1122

IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 48, NO. 7, JULY 2003

A Separation Principle for a Class of Non-UCO Systems

Manfredi Maggiore, Member, IEEE, and Kevin M. Passino, Senior Member, IEEE
AbstractThis paper introduces a new approach to output feedback stabilization of single-inputsingle-output systems which, unlike other techniques found in the literature, does not use quasilinear high-gain observers and control input saturation to achieve separation between the state feedback and observer designs. Rather, we show that by using nonlinear high-gain observers working in state coordinates, together with a dynamic projection algorithm, the same kind of separation principle is achieved for a larger class of systems which are not uniformly completely observable. By working in state coordinates, this approach avoids using knowledge of the inverse of the observability mapping to estimate the state of the plant, which is otherwise needed when using high-gain observers to estimate the output time derivatives. Index TermsNonlinear control, nonlinear observer, output feedback, separation principle.

I. INTRODUCTION HE area of nonlinear output feedback control has received much attention after the publication of the work [3], in which the authors developed a systematic strategy for the output feedback control of input-output linearizable systems with full relative degree, which employed two basic tools: an high-gain observer to estimate the derivatives of the outputs (and hence the system states in transformed coordinates), and control input saturation to isolate the peaking phenomenon of the observer from the system states. Essentially the same approach has later been applied in a number of papers by various researchers (see, e.g., [9], [12], [7], [13], and [1]) to solve different problems in output feedback control. In most of the papers found in the literature, (see, e.g., [3], [9], [12], [13], [7], and [14]) the authors consider input-output feedback linearizable systems with either full relative degree or minimum phase zero dynamics. The work in [21] showed that for nonminimum phase systems the problem can be solved by extending the system dynamics with a chain of integrators at the input side. However, the results contained there are local. In [19], by putting together this idea with the approach found in [3], the authors were able to show how to solve the output feedback stabilization problem for smoothly stabilizable and uniformly completely observable (UCO) systems.
Manuscript received October 27, 1999; revised February 5, 2002 and March 10, 2003. Recommended by Associate Editor W. Kang. This work was supported by NASA Glenn Research Center under Grant NAG3-2084. M. Maggiore was with The Ohio State University, Columbus, OH 43210 USA. He is now with the Department of Electrical and Computer Engineering, University of Toronto, Toronto, ON M5S 3G4, Canada (e-mail: [email protected]). K. M. Passino is with the Department of Electrical Engineering, The Ohio State University, Columbus, OH 43210 USA (e-mail: [email protected]). Digital Object Identifier 10.1109/TAC.2003.814110

The work in [1] unifies these approaches to prove a separation principle for a rather general class of nonlinear systems. The recent work in [16] relaxes the uniformity requirement of the UCO assumption by assuming the existence of one control input for which the system is observable. On the other hand, however, [16] requires the observability property to be complete, i.e., to hold on the entire state space. Another feature of that work is the relaxation of the smooth stabilizability assumption, replaced by the notion of asymptotic controllability (which allows for possibly nonsmooth stabilizers). A common feature of the papers previously mentioned is their input-output variable approach, which entails using the vector as feedback, for some integers , , where and denote the system output and input, respectively. This in particular implies that, when dealing with systems which are not input-output feedback linearizable, such approach requires the explicit knowledge of the inverse of the observability mapping, which in some cases may not be available. This paper develops a different methodology for output feedback stabilization which is based on a state-variable approach and achieves a separation principle for a class of non-UCO systems, specifically systems that are observable on an open region of the state space and input space, rather than everywhere. We impose a restriction on the topology of such an observability region assuming, among other things, that it contains a sufficiently large simply connected neighborhood of the origin. The main contributions are the development of a nonlinear observer working in state coordinates (which is proved to be equivalent to the standard high-gain observer in output coordinates), and a dynamic projection operating on the observer dynamics which eliminates the peaking phenomenon in the observer states, thus avoiding the need to use control input saturation. One of the benefits of a state-variable approach is that the knowledge of the inverse of the observability mapping is not needed. It is proved that, provided the observable region satisfies suitable topological properties, the proposed methodology yields closed-loop stability. In the particular case when the plant is globally stabilizable and UCO, this approach yields semiglobal stability, as in [19], provided a convexity requirement is satisfied. As in [21], [19], and [1], our results rely on adding integrators at the input side of the plant and designing a stabilizing control law for the resulting augmented system. Thus, a drawback of our approach (as well as the approaches in [21], [19], and [1]) is that separation is only achieved between the state feedback control design for the augmented system and the observer design and not between the state feedback control design for the original system and the observer design.

0018-9286/03$17.00 2003 IEEE

MAGGIORE AND PASSINO: SEPARATION PRINCIPLE FOR A CLASS OF NON-UCO SYSTEMS

1123

II. PROBLEM FORMULATION AND ASSUMPTIONS Consider the following dynamical system:

(1) , , and are known smooth functions, where . Our control objective is to construct a staand bilizing controller for (1) without the availability of the system states . In order to do so, we need an observability assumption. Define the observability mapping

, , and . Now, we are ready to state our first assumption. Assumption A1 (Observability): System (1) is observable containing the origin, i.e., the over an open set [where ] defined by mapping

where

(4)

has a smooth inverse (5)

. . . Remark 1: In the existing literature, an assumption similar to A1 can be found in [21] and [19]. It is worth stressing, howto be ever, that in both works the authors assume the set . When that is the case, the system is said to be [20], [19] UCO. In many practical applications, the system under consideration may not be UCO, but rather be observable in some only, thus preventing the use of most of subset of the output feedback techniques found in the literature, including the ones found in [21], [19], and [1]. On the other hand in A1 the mapping , viewed as a mapping acting on parameterized by , is assumed to be square (i.e., it maps spaces of equal dimension), thus implying that can be expressed as time derivatives and , i.e., a function of , its . In the works [19], [16], is allowed to be a function depending on a possibly higher number . In our setting, this is of derivatives of , rather than just equivalent to assuming that in A1, rather than being invertible, is just left-invertible. We are currently working on relaxing A1 and replace it by the weaker left-invertibility of . Assumption A2 (Stabilizability): The origin of (1) is locally stabilizable (or stabilizable) by a static function of , i.e., there such that the origin is an asympexists a smooth function totically stable (or globally asymptotically stable) equilibrium . point of Remark 2: Assumption A2 implies that the origin of the extended system (3) is locally stabilizable (or stabilizable) by a function of as well. A proof of the local stabilizability property for (3) may be found, e.g., in [17], while its global counterpart is a well known consequence of the integrator backstepping lemma (see, e.g., [5, Th. 9.2.3] or [10, Corollary 2.10]). Therefore, we conclude that for the extended system (3) there such that its origin is asymptotiexists a smooth control cally stable under closed-loop control. Let be the domain of attraction of the origin of (3), and notice that, when A2 holds . globally, Remark 3: In [21], the authors consider affine systems and use a feedback linearizability assumption in place of our A2. Here, we consider the more general class of nonaffine systems for which the origin is locally stabilizable (stabilizable). In this respect, our assumption A2 relaxes also the stabilizability assumption found in [19], while it is equivalent to [1, Assumption 2].

. . .

is the

th derivative) where

. . .

where ( indicates that there is no depen, dence on ). In the most general case, . In some cases, however, we may have for all and some inthat . This happens in particular when system (1) has teger a well-defined relative degree . Here, we do not require the system to be input-output feedback linearizable, and hence to possess a well-defined relative degree. In the case of systems corresponds to having with well-defined relative degree, , while corresponds to having . Next, augintegrators on the input side, ment the system dynamics with which corresponds to using a compensator of order . System (1) can be rewritten as follows: (2) Define the extended state variable and the associated extended system ,

(3)

1124

IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 48, NO. 7, JULY 2003

III. NONLINEAR OBSERVER: ITS NEED ANALYSIS

AND

STABILITY

Assumption A2 allows us to design a stabilizing state feed. In order to perform output feedback back control control should be replaced by its estimate. Many researchers in the past adopted an input-output feedback linearizability assumption [3], [9], [13], [7], [14] and transformed the system into normal form

(6) In this framework the problem of output feedback control finds a very natural formulation, as the first derivatives of are equal to the states of the -subsystem (i.e., the linear subsystem). The works [3], [9], and [7] solve the output feedback control ), so problem for systems with no zero dynamics (i.e., derivatives of provide the entire state of the that the first system. In the presence of zero dynamics ( -subsystem), the use of input-output feedback linearizing controllers for (6) forces the use of a minimum phase assumption (e.g., [13]) since the states of the -subsystem are made unobservable by such controllers and, hence, cannot be controlled by output feedback. For this reason the output feedback control of nonminimum phase systems has been regarded in the past as a particularly challenging problem. Researchers who have addressed this problem in (5), (e.g., [21], [19]) rely on the explicit knowledge of , so that estimation of the first derivatives of (the vector ) provides an estimate of , , since the vector , being the state of the controller, is known. Next, to estimate the derivatives of , they employ an high-gain observer. Both the works [21] and [19] (the latter dealing with the larger class of stabilizable to prove closed-loop systems) rely on the knowledge of stability. In addition to this, the work in [1] proves that a separation principle holds for a quite general class of nonlinear sysis explicitly known tems which includes (1) provided that and that the system is uniformly completely observable. Somecannot be explicitly calcutimes however, even if it exists, lated (see, e.g., the example in Section V) thus limiting the applicability of existing approaches. Hence, rather than estimating and using to get , the approach adopted here is to estimate directly using a nonlinear observer for (1) and using the fact that the -states are known. The observer has the form1

is a vector, , and is a fixed design constant. and Notice that (7) does not require any knowledge of has the advantage of operating in -coordinates. The observability assumption A1 implies that the Jacobian of the mapwith respect to is invertible, and hence the inverse ping in (7) is well defined. In the work [2], the auof thors used an observer structurally identical to (7) for the more restrictive class of input-output feedback linearizable systems with full relative degree. Here, by modifying the definition of and by introducing a dynamic projection, we the mapping considerably relax these conditions by just requiring the general observability assumption A1 to hold. Furthermore, we propose a different proof than the one found in [2] which clarifies the relationship among (7) and the high-gain observers commonly found in the literature. Theorem 1: Consider system (2) and assume that A1 is sat, the state belongs to a positively inisfied for , and that there exists variant, compact set , , which is positively invariant for and a set , is a compact set. Choose such that such that is Hurwitz. Under these conditions and using observer (7), the following two properties hold. i) Asymptotic stability of the estimation error: There exists , , such that for all , and all , as . ii) Arbitrarily fast rate of convergence: For each positive , there exists , , such that for all , . In Section IV, we show that, by applying to the vector field a suitable dynamic projection onto a fixed compact set, the existence of the compact positively invariant set is guaranteed. Proof: Consider the smooth coordinate transformation

where

which maps (1) to (8) where and are smooth functions and the pair is in Brunovsky canonical form. Similarly, it is not difficult to show that the coordinate transformation

maps the observer dynamics (7) to (9) Define the observer error in the new coordinates, Then, the observer error dynamics are given by .

(7)
this section, we assume A1 to hold globally, since we are interested in the ideal convergence properties of the state estimates. In the next section, we will show how to modify the observer equation in order to achieve the same convergence properties when A1 holds over the set .
1Throughout

(10) . Note that, in coordinates, the obwhere server (7) is almost identical to the standard high-gain observer

O

MAGGIORE AND PASSINO: SEPARATION PRINCIPLE FOR A CLASS OF NON-UCO SYSTEMS

1125

used in, e.g., [19] with an important difference: while and in [where denotes the saturation func[19] depend on tion] and, hence, the nonlinear portion of the observer error dy, and in (10) namics is globally bounded when do not contain saturation and, thus, one cannot guarantee global boundedness of the last term in (10). Since our observer operates in coordinates, saturation cannot be induced in coordinates and therefore, without any further assumption, we cannot prove stability of the origin of (10). The assumption concerning the existence of addresses precisely this issue since it guarantees that is contained in a compact set for all . Define the coordinate transformation (11) In the new coordinates the observer error dynamics are given by

that over, from (15)

implies that

, and . More-

Therefore, by the comparison lemma (see, e.g., [22]), isfies the following inequality:

sat-

(16) which, for sufficiently small , can be written as (12) is Hurwitz. Let where, by assumption, lution to the Lyapunov equation be the so(13) and consider the Lyapunov function candidate along the trajectories Calculate the time derivative of . An upper estimate of the time such that (and, thus, ) for all follows: , is calculated as

(14) Since for all , we have that and, hence, for all , which, by assumption, is , a compact set. This, together with the fact that, for all and , implies that there exists , independent of , such that the time a fixed scalar can be bounded as derivative of

. Noticing that as , we conclude that can be made arbitrarily small by choosing a sufficiently small , thus concluding the proof of part ii). Remark 4: Using inequality (16), we find the upper bound for the estimation error in -coordinates

for all

(17) (15) , we conclude that, for all Defining , the origin of (12) [and, hence, also the origin of (10)] is asymptotically stable which, by the smoothness of , as . This concludes the proof implies that of part i). As for part ii), note that may exhibit peaking, and Hence, during the initial transient, the size of the peak grows larger as decreases and the convergence rate is made faster. Refer to [18] for more details on the peaking phenomenon and to [3] for a study of its implications on output feedback control. The analysis in the latter paper shows that a way to isolate the peaking of the observer estimates from the system states is to saturate the control input outside of the compact set of interest. The same idea has then been adopted in several other works in the output feedback control literature (see, e.g., [3], [9], [19], [12][14], [7], and [1]). Rather than following this approach, in the next section we will present a new technique to eliminate (rather than just isolate) the peaking phenomenon which allows for the use of the weaker assumption A1. IV. OUTPUT FEEDBACK STABILIZING CONTROL Define so that implies that (the smoothness of guarantees that is well-defined). By the aforementioned inequality, we have Consider system (3), by using assumption A2 and Remark 2 we conclude that there exists a smooth stabilizing control which makes the origin of (3) an asymptotically stable equilibrium point with domain of attraction .

since

. Next

since

. Therefore

1126

IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 48, NO. 7, JULY 2003

Fig. 1. Mechanism behind the observer estimates projection.

Fig. 2.

Domains C , C , C violate some of the requirements i)iv) in A3, while C does not.

By the converse Lyapunov theorem found in [11], there exists a continuously differentiable function defined on satisfying, , for all (18) (19) (20) are class functions (see [8] for a defiwhere , stands for the boundary of the set . Given any nition), and , define scalar

Clearly, for all and, from (19), becomes . Next, the following assumption arbitrarily close to as is needed. Assumption A3 (Topology of ): Assume that there exists a and a set such that constant (21) where has the following properties. dimensional, subi) The boundary of , , is an function manifold of , i.e., there exists a such that , and on . is convex for all . ii) for each fixed , i.e., iii) 0 is a regular value of , . for all is compact. iv) Remark 5: See Fig. 1 for a pictorial representation of condition (21). This assumption requires in primis that there ex-

ists a set in coordinates which contains the image under of a level set of the Lyapunov function and is contained in the image under of the observable set . This guarantees , and thus, when the state feedback in particular that never controller is employed, the phase curves leaving from exit the observable region . Furthermore, it is required that possesses some basic topological properties: It is asked that the boundary of be continuously differentiable [part i)], every slice of obtained by holding constant at , , is convex (which is given by [part ii)], the normal vector to each slice ) does not vanish anywhere on the slice [part iii)] and, finally, it is asked that the set is compact in the direction [part iv)]. Part iv) can be replaced by the slightly weaker is compact ( is defined in the next requirement that section), with minor changes in the analysis to follow. To further clarify the topology of the domains under consideration, consider the sets to in Fig. 2, corresponding to the , . While they all satisfy part i), does not case satisfy part ii) since its slices along are not convex. satisfies part ii) but violates requirement iii) because the normal vector direction. to one of the slices has no components in the satisfies parts i)iii) but violates iv) since the area of its slices . satisfies all the aforementioned grows unbounded as requirements. ) Note that, when the plant is UCO (and, thus, , A3 is always satisfied by a suffiand . In order to see that, pick any ciently large set and any and choose to be any cylinder , , containing . The existence of is guarwhere is bounded. More generanteed by the fact that the set is not all of and ally, the same holds when is convex for all .

MAGGIORE AND PASSINO: SEPARATION PRINCIPLE FOR A CLASS OF NON-UCO SYSTEMS

1127

Finally, notice that when the plant in not UCO but , where is an open set which is not all of , and the origin is globally stabilizable (i.e., A2 holds globally), one can , where is any convex compact set choose with smooth boundary contained in and containing the point (i.e., the origin in coordinates). The scalar is then (notice, however, that the largest number such that does not need to be known for design purposes). Consequently, (and hence the control input in the particular case when does not affect the mapping ) and A2 holds globally, one can , where is defined above. choose A. Observer Estimates Projection As we already pointed out in Remark 4, in order to isolate the peaking phenomenon from the system states, the approach generally adopted in several papers is to saturate the control input to prevent it from growing above a given threshold. This technique, however, does not eliminate the peak in the observer estimate and, hence, cannot be used to control general systems like the ones satisfying assumption A1, since even when the system , the observer state lies in the observable region estimates may enter the unobservable domain where (7) is not well-defined. It appears that in order to deal with systems that are not completely observable, one has to eliminate the peaking from the observer by guaranteeing its estimates to be confined in a prespecified observable compact set. A common procedure used in the adaptive control literature (see [4]) to confine vectors of parameter estimates within a desired convex set is gradient projection. This idea cannot be directly applied to our problem, mainly because is not proportional to the gradient of the observer Lyapunov function and, thus, the projection cannot be guaranteed to preserve the convergence properties of the estimate. Inspired by this idea, however, we propose a way to modify the vector field which confines to within a prespecified compact set while preserving its convergence properties. Recall the coordinate transformation defined in (4) and let

represent the and components of the normal vector to the boundary of at , i.e., (the function is defined in A3). and are the Lie derivatives of and Further, along the vector fields and , respectively, i.e.,

and , where denotes the matrix square root of [defined in (13)]. The dynamic projection (23) is well-defined since A3, part does not vanish (see also Remark 5). iii), guarantees that The following lemma shows that (23) guarantees boundedness . and preserves convergence of Lemma 1: If A3 holds and (23) is used: : if i) Positive invariance of , then for all . for all and the asIf, in addition, sumptions of Theorem 1 are satisfied, then the following property holds for the integral curves of the projected observer dynamics (7), (23). ii) Preservation of original convergence characteristics: prop. erties i) and ii) established by Theorem 1 remain valid for Proof: We begin by introducing another coordinate trans, (similarly, let , formation, ), letting , and letting be the image of the set under the linear map , i.e., . Let , be the and components of the normal vector to the boundary of . The reader may refer to Fig. 1 for a pictorial representation of the sets under consideration. In order to prove part i) of the Lemma, we have to show that the projection (23) renders a positively . The coordinate transformation (22) maps invariant set for (23) to (24)

(22) where is the state of the projected observer defined as2

(25) if otherwise. if otherwise (23) where and and to and In order to relate , recall from A3 that the boundary of is expressed as and hence the boundary the set . From of is the set and as this definition we find the expression of and

2The projection defined in (23) is discontinuous in the variable y , therefore ^ raising the issue of the existence and uniqueness of its solutions. We refer the reader to Remark 7, were this issue is addressed and a solution is proposed.

1128

IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 48, NO. 7, JULY 2003

The expression of the projection (23) in by noting that

coordinates is found

where

is nonnegative since, by assumption (26) Using the fact that and that lies on the boundary of , we have that the difference vector points outside of or, equivalently, points outside of the . Using the definition of , slice is the image of the convex compact set we have that the set , defined in A3, under the linear map and is, therefore, compact and convex as well. Combining these two facts we have , thus proving that , which that concludes the proof of part ii). Remark 6: Part i) of Lemma 1 implies that, with dynamic projection, the requirement in Theorem 1 that there exists a posis satisfied with . itively invariant set for To see that, note that

if otherwise and then substituting , to find that

and

, and

if and otherwise. (27) Next, we show that the boundary of the domain is positively invariant with respect to (27). In order to do that, consider the and continuously differentiable function calculate its time derivative along the vector field of (27) when

and, thus

which, by part iv) in A3, is a compact set. Remark 7: In order to avoid the discontinuity in the right-hand side of (23) one can employ the smooth projection idea introduced in [15]. In this case, (21) in A3 should be replaced by the following: Since is positively invariant, so is and, therefore, . The proof of part ii) is based on the knowledge of a coordinates [see Lyapunov function for the observer in , we have that, in new coordinates, (11)]. Letting . Now let be a Lyapunov function candidate for the projected observer error dynamics in transformed coordinates and recall that, by and, thus, . The latter assumption, or, what is the same, . fact implies that is in the interior of , or is From (27), when (i.e., on the boundary of and the unprojected update is pointed to the interior of ), we have . Let us now consider all remaining cases, that and i.e., (29) satisfy properties i)iv) in A3 and , . The in (23) should be multiplied by , and ratio a slight modification of Lemma 1 would show that, by applying (rather than the set the smooth projection to , the set ) is made positively invariant. Remark 8: From the proof of Lemma 1, we conclude that within the set which is, in gen(23) confines , and is generally not eral, unknown since we do not know convex (see Fig. 1). It is interesting to note that applying a standard gradient projection for over an arbitrary convex domain does not necessarily preserve the convergence result ii) in Theorem 1. where B. Closed-Loop Stability To perform output feedback control we replace the state feedwith which, by the smoothback law is guaranteed to belong to , ness of and the fact that and

(28)

MAGGIORE AND PASSINO: SEPARATION PRINCIPLE FOR A CLASS OF NON-UCO SYSTEMS

1129

is bounded provided that is confined to within a compact set. In the following we will show that makes the origin of (3) is contained in its region of asymptotically stable and that .3 The proof is divided in three attraction, for all steps. and uniform ultimate 1) (Lemma 2). Invariance of boundedness: Using the arbitrarily fast rate of convergence of the observer [see part ii) in Theorem 1], we cannot exit show that any phase curve leaving from and converges in finite time to an arbitrarily the set small neighborhood of the origin. Here, Lemma 1 plays an important role, in that it guarantees that the peaking phenomenon is eliminated and thus it does not affect , allowing us to use the same idea found in [3] to prove stability. 2) (Lemma 3). Asymptotic stability of the origin: By using Lemma 2 and the exponential stability of the observer estimate, we prove that the origin of the closed-loop system is asymptotically stable. 3) (Theorem 2). Closed-loop stability: Finally, by putting together the results of Lemma 2 and 3, we conclude the closed-loop stability proof. The arguments of the proofs are relatively standard (see, e.g., [1] and [19]) and so are sketched. , and such that Lemma 2: For all

seen that . We have, thus, leaving from the set shown that for any phase curve there exists a uniform lower bound to the exit time which is independent of . Choose , let from , and be the Lipschtiz constant of over the compact set

By A3, implies and thus from Theorem 1, part ii), and Lemma 1 there exists such that, for all , for all . Now, using the function defined in (18)(20), a standard Lyapunov analysis shows that, for all

there exists a number for all

such that, for all , where

the phase curves of the closed-loop system remain con, the set is positively invariant, and is fined in reached in finite time. , by Sketch of the Proof: Since for all . Let Lemma 1, part i), and notice that and

which, since , implies that is positively invariant ), and that the phase curves enter and stay in ( in finite time. the set Remark 9: The use of the projection for the observer estimate plays a crucial role in the proof of Lemma 2. As is made smaller, the observer peak may grow larger, thus generating a large control input, which in turn might drive the system states outside of in shorter time. The dynamic projection makes sure that the exit time is independent of , thus allowing one to choose independently of . Lemma 3: Under the assumptions of Lemma 2, there exists a the origin positive scalar such that for all is asymptotically stable. . By (17) and Sketch of the Proof: Let dynamics is exponentially Lemma 1, the origin of the , from stable. Recalling that dythe smoothness of we conclude that the origin of the namics is exponentially stable as well. By the converse Lya, a pospunov theorem there exists a Lyapunov function such that, for itive number , and positive constants all

which is a compact set independent of . Hence, there exists a bounded positive real number independent of such that, for and all , all

Choose such that , choose as in Lemma 2 and pick any any Then, the following Lyapunov function:

. For .

Therefore, for all such that , , has a positive proving that the exit time from the set , independent of . Further, letting lower bound, , it is readily
3Recall that c is a positive constant satisfying A3 and hence its size is constrained by the topology of the observable set .

readily allows one to conclude that the origin is asymptotically stable. We are now ready to state the following closed-loop stability theorem. Theorem 2: Suppose that assumptions A1, A2, and A3 are satisfied. Then, for the closed-loop system (3), (7), (23), with , there exists a scalar such control law

1130

IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 48, NO. 7, JULY 2003

that, for all , the set is contained in the region of . attraction of the origin Sketch of the Proof: By Lemma 3, there exists such that, for all , the origin is be its domain of attraction. For asymptotically stable. Let , use Lemma 2 to find so any the phase curves of the closed-loop system that for all in finite time. enter Remark 10: Theorem 2 proves regional stability of the closed-loop system, since given an observability domain , and provided A3 is satisfied, the control law , together with (7) and (23), make the compact set an estimate of the domain of attraction for the origin of the closed-loop system. The difference between Theorem 2 and a local stability result lies in the fact that here the domain of attraction for is at least as and not restricted to be a small unknown neighborlarge as is independent of hood of the origin. Further, the size of and thus the domain of attraction does not shrink as the rate of convergence of the observer is made faster (that is, when ). The next corollary gives conditions for the recovery of the domain of attraction of the state feedback controller by output feedback. , Corollary 1: Assume that A1 is satisfied on is convex for all , and A2 holds. Then, , there exists a scalar such given any set , the set is contained in the region of that, for all . Moreover, if , the attraction of the origin is semiglobally asymptotically stable. origin , Proof: By property (19) of the Lyapunov function , there exists such that given any set . By Remark 5, if , A3 is satisfied by any and a sufficiently large . Choose . By Theorem such that, for all 2 we conclude that there exists , the origin of the closed-loop system is asymptot, ically stable and the set is contained in its domain of attraction. is contained in the domain of attraction Thus, in particular, . If then can be chosen to be any of is semiglobally asympcompact set and thus the origin totically stable. Remark 11: A drawback of the result in Theorem 2, shared by the works in [21], [19], and [1], is that separation is achieved between the observer design and the state feedback control design for the augmented system (3). In order to avoid increased complexity of the controller, it would be more desirable to achieve separation between the observer design and the state feedback control design for the original system (1). Remark 12: As mentioned in Remark 5, if the plant is UCO ) and is a convex set (and hence , assumption A3 is automatically satisfied by a for all is sufficiently large cylindrical set . Even in this case, if not explicitly known and one wants to directly estimate the state of the plant, one should employ the dynamic projection (23) since the standard saturation used, e.g., in [3] and [19] can only coordinates. Clearly, be applied to a high-gain observer in the only instance when dynamic projection can be replaced by saturation of the observer estimates is when the observability mapping is the identity.

V. EXAMPLE Consider the following inputoutput linearizable system:

(30) The control input appears in the first derivative of the output

Notice, however, that the coefficient multiplying vanishes , and hence system (30) does not have a well-dewhen fined relative degree everywhere. Since appears in , we have , therefore, we add one integrator at the input side that (31) The mapping is given by

(32) , and its The first equation in (32) is invertible for all . Substituting into the second inverse is given by equation in (32) and isolating the term in , we get (33) is a strictly increasing function, it folSince , however, an analytlows that (33) is invertible for all ical solution to this equation cannot be found. In conclusion, Assumption A1 is satisfied on the domain , but an explicit inverse is not known. The fact that system (30) is not UCO, together with the nonexistence of an explicit inverse to (32), prevents the application of the output feedback control approaches in [3], [21], [9], [19], [12], [13], [7], [14], and [1]. To find a stabilizing state feedback controller, note that the extended system (30), (31) can be feedback linearized by letting and rewriting the system in : new coordinates

(34) Choose , where , so that the closed-loop system becomes with poles placed at 1. Then, the origin is a globally asymptotically equilibrium point of (34), and As. Let be the solution sumption A2 is satisfied with , so that a of the Lyapunov equation associated to

MAGGIORE AND PASSINO: SEPARATION PRINCIPLE FOR A CLASS OF NON-UCO SYSTEMS

1131

Lyapunov function for system (34) is , and any set , with , is contained in the region of attraction for the origin. Equation (35) is shown at the bottom of the page. Next, we seek to find a set satisfying Assumption A3. To this end, notice that

Next recall from Remark 5 that, since A2 holds globally, can , where is any compact be chosen to be the cylinder containing convex set in the upper half plane . For the sake of simplicity choose the point to be the disk of radius centered at , so that , and , as depicted in Fig. 3. We stress that our choice is quite conservative and is made exclusively for the sake of illustration. Once has been chosen, the control design is complete and the output feedback controller is given by
Fig. 3. Projection domain C .

(36) let where (35) and is the solution of (23) with defined in and note that, for all ,

Next, we seek to find such that ities above we have that if the largest scalar satisfying (37) Using controller (37), Theorem 2 guarantees that the origin of the closed-loop system is asymptotically stable and it provides an estimate of its domain of attraction. , there exists Specifically, given any positive scalar such that is contained in the domain of attraction . In what follows we will find the set for all satisfying A3. Recalling that, in -coordinates, is , we have that expressed as implies , , and hence . Now

. Using the inequal, where is

(38) satisfying the inequality above always exists) (note that and, hence, satisfies A3. then and, from (38), we For our simulations we choose . The initial condition of the extended system is get [or set to ], which is contained inside so that Theorem 2 can be , applied. Finally, we choose the observer gain to be

(35)

1132

IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 48, NO. 7, JULY 2003

(a)

Fig. 5. Observer states during the initial peaking phase with and without projection,  = 10 .

(b)

Fig. 4. Integral curve of the closed-loop system under output feedback. (a) = 0:2. (b)  = 0:05.

Fig. 6. Phase curve of the observer with dynamic projection in the transformed domain Y = F (X ).

so that its associated polynomial is Hurwitz with both poles placed at 2. We present four different situations to illustrate four features of our output feedback controller. 1) Arbitrary fast rate of convergence of the observer. , as Fig. 4 shows the evolution of the integral curve and . well as the control input , for The convergence in the latter case is faster, as predicted by Theorem 1 (see Remark 4). 2) Observer estimate projection. Fig. 5 shows the evoluwith and without projection of and for tion. The dynamic projection successfully eliminates the peak in the observer states, thus yielding a bounded control input, as predicted by the result of Lemma 1. Fig. 6 is contained shows that the phase curve , confirming the result of within the set for all Lemma 1. In particular, Fig. 6 shows the operation of the projection when the phase curve of the observer (in coordinates) hits the boundary of : it forces to

slide along the boundary of and preserves its convergence characteristics. This is equivalent, in the domain, slides along to saying that the phase curve and converges to . the boundary of 3) Observer estimate projection and closed-loop stais shown bility. In Fig. 7 a phase plane plot for . with and without observer projection when The small value of generates a significant peak which, if projection is not employed, drives the phase curve of the output feedback system away form that of the state feedback system and, in general, may drive the system to instability (see Remark 9). On the other hand, using dynamic projection, the phase curve of the output feedback system is almost indistinguishable from the phase curve of the state feedback system. 4) Trajectory recovery. The evolution of the phase curve for decreasing values of , in Fig. 8, shows that the phase curve of the output feedback system approaches (see the phase curve of the state feedback one as Remark 11).

MAGGIORE AND PASSINO: SEPARATION PRINCIPLE FOR A CLASS OF NON-UCO SYSTEMS

1133

Fig. 7.

Phase curve x(t) with and without projection,  = 10

[10] M. Krstic, I. Kanellakopoulos, and P. Kokotovic, Nonlinear and Adaptive Control Design. New York: Wiley, 1995. [11] J. Kurzweil, On the inversion of Ljapunovs second theorem on stability of motion, Amer. Math. Soc. Trans., vol. 24, pp. 1977, 1956. [12] Z. Lin and A. Saberi, Robust semi-global stabilization of minimumphase input-output linearizable systems via partial state and output feedback, IEEE Trans. Automat. Contr., vol. 40, pp. 10291041, June 1995. [13] N. Mahmoud and H. Khalil, Asymptotic regulation of minimum phase nonlinear systems using output feedback, IEEE Trans. Automat. Contr., vol. 41, pp. 14021412, Oct. 1996. , Robust control for a nonlinear servomechanism problem, Int. J. [14] Control, vol. 66, no. 6, pp. 779802, 1997. [15] J. Pomet and L. Praly, Adaptive nonlinear regulation: Estimation from the Lyapunov equation, IEEE Trans. Automat. Contr., vol. 37, pp. 729740, June 1992. [16] H. Shim and A. R. Teel, Asymptotic controllability imply semiglobal practical asymptotic stabilizability by sampled-data output feedback,, to be published. [17] E. D. Sontag, Remarks on stabilization and input to state stability, in Proc. IEEE Conf. Decision Control, Tampa, FL, Dec. 1989, pp. 13761378. [18] J. H. Sussmann and P. V. Kokotovic, The peaking phenomenon and the global stabilization of nonlinear systems, IEEE Trans. Automat. Contr., vol. 36, pp. 424440, Apr. 1991. [19] A. Teel and L. Praly, Global stabilizability and observability imply semi-global stabilizability by output feedback, Syst. Control Lett., vol. 22, pp. 313325, 1994. [20] , Tools for the semiglobal stabilization by partial state and output feedback, SIAM J. Control Optim., vol. 33, no. 5, pp. 14431488, 1995. [21] A. Tornamb, Output feedback stabilization of a class of nonminimum phase nonlinear systems, Syst. Control Lett., vol. 19, pp. 193204, 1992. [22] T. Yoshizawa, Stability Theory by Lyapunovs Second Method. Tokyo, Japan: The Mathematical Society of Japan, 1966.

Fig. 8. Phase curve X (t) of the closed-loop system for decreasing values of .

Manfredi Maggiore (M00) received the Laurea degree in electronic engineering from the University of Genoa, Italy, and the Ph.D. degree in electrical engineering from The Ohio State University, Columbus, in 1996 and 2000, respectively. He is currently an Assistant Professor in the Edward S. Rogers, Sr. Department of Electrical and Computer Engineering, the University of Toronto, Canada. His research interestsare in nonlinear control, including output feedback control, output tracking, and nonlinear estimation. Kevin M. Passino (S79M90SM96) received the Ph.D. degree in electrical engineering from the University of Notre Dame, Notre Dame, IN, in 1989. He has worked on control systems research at Magnavox Electronic Systems Co., Fort Wayne, IN, and McDonnell Aircraft Co., St. Louis, MO. He spent a year at Notre Dame as a Visiting Assistant Professor and is currently Professor of Electrical Engineering at The Ohio State University (OSU), Columbus. Also, he is the Director of the OSU Collaborative Center of Control Science that is funded by AFRL/VA and the Air Force Office of Scientific Research. He is coeditor or coauthor of many books, including An Introduction to Intelligent and Autonomous Control (Norwell, MA: Kluwer, 1993), Fuzzy Control (Reading, MA: Addison-Wesley, 1998), and Stability Analysis of Discrete Event Systems (New York: Wiley, 1998). Dr. Passino has served as the Vice President of Technical Activities of the IEEE Control Systems Society (CSS), was an Elected Member of the IEEE Control Systems Society Board of Governors, was Chair for the IEEE CSS Technical Committee on Intelligent Control, has been an Associate Editor for the IEEE TRANSACTIONS ON AUTOMATIC CONTROL and also for the IEEE TRANSACTIONS ON FUZZY SYSTEMS, served as the Guest Editor for the 1993 IEEE Control Systems Magazine Special Issue on Intelligent Control, was a Guest Editor for a special track of papers on Intelligent Control for IEEE Expert Magazine in 1996, and was on the Editorial Board of the International Journal for Engineering Applications of Artificial Intelligence. He was a Program Chairman for the 8th IEEE International Symposium on Intelligent Control 1993, and was the General Chair for the 11th IEEE International Symposium on Intelligent Control. He was the Program Chair of the 2001 IEEE Conference on Decision and Control. He is currently a Distinguished Lecturer for the IEEE Control Systems Society.

ACKNOWLEDGMENT The authors would like to thank M. Broucke for fruitful discussions concerning the dynamic projection. REFERENCES
[1] A. Atassi and H. Khalil, A separation principle for the stabilization of a class of nonlinear systems, IEEE Trans. Automat. Contr., vol. 44, pp. 16721687, Sept. 1999. [2] G. Ciccarella, M. Dalla Mora, and A. Germani, A Luenberger-like observer for nonlinear systems, Int. J. Control, vol. 57, no. 3, pp. 537556, 1993. [3] F. Esfandiari and H. Khalil, Output feedback stabilization of fully linearizable systems, Int. J. Control, vol. 56, no. 5, pp. 10071037, 1992. [4] P. Ioannou and J. Sun, Stable and Robust Adaptive Control. Upper Saddle River, NJ: Prentice-Hall, 1995. [5] A. Isidori, Nonlinear Control Systems, 3rd ed. London, U.K.: Springer-Verlag, 1995. [6] A. Isidori, A. R. Teel, and L. Praly, A note on the problem of semiglobal practical stabilization of uncertain nonlinear systems via dynamic output feedback, Syst. Control Lett., vol. 39, pp. 165179, 2000. [7] M. Jankovic, Adaptive output feedback control of nonlinear feedback linearizable systems, Int. J. Adapt. Control Signal Processing, vol. 10, pp. 118, 1996. [8] H. Khalil, Nonlinear Systems, 2nd ed. Upper Saddle River, NJ: Prentice-Hall, 1996. [9] H. Khalil and F. Esfandiari, Semiglobal stabilization of a class of nonlinear systems using output feedback, IEEE Trans. Automat. Contr., vol. 38, pp. 14121415, Sept. 1993.