Optimality conditions for asymptotically stable control processes

OPTIMALITY CONDITIONS FOR ASYMPTOTICALLY STABLE CONTROL PROCESSES Fernando Lobo Pereira ∗,1Geraldo Nunes Silva ∗∗,2 ∗DEEC, FEUP, Rua Dr. Roberto Frias s/n, 4200-465 Porto, Portugal, [email protected] ∗∗ DCCE, IBILCE, UNESP, R. Cristovao Colombo, 2265, 15054-000, S.J. Rio Preto, SP, Brasil, [email protected] Abstract: In this article, we present and discuss the infinite horizon optimal control problem subject to stability constraints. First, we consider optimality conditions of the Hamilton-Jacobi-Bellman type, and present a method to define a feedback control strategy. Then, we address necessary conditions of optimality in the form of a maximum principle. These are derived from an auxiliary optimal control problem with mixed constraints. Keywords: Optimality Conditions, Optimal Control, Stability Constraints 1. INTRODUCTION In this article, we address the infinite horizon problem of optimizing a performance criterion by choosing control strategies whose trajectories are asymptotically stable. In a first stage, we state and discuss sufficient conditions of optimality in the form of an Hamilton-Jacobi-Bellman equation, and, based on them, we also present a method to synthesize a feedback control strategy. Then, we present necessary conditions of optimality in the form of a maximum principle and show how it can be derived from an auxiliary optimal control problem with mixed constraints. 1Partially supported by FCT under the research project “Constrained Optimal Control and Applications” and convenio FCT-FAPESP under the research project “Aplicacoes da Teoria do Controlo Impulsional e de Visao Computacional para Sistemas Roboticos Autonomos”. 2Partially supported by convenio FCT-FAPESP under the research project “Aplicacoes da Teoria do Controlo Impulsional e de Visao Computacional para Sistemas Roboticos Autonomos”. In many references, by optimal stabilization it is meant time-optimal stabilization, i.e. finding a control that steers the state of the system to the origin in minimum time. However, here, our goal is substantially different. Given a dynamic control system and a set in the state space containing at least one equilibrium point, we are interested in finding a feedback control strategy which stabilizes the system in the given set (in the sense that the corresponding trajectory converges asymptotically to an equilibrium point) and, simultaneously, minimizes the given cost functional. Notice that the value of the optimal cost depends not only on the equilibrium point, but also on the particular trajectory driving the system to this point. There has been a significant demand for results for this problem. A small sample of optimal stabilization application problems include micro-electromechanical (MEMS) control systems (Chase and Bhashyam, 1999), economic systems under a variety of constraints and assumptions, (Benigno and Woodford, 2004; Adam, 2002; Suescun, 1998), rigid body mechanical systems (El-Gohary, 2003), biological, medical, health care systems (Gomez and McLaughlin, 1991), and general chaotic systems (Basso et al., 1998), to name just a few. This contrasts with what appears to be a small body of results available for the general nonlinear dynamic optimization framework addressing the pertinent issues. See for example, (Ugrinovskii and Petersen, 1999) for results on the stabilization and minimax optimal control in the context of stochastic control systems, and (Prieur and Trelat, 2004) for a very specific problem and approach. The problem of stabilizing general dynamic nonlinear control systems has been receiving a considerable attention in the control literature, (Brockett, 1983; Clarke et al., 1997; Sontag, 1998; Clarke et al., 1998; Sontag, 1999; Prieur, 2000) and references cited therein. It has also emerged the important role of dynamic optimization and methods of nonsmooth analysis to derive stability results, see (Clarke et al., 1997; Clarke et al., 1998). However, to the best of our knowledge, no results have been derived for optimal control problems where control strategies are restricted to the subset of stabilizing ones. This article is organized as follows. In the next section, we introduce a precise and detailed statement of our problem. Then, in section three, the Hamilton-Jacobi-Bellman equation corresponding to this problem is presented followed in the ensuing section by the definition and pertinent results concerning the verification function. In section five, a mechanism of feedback synthesis based on a dynamic programming approach is presented and a result on the convergence of a sequence of sampled feedback control processes is discussed. In section six, we present and outline the derivation of the necessary conditions of optimality. Finally, some brief conclusions are presented. 2. PROBLEM STATEMENT Let us consider the following dynamic control system    ˙x(t) = f(x(t), u(t)), t ∈[τ, +∞) a.e., x(τ) = z, u∈ U, (1) where U:= {u(·)∈L∞[τ, ∞): u(t)∈Ω⊂IRma.e.} with Ω being a closed convex set. Assume that the closed set S⊂IRn, called target set, contains at least an asymptotic equilibrium of the former, i.e., ∃ξ∈Sand ¯u(·)∈U such that x¯u(·), the response of the system to the control ¯uwith x¯u(τ) = z, satisfies x¯u(t)→ξas t→ ∞. Let us consider the following optimal control problem: P∞(τ, z) Minimize g(ξ)+ ∞ Z τ e−δtf0(x(t), u(t))dt (2) subject to (1),and x(t)→ξas t→ ∞,(3) ξ∈S⊂IRn(4) Here, the constant δ > 0 is the discount rate, f0:IRn×IRm→IR and f:IRn×IRm→IRn are given functions, and S⊂IRnis a closed set. This is an infinite time horizon problem in which the optimization is taken over arcs xsuch that x(t) converges to an equilibrium point, ξ, of (1), where ξis also a choice variable. The pair (x, ξ) satisfying the constraints (1), (3), and (4) is called an admissible process of P∞(τ, z). We sometimes refer to an admissible arc xleaving implicit the existence of ξ∈Ssuch that the pair (x, ξ) is an admissible process of P∞(τ, z). Now, we specify the sense of the convergence x(t)→ξas t→ ∞. By this we mean that, ∃γ>0, lim t→∞ t Z τ eγskx(s)−ξkds<∞. Therefore, the optimal trajectory has to approach an equilibrium point in the given set S. Our approach consists in considering a family of auxiliary optimal control problems where this asymptotic convergence constraint gives rise to a penalization term added to the cost function of the original problem, i.e., we consider the problem: (Pl ∞(τ, z)) Minimize g(ξ)+R∞ τe−δtf0(x(t), u(t))dt +R∞ τ+leγskx(s)−ξkds subject to (1),and ξ∈S⊂IRn. Note that we should have ∞ Z τ+l eγskx(s)−ξkds →0 as l→ ∞, thus recovering the original optimization problem without the explicit constraint. Then, we show how to construct an (almost) optimal feedback control for problem Pl ∞(τ, ξ). This framework allows us to construct stabilizing optimal feedback controls. In order to state the assumptions required by the data of our problem, let F:[0,∞)×IRn×IRm→ IRn+1 be defined by F(t, x, u) = ·e−δtf0(x, u) f(x, u)¸. They are as follows: H1) Fis continuous and locally Lipschitz in x. H2) There exists c > 0 such that F(t, x, u)∈c(1 + kxk)B, ∀(t, x)∈[0,∞)×IRn. H3) ∀(t, x)∈[0,∞)×IRnthe set F(t, x, Ω) is convex-valued. H4) The set Ω is compact. H5) gis lower semicontinuous. 3. HAMILTON-JACOBI-BELLMAN EQUATION In this section we present a number of preliminary concepts and results needed in order to construct an optimal solution of optimal stabilization problem by using the concept of verification function, which can be shown to be a solution to the so called Hamilton-Jacobi-Bellman (HJB) partial differential equation associated with the optimal control problem. Let H: [0,∞)×IRn×IRn→IR be the Hamiltonian function for this problem defined by H(t, x, η) := sup v∈Ω {hη, f(x, v)i+e−δtf0(x, v)}.(5) Then, the continuous function φ: [τ, ∞)×IRn→ IR is a viscosity solution to the HJB equation if φt(t, x)−H(t, x, −φx) = 0,∀(t, x)∈[τ, ∞)×IRn, wherever ∇tw(t, x)−H(t, x,−∇xw(t, x))½≤0∀(t, x)∈A− φ−w ≥0∀(t, x)∈A+ φ−w for any C1function w:IR ×IRn→IR. Here A+ φ−wand A− φ−wdenote, respectively, the argmax and the argmin of the function (φ−w)(·,·) in [0,∞)×IRn. This solution concept satisfies the uniqueness and nonsmoothness requirements of the generalized solution to the HJB equation, but a characterization of an extended valued, lower semicontinuous solution is needed when endpoint state constraints are present. So, now we introduce the proximal sub-gradient. The proximal sub-gradient of Φ at (t, x), denoted by ∂PΦ(t, x), is the set of all vectors (α, ξ)∈IRn+1 such that ∃σ > 0 and a neighborhood Uof (t, x) satisfying Φ(τ, y)≥Φ(t, x) + α(τ−t) + hξ, y −xi −σ(kτ−tk2+ky−xk2), (6) ∀(τ, y)∈U. Analogously, the proximal supergradient of Φ at (t, x), denoted by ∂PΦ(t, x), is the set of all vectors (α, ξ)∈IRn+1, such that ∃σ > 0 and a neighborhood Uof (t, x) satisfying Φ(τ, y)≤Φ(t, x) + α(τ−t) + hξ, y −xi −σ(kτ−tk2+ky−xk2), (7) ∀(τ, y)∈U. The proximal super-gradient can also be defined by ∂PΦ(t, x) = −∂P(−Φ)(t, x). A lower semicontinuous function v: [τ, ∞)× IRn→IR ∪ {+∞} is a proximal solution to the HJB equation if ∀(t, x)∈[τ, ∞)×IRn, such that ∂Pv(t, x)6=∅, η0− H(t, x, −η) = 0,∀(η0, η)∈∂Pv(t, x).(8) There are well known results in the literature providing a characterization of the value function, V:IR ×IRn→IR, for an optimal control problem (for our problem in this article V(τ, z) := Inf{P∞(τ, z)}) as a generalized lower semicontinuous solution to the HJB equation (see for example Theorem 12.3.7 in (Vinter, 2000)). Such a result was derived for the infinite time horizon in (Baumeister et al., submitted in 2005). Clearly, invariance type results provide more detailed information on optimal control processes than this characterization of the value function and thus we proceed with the definition and properties of verification functions in the next section. 4. LOCAL VERIFICATION FUNCTIONS Next, we provide a result, standard in Dynamic Programming, for the conventional optimal control problem. In fact, we extend the concept of local verification function for this new problem formulation and provide conditions under which the existence of a verification function for a reference process (¯x, ¯ ξ, ¯u) is necessary and sufficient for its optimality. Let ¯xbe an admissible arc of problem P∞(τ, z). Let T(¯x, ²) be a tube centered at ¯xdefined by T(¯x, ²) := {(t, x)∈[τ, ∞)×IRn:kx−¯x(t)k ≤ ²}. A function φ:T(¯x, ²)→IR ∪+∞is a lower semicontinuous local verification function for (¯x, ¯ ξ, ¯u) if φis lower semicontinuous and the following conditions are satisfied. (1) ∀(t, x)∈int T(¯x, ²) such that ∂Pφ(t, x)6=∅, η0+ min u∈Ω{hη, f(x, u)i+e−δtf0(x, u)} ≥ 0, ∀(η0, η)∈∂Pφ(t, x). (2) lim inf t→∞ φ(t, ξ)≤g(ξ)+ ∞ Z τ e−δtf0(x(t), u(t))dt, ∀ξ∈Sand admissible control process (x, u). (3) lim inf t↑∞,ξ0→ξ φ(t, ξ0) = lim inf t↑∞ φ(t, ξ) for all ξ∈ S∩[¯ ξ+²B]. (4) φ(τ, z) = g(¯ ξ) + ∞ Z τ e−δtf0(¯x(t),¯u(t))dt. We have the following necessary and sufficient conditions of optimality. Theorem. Let (¯x, ¯ ξ, ¯u) be an admissible process of problem P∞(τ, z). Assume that the basic hypotheses (H1) −(H5) hold. We have the following: (1) If there exists a lower semicontinuous local verification function for (¯x, ¯ ξ, ¯u), then this control process is a strong local minimizer for P∞(τ, z). (2) Conversely, if |g(ξ)|+¯¯¯ ∞ Z τ f0(x(t), u(t))dt¯¯¯is bounded for all admissible processes (x, ξ, u), and (¯x, ¯ ξ, ¯u) is a strong local minimizer of P∞(τ, z), then there exists a lower semicontinuous local verification function for (¯x, ¯ ξ, ¯u). The proof is a slight modification of a similar result for finite time interval problems in (Vinter, 2000). Take a large T > 0, with T > τ +l, and consider the approximate problem PT(τ, z) of P∞(τ, z). Min J(x, u) subject to (1),and x(T)∈S. Here Bis the unit ball of IRnand J(x, u) is g(x(T))+ T Z τ e−δt f0(x(t), u(t))dt+ T Z T−l eγskx(s)−x(T)kds. The rationale behind this approximating problem is that, when T→ ∞, the last term in the cost function goes to zero and we can show that there is subsequence x(Tk) converging to some ξ∈S. Now, we define verification function for a process (¯x, ¯u) of problem PT(τ, z). The function φ: T(¯x, δ)→IR∪{∞} is a lower semicontinuous local verification function for ¯xwith parameter δ > 0 if φis lower semicontinuous and satisfies: a) ∀(t, x)∈intT(¯x, δ) such that ∂Pφ(t, x)6=∅, η0+ min u∈Ω{hη, f(x, u)i+f0(x, u)} ≥ 0 (9) ∀(η0, η)∈∂P φ(t, x). b) φ(T, ζ)≤g(z),∀ζ∈S. c) lim t0↑T,x0→ζ inf φ(t0, x0)=φ(T, ζ),∀ζ∈S∩[¯x(T)+δB]. d) φ(τ, z)=g(¯x(T))+ T Z τ e−δtf0(¯x(t),¯u(t))dt + T Z T−l eγsk¯x(s)−¯x(T)kds. We have the following result which can be found in (Vinter, 2000). Theorem. Let (¯x, ¯u) be an admissible process of problem PT(τ, z). Assume that the basic hypotheses (H1) −(H5) hold. We have the following: (1) If there exists a lower semicontinuous local verification function for (¯x, ¯u), then this control process is a strong local minimizer for PT(τ, z). (2) Conversely, if (¯x, ¯u) is a strong local minimizer of PT(τ, z) and |RT τf0(¯x(t),¯u(t))dt + RT T−leγskx(s)−x(T)kds|+|g(¯x(T))|is bounded for all admissible processes (¯x, ¯u), then there exists a a lower semicontinuous local verification function for (¯x, ¯u). 5. FEEDBACK SYNTHESIS Here, we present and discuss an algorithm for feedback control synthesis for problem P∞(τ, z) that, essentially, is a version of the procedure in (Vinter, 2000) modified in order to force the state to reach the target set S. A partition π={tk}of [τ, ∞) is a countably, strictly increasing sequence tksuch that ti> tj, whenever i > j,tk→ ∞ as k→ ∞. The diameter of π, denoted by hπ, is defined by sup k≥0 {∆k}, where ∆k=tk+1 −tk. Let us assume that τ= 0. Let φbe a given local verification function as defined in the previous section and let x∈IRn be a given state. Define U(x) := {u∈Ω,hNP S(pS(x)), f(x, u)i ≤ 0} where pS(x) is the proximal point of xat S. Let us start with x(0) = x0. Then, an approximating optimal control process is constructed recursively by computing a piecewise constant control function given, for each k= 0,1, . . . by ¯uπ k∈arg max u∈U(xπ(tπ k))nφ³tπ k, xπ(tπ k)+∆kf(xπ(tπ k), u)´ +∆kf0(xπ(tπ k), u)o and the corresponding trajectory is obtained by integrating the dynamics differential equation with the boundary condition given by the last value of the state variable in the previous time subinterval of the partition. Namely, xπ(t) is defined on [tπ k, tπ k+1) as the solution of ˙x(t) = f(t, x(t),¯uπ k) a.e.t∈(tπ k, tπ k+1], with initial value x(tπ k) given by the value of the state variable in the previous interval. We have the following main result of this work. Theorem. Assume that (H1)−(H5) hold. Let φbe a lower semicontinuous solution to the HamiltonJacobi-Bellman equation. Take (xπ, uπ), the control process obtained by the recursive procedure described above. Then, xπhas a cluster point 3 with respect to the topology of uniform convergence on compact intervals, and, associated with such a point x(·), there is a pair, control u(·) and limit point ξ, such that (x(·), ξ, u(·)) is an optimal process of P∞(0, x0). Here, we just outline the proof. Given N∈IN, there exist a partition π(N), and a corresponding process (xπ, uπ) constructed by the procedure described above, such that (xπ, uπ) restricted to the time interval [0, TN] is admissible for the slightly perturbed problem PTN(0, x0), defined by: (PN) Minimize J(x, u) subject to (1),and x(TN)∈S+1 NB. 3A cluster point of a given sequence is a point to which there is a convergent subsequence. Our aim is to extract the limit as N→ ∞ to recover the original problem. So, we assume that the sequence TNsatisfies TN→ ∞ as N→ ∞. Following arguments in (Vinter, 2000), it is possible to show that, under assumptions (H1)−(H5), {xπ}(·) has a cluster point, ¯xN(·), on [0, TN]. Moreover, by using Filippov’s selection theorem, the existence of ¯uNsuch that (¯xN(·),¯uN(·)) is an optimal process for PNis asserted. Since ¯xN(TN)∈S+(1/N)Band Sis compact, there exists a subsequence of ¯xN(TN) converging to some point ξ∈S. We denote it by {¯xN k}and consider the corresponding processes (¯xN k(·),¯uN k(·)) on the respective intervals [0, Tk], k= 1,2, . . ..Tkis a subsequence of TN. Now, we show how to obtain the optimal control process for the original problem. Restrict (¯xN k(·),¯uN k(·)) to [0, T1]. Again it is possible to show that ¯xN k(·) has a cluster point, x1(·), in the uniform convergence topology on [0, Tk] and there exists a corresponding control function u1(·). Now, by considering this subsequence (¯xN k(·),¯uN k(·)) restricted to [0, T2] for k= 2,3, . . . and repeating the same argument as before, we can find a process (x2(·), u2(·)) satisfying (1) and (2) restricted to [0, T2], in which (x2(·) is a cluster point of (¯xN k(·). By continuing this process, for all k∈N, we can show (xk(·), uk(·)) satisfying (1) and (2) restricted to [0, Tk], in which (xk(·) is a cluster point of (¯xN k(·). For each T > 0, there exists k∈Nsuch that T∈[Tk−l, Tk]. Define (¯x(·),¯u(·)) : [0,∞)→IRn×IRmby (¯x(t),¯u(t)) ≡(xk(t), uk(t)) t∈[0, T].(10) Lemma. We have the following. (1) The function given by (10) is well defined. (2) (¯x(t),¯ ξ, ¯u(t)) is an optimal process for P∞(0, x0). By construction (¯x(t),¯u(t)) is well defined. The second assertion follows from the existence of φ and from the first theorem in the previous section. 6. NECESSARY CONDITIONS OF OPTIMALITY Consider problem (P∞(τ, z)) formulated in section 2. In order to derive the necessary conditions we specify further the constraint of asymptotic convergence, x(t) to ξas t→ ∞. We impose that the rate of asymptotic convergence is not smaller than some given positive number γ. It is not difficult to see that this condition can be expressed as the following inequality mixed constraint h(x, u) := xTf(x, u) kxk2+γ≤0. Let us fix τ= 0 and z=x0, and consider the following optimal control problem (P) MinimizeJ(u) (11) subject to ˙x(t) = f(x(t), u(t)) L−a.e.(12) x(0) = x0, x(t)→ξ∈S(13) h(x(t), u(t)) ≤0∀t≥0 (14) u(t)∈Ω∀t≥0 (15) where J(u) := g(ξ) + ∞ Z 0 f0(x(t), u(t))dt. Notice that it is enough to specify u∗since it will follow that x∗(t)→ξ∗∈Sas t→ ∞. Let us state now our optimality conditions. Let the Pontryagin function or pseudo-Hamiltonian be defined as H(x, p, q, λ, u) := pTf(x, u)+qh+(x, u)+λf0(x, u)) where h+(x, u) = max{0, h(x, u)}. Theorem. Let (x∗, u∗) be an optimal control process for problem (P). Then, there exists an absolutely continuous function p: [0,∞)→IRn, a monotonically increasing function q: [0,∞)→IR, and a number λ≥0 satisfying: −˙p(t)∈∂xH(x∗(t), p(t), q(t), λ, u∗(t))(16) −lim s→∞ p(s)∈λ∂xg(ξ∗) + NS(ξ∗) (17) q(t) = −Z [t,∞) dν(s) (18) lim s→∞ q(s) = 0 (19) u∗(t) maximizes the mapping (20) v→H(x∗(t), p(t), q(t), λ, v) on Ω.(21) Here, dν is an arbitrary positive measure supported on [0,∞). Note that, by denoting the generalized gradient (in the sense of Clarke, see (Clarke et al., 1998)) by ∂f, we have ∂xh+(x, u) = ½{0}if h(x, u)<0 co{0, ∂> xh(x, u)}if h(x, u) = 0, where ∂> xh(x, u) = lim x0→xsup h(x0,u)>0 ∂xh(x0, u). We will need a refinement of the previously stated assumptions on the data of the optimal control problem H1) The functions g,f0,fand hare Lipschitz continuous in xuniformly w.r.t. all other variables. H2) The functions f0,fand hare Borel measurable w.r.t. the control variable. H3) The sets S∈IRnand Ω ∈IRmare closed and bounded. H4) There is at least one equilibrium point in S. Now, we outline the proof. We consider the following steps: a) The result is proved for an auxiliary optimal control problem with mixed constraints and a certain finite time T, denoted by (PT). b) The infinite horizon is regarded as the limit of the conditions for finite time for the problem (PT). Given an optimal control process for the infinite time horizon, its truncation to some finite interval [0, T ] for Tsufficiently large is proved to be an almost minimizer of the auxiliary finite time optimal control problem. Then, Ekeland’s variational principle is applied and the necessary conditions of optimality proved in a) are applied. Finally, limits are extracted in order to get the desired result. c) By extending the dynamics, we produce another auxiliary problem (QT) exhibiting only conventional state constraints and without mixed constraints. Well known necessary conditions of optimality in the form of a maximum principle, (Vinter, 2000), can be readily written down for (QT) and expressed in terms of the data of (PT). 7. CONCLUSIONS Here we presented and discussed an infinite time horizon control optimization problem in which a given objective functional is optimized by choosing control strategies which ensure the stabilization of the dynamic control system within a given target set with respect to which the system is invariant. Therefore, the trajectory associated to the optimal control process converges asymptotically to an optimal equilibrium within a given target set. We provided a dynamic programming based algorithm which yields a control process defined in a feedback form that approximates the optimal process. The method proposed here is modification of previous construct in (Rowland and Vinter, 1991) for a simpler problem. In this article, the model is finite time interval and there are no target set or set constraints. We also present necessary conditions of optimality in the form of a maximum principle for an optimal control process satisfying a prescribed minimum rate for the asymptotic convergence towards the optimal equilibrium point in a given target set. REFERENCES Adam, K. (2002). Optimal stabilization policy when the private sector has information processing constraints. Number 81. Centre for Studies in Economics and Finance, Dept of Economic Sciences - University of Salerno. Basso, M., R. Genesio, L. Giovanardi, A. Tesi and G. Torrini (1998). Stabilization of periodic orbits via time delayed feedback control. Int. J. of Bifurcation and Chaos 8, 1699–1706. Baumeister, J., A. Leitao and G. N. Silva (submitted in 2005). On the value function for control problems with infinite horizon. IMA Journal of Mathematical Control and Information. Benigno, P. and M. Woodford (2004). Optimal stabilization policy when wages and prices are sticky: The case of a distorted steady state. pp. 125–247. NBER Working Paper n 10893. Brockett, R. W. (1983). Asymptotic stability and feedback stabilization. In: Differential Geometric Control Theory (R. Millman R. Brockett and H. Sussmann, Eds.). pp. 181–191. Birkhauser, Boston. Chase, J. G. and S. Bhashyam (1999). Optimal stabilization of plate buckling. Smart Materials and Structures 8, 204–211. Clarke, F., Y. Ledyaev, E. Sontag and N. Subbotin (1997). Asymptotic controllability implies feedback stabilization. IEEE Trans. Automat. Control 42, 1394–1407. Clarke, F., Y. Ledyaev, R. Stern and P. Wolenski (1998). Nonsmooth analysis and control theory. Springer-Verlag. New York. El-Gohary, A. (2003). Optimal stabilization of a rigid body motion using rotors system. Appl. Math. and Computation 136, 229–239. Gomez, C. R. and J. McLaughlin (1991). New method for optimal stabilization of the transcranial doppler transducer during monitoring: normal controls and patient application. J. Neuroimaging 1, 91–94. Prieur, C. (2000). A robust globally asymptotically stabilizing feedback: The example of the artstein ’s circles. In: Nonlinear Control in the Year 2000 (W. Respondek A. Isidori, F. Lamnabhi-Lagarrigue, Ed.). Prieur, C. and E. Trelat (2004). Robust optimal stabilization of the brockett integrator via a hybrid feedback. In: Proceedings of MTNS 2004. Leuven, Belgium. Rowland, J. and R. B. Vinter (1991). Construction of optimal feedback controls. Systems and Control Letters 16, 337–367. Sontag, E. (1998). Mathematical Control Theory. Springer-Verlag. New York. Sontag, E. (1999). Stability and stabilization: Discontinuities and the effect of disturbances. In: Nonlinear Analysis, Differential Equations, and Control. pp. 551–598. Kluwer. Dordrecht. Suescun, R. (1998). Optimal commodity price stabilization over the business cycle. In: Procs IEEE Conference on Decision and Control. Banco de la Republica. Bogota, Colombia. Ugrinovskii, V. and I. Petersen (1999). Absolute stabilization and minimax optimal control of uncertain systems with stochastic uncertainty. SIAM J. Control Optim. 37, 1089– 1122. Vinter, R. B. (2000). Optimal Control: Foundations and Applications. Birkhauser. Boston.