
(dashed line) trajectories.
V. CONCLUSION
In this note, we analyze a fundamental problem: the stability of a
Fig. 3. Single link: the percent of stable cases as function of . The system nonlinear networked control system, i.e., a nonlinear control system
loses stability when is around 0.015 having a feedback loop closed through a communication network. A
networked control architecture has many advantage compared to the
Consider a link driven by a motor through a torsional spring in the traditional point-to-point design, but there are also some new problems
vertical plane as shown in Fig. 2. The equations of motion are as fol- due to the presence of a communication network. The main compli-
lows: cation is the presence of the random time-delays, but there are also
I q1 + Mg sin q1 + k(q1 0 q2 ) = 0
undesired effects due to the discrete nature of the messages that can be
exchanged over the network.
J q2 0 k(q1 0 q2 ) = u: Our paradigm in this note is that the controller for the plant is de-
signed regardless of the network, and it exponentially stabilizes the
The system is linearized exactly with the following transformation: plant (locally or globally) if the network is not present. We study a
q1 newly introduced network protocol (TOD) that resolves collisions by
q_1 transmitting the channel with highest error between its last transmis-
z= sion and its current data.
0(MgL=I ) sin q1 0 (k=I )(q1 0 q2) We prove that if certain assumptions hold, the plant will be stabilized
0(MgL=I )q_1 cos q1 0 (k=I )(q_1 0 q_2 ) even in the presence of the network, although the region of attraction
may be smaller (and we give an estimate for this region). The main
IJ
u= (v 0 a(x)) assumptions that we make are: 1) that the intertransmission interval
k is bounded, and the bound is sufficiently small,and 2)that the collision
where a(x) = (MgL=I ) sin q1 [q_1 2 +(MgL=I ) cos q1 +(k=I )] resolution is perfect, meaning that the highest error channel will always
+(k=I )(q1 0 q2 )[(k=I ) + (k=J ) +(MgL=I ) cos q1 ]. The resulting be transmitted.
equations are in Brunowski cannonical form, z_1 = z2 , z_2 = z3 , Future research includes taking into account the presence of the com-
z_3 = z4 , z_4 = v . We then design a state feedback controller for this munication network when the controller is designed.
linear system in order to stabilize the origin, and then map it back to
the original system.
REFERENCES
The goal of this example is to show what happens to the stability of
the system when the average amount of bandwidth allocated to the con- [1] H. K. Khalil, Nonlinear Systems, second ed. Upper Saddle River, NJ:
Prentice-Hall, 1996.
trol loop is varied. We are going to model this variation by changing the [2] U. zgner, H. Goktag, and H. Chan, Automotive suspension control
average of the Poisson distribution used for the transmission times. through a computer communication network, in Proc. IEEE Conf. Con-
Due to the random nature of the transmission times distribution, we run trol Applications, 1992.
IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 46, NO. 7, JULY 2001 1097
[3] D. Radford, Spread-spectrum data leap through ac power wiring, IEEE in a form of a reduced and boundary-layer solutions. On the other hand,
Spectrum, pp. 4853, Nov. 1996. sufficient conditions will be given for proving the validity of the ap-
[4] R. S. Raji, Smart networks for control, IEEE Spectrum, pp. 4955, proximation. In Section III, we turn our attention to optimal control
June 1994.
[5] A. Ray, Performance evaluation of medium access control protocols problems of discrete-time singularly perturbed nonlinear systems. Fol-
for distributed digital avionics, ASME J. Dyn. Syst., Meas., Control, lowing the spirit and philosophy of Section II, we provide significant
vol. 109, pp. 370377, Dec. 1987. order reduction of optimal control problems of such systems.
[6] J. A. Sparks, Low cost technologies for aerospace applications, Mi-
croprocessors and Microsystems, vol. 20, pp. 449454, 1997.
[7] G. C. Walsh, H. Ye, and L. G. Bushnell, Stability analysis of networked II. SYSTEM DESCRIPTION AND DECOMPOSITION
control systems, in Proc. Amer. Control Conf., June 1999.
Consider a discrete-time nonlinear systems of the form
x(k + 1) = f (x(k); y(k); ) = 1 f (x(k); y(k)) (1)
1
y(k + 1) = g(x(k); y(k); ) = g(x(k); y(k)) (2)
where f (1) and g (1) are smooth functions for x 2 Dx n and y 2
Optimal Control of a Particular Class of Singularly Dy m satisfying f (0; 0) = g(0; 0) = 0. Here, Dx and Dy are
Perturbed Nonlinear Discrete-Time Systems open convex neighborhoods of the origins of the respective Euclidian
spaces. k 2 and is a smallpositive perturbation parameter. This
Rachid Bouyekhf, Abdelkhalek El Hami, and Abdellah El Moudni system will be studied under the following assumption.
Assumption 1: For all (x; y )T 2 Dx 2 Dy we assume that
i) supk0 kx(k)k a1 < 1;
AbstractThis note studies a class of discrete-time nonlinear systems ii) supk0 ky (k)k a2 < 1;
iii) ((@f=@x)(x; y )) < where (1) denotes the spectrum of
which depend on a small parameter. Using the singular perturbation theory
in a systematic way, we give a trajectory approximation result based on the
decomposition of the model into reduced and boundary layer models. This a matrix and 0 < < 1.
decomposition is used to analyze optimal control via maximum principle System (1) and (2) is characterized by an explicit separation of vari-
of such systems. As a result, significant order reduction of optimal control ables x(k) and y (k), owing to the presence of the small parameter
problems is achieved.
that multiplies the state y(k). This fact allows decomposing such
Index TermsDiscrete-time nonlinear systems, optimal control, singular system into separate reduced-order subsystems. Indeed, the reduced
perturbation. subsystem is obtained by setting = 0, then the dimension of the
system (1) and (2) reduces from n + m to n because the system (1)
I. INTRODUCTION and (2) degenerates into the n-dimensional system
One of the most popular methods for model reduction of large-scale xs (k + 1) = f (xs (k); 0) (3)
systems is the singular perturbation method. This approach has at- ys (k + 1) = g(xs (k); 0) (4)
tracted much attention because of its simplicity and good performance where xs and ys refer to the reduced (or slow) solutions.
in most experimental situations. In the case of discrete-time linear sys- Assumption 2: Assume that, the Jacobian matrix (@f (x; 0))=@x is
tems, the theory of the singular perturbation has been probably initiated nonsingular in Dx .
by [1], [2]. In these papers, the method is treated by using the similar Thus, in view of the implicit function theorem implies that there ex-
ideas used in continuous-time case. After these results, the singular per- ists a unique smooth function h(1) mapping Dx onto itself such that
turbation discrete-time linear system has received a great deal of atten-
tion in the literature, and significant progress has been made toward the xs (k) = h(xs (k + 1)): (5)
optimal feedback control of large-scale linear systems. Among several To obtain the reduced model we substitute (5) into (4)
contributions to the subject, we would-like to point out those related to
xs (k + 1) = f (xs (k); 0) = 1 f (x (k))
[3]. In that paper, the authors have derived sufficient conditions for the s s (6)
existence of a suboptimal solution, and a near-optimum design method
1
ys (k) = g(h(xs (k)); 0) = gs (xs (k)): (7)
has also been proposed. Another notable work in this area is the book
It is straightforward to see that xs starts from the same initial condition
of Naidu [4] in which the author gives a particularly excellent account
as x. By contrast ys is not free to start from y0 , hence the approximation
of the application of the singular perturbation in discrete-time linear
of x by xs may be uniform for all k 2 that is
control theory.
This paper is entirely devoted to the singularly perturbed nonlinear x(k) = xs (k) + O(): (8)
discrete-time systems and develops a discrete-time counterpart of cer- However, the approximation
tain results of singularly perturbed continuous-time nonlinear control
problems. More precisely, a method will be developed in Section II y(k) = ys (k) + O() (9)
which parallels the corresponding method in [6]. In particular, a mode hold only on an interval excluding 0, that is, for k 2 fk1 ; k1 + 1; . . .g,
decoupling approach will be given for the solution to be approximated where k1 > 0. Thus, following the nomenclature of the singular per-
turbation literature [5], (1) and (2) is said to be the singularly perturbed
form and the boundary layer occurs at k = 0. Layer correction terms
must thus be added to the reduced solution ys (k) for the approximation
Manuscript received October 1, 1999; revised December 11, 2000. Recom-
mended by Associate Editor M. Krstic
R. Bouyekhf and A. El Moudni are with the University of Technology Belfort- (9) to be valid over the entire interval. To investigate this initial value
Montbeliard (UTBM), Laboratory SET, 90010 Belfort Cedex, France. problem we introduce the following transformation as suggested by [4]
A. El Hami is with INSA of Rouen-LMR-UPRESA 6104, 76801 Saint-Eti- for nonlinear systems
x(k) = k+1 xf (k); y(k) = k yf (k)
enne du Rouvroy Cedex, France.
Publisher Item Identifier S 0018-9286(01)06613-2. (10)