Newton-Like Extremum-Seeking Part I: Theory

Size: px
Start display at page:

Download "Newton-Like Extremum-Seeking Part I: Theory"

Transcription

1 Joint 48th IEEE Conference on Decision and Control and 28th Chinese Control Conference Shanghai, P.R. China, December 16-18, 29 ThAIn1.9 Newton-Like Extremum-Seeking Part I: Theory William H. Moase, Chris Manzie and Michael J. Brear Abstract In practice, the convergence rate and stability of perturbation based extremum-seeking ES schemes can be very sensitive to the curvature of the plant map. This sensitivity arises from the use of a gradient descent adaptation algorithm. Such ES schemes may need to be conservatively tuned in order to maintain stability over a wide range of operating conditions, resulting in slower optimisation than could be achieved for a fixed operating condition. This can severely reduce the effectiveness of perturbation based ES schemes in some applications. It is proposed that by using a Newton-like step instead of a more typical gradient descent adaptation law, then the behaviour of the ES scheme near an extremum will be independent of the plant map curvature. In this paper, such a Newton-like ES scheme is developed and its stability and convergence properties are explored. I. INTRODUCTION Consider a plant with an output, y, which is an unknown function of an input,. The goal of extremum-seeking is to find the input,, which minimises or maximises y using only measurements of the output. Without loss of generality, this paper will deal with the problem of minimising y. Furthermore this paper will deal only with the problem of seeking a local minimum of y. Fig. 1 shows a basic schematic of sinusoidally perturbed ES. The control input is the superposition of a slowly changing component,, and a small dither signal,asinω t. The dither signal is used to determine ˆy, where y = dy/d, is a quantity evaluated at =, and ˆ is an estimate. The exact method for estimating y varies between different schemes but requires to be slowly changing compared to the dither signal. With an estimate ofy available, can be driven towards using an approximated gradient descent law, d dt = ω k ˆy. 1 Sinusoidally perturbed ES schemes were amongst the first adaptive controllers developed and were popular in the 195s and 196s [1]. In 2, there was a resurgence of interest in ES, largely due to the development of the first stability analysis of sinusoidally perturbed ES on a general, nonlinear plant in [2]. In the same year, a number of other papers on sinusoidally perturbed ES were published [3] including extensions to multi-parameter optimisation [4], [5]. There have been a number of developments in sinusoidally perturbed ES since 2 including: extension to discrete This research was partially supported under Australian Research Council s Discovery Projects funding scheme project number DP W. H. Moase, C. Manzie and M. J. Brear are with the Department of Mechanical Engineering, The University of Melbourne, 31, Victoria, Australia moasew@unimelb.edu.au, manziec@unimelb.edu.au, mjbrear@unimelb.edu.au Fig Plant a sinω t y Gradient estimator k ω Basic schematic of sinusoidally perturbed ES. time [6]; semi-global stability results [7]; the use of periodic non-sinusoidal dither signals [8]; the use of a time-dependent dither signal amplitude [9]; and the development of stochastically perturbed ES [1]. As a result of using an approximated gradient descent adaptation law, the local convergence speed of a perturbation based ES scheme is typically proportional to y, where is a quantity evaluated at = and y = d 2 y/d 2. This dependence is evident in the mathematical analysis of particular ES schemes [11], and can severely reduce the effectiveness of perturbation based ES in applications with a wide range of plant behaviours [12]. If the operating condition changes, then an increase in y may destabilise the scheme since the rate of change of must remain sufficiently small in order to estimate y whereas a decrease in y will result in a reduced rate of convergence. This curvature dependence may be reduced by introducing further compensators to the ES scheme [11]. By appropriate selection of these compensators, the behaviour of the averaged, linearised closed-loop system can be tuned. This tuning could aim to increase the speed at which the extremum is tracked or, of more interest here, it could reduce the sensitivity of the closed-loop system to perturbations iny by using, for example, H techniques. Although this reduced sensitivity to y is an improvement over more simple ES schemes, this technique requires an a priori estimate of y and is most effective for small variations in y. Some schemes are capable of seeking minima in a fashion which is independent of the curvature at that minimum. One such scheme is the discrete-time triangular search algorithm [13] which uses information from previous steps rather than a dither signal in order to determine its next step. The magnitude of each step is not based on any estimate of the gradient, so the performance of the scheme is not related to the magnitude of y. This is evident in [14], where triangular search and sinusoidally perturbed ES are compared in the role of minimising the thermoacoustic limitcycle pressure oscillations in a gas-turbine combustor. The adaptation gain, k, used in the sinusoidally perturbed ES ŷ /9/$ IEEE 3839

2 scheme had to be changed between operating conditions, whereas triangular search required no such tuning. The Kiefer-Wolfowitz KW algorithm [15] is a popular stochastic approximation scheme for online minimisation. It uses a gradient descent law, n+1 = n k nˆy n, where n denotes a value at the n-th step of the algorithm and k n is a sequence of positive numbers satisfying certain, wellknown, conditions which include k n as n. The gradient is estimated using a finite difference which is performed over an interval [ n a n, n + a n ] where a n as t. There have been a number of extensions to the KW algorithm which include multi-variable optimisation [16], [17], [18], methods for handling constrained optimisation problems [19], and a deterministic result for non-smooth optimisation [2]. Since the KW algorithm, and its mentioned extensions, use a gradient descent algorithm, optimal selection of the series k n and a n requires knowledge of y [21]. This problem is largely solved through the use of a Newton step [22], [23]. In the one-dimensional case, this involves estimating both y and y before progressing n an amount proportional toy /y. In a region near the extremum, the Newton step approximates the difference between the current input and the optimal input, so in the case of perfect estimation of y and y, it has a local rate of convergence which is independent of y. Newton-like SA and triangular search achieve convergence independently of y, however, they lack one major advantage offered by perturbation based ES: the ability to achieve convergence to the extremum on a time-scale comparable to that of the plant dynamics [1] instead, they require the plant dynamics to settle between each step of the algorithm. This motivates the present work: to develop a sinusoidally perturbed ES scheme which uses a Newton-like step. The proposed scheme is described in Section II, and as well as using a Newton-like step, features a dither signal amplitude schedule DSAS. This idea is similar to the shrinking interval used in the KW algorithm for estimating the gradient, or the time-varying dither signal amplitude for ES used in [9]. By initialising a to be large and reducing it as the extremum is approached, then fast convergence rates and accurate convergence are simultaneously achievable. The most significant difference between the proposed DSAS and the aforementioned schemes is the ability for the DSAS to increase the dither signal amplitude should change after some time. In Section III, it is shown that the proposed scheme is stable for a noiseless plant with no dynamics and that arbitrarily accurate convergence to the extremum can be achieved. For a plant with linear time-invariant LTI input and output dynamics, conditions for local stability of the ES scheme are given, and it is shown that the local rate of convergence is independent of y. II. PROPOSED SCHEME Fig. 2 shows a schematic of the proposed scheme. The plant is subject to the input = +asinω t, 2 sinω t + Fig. 2. a Plant y DSAS Gradient estimator Adaptation law aŷ Basic schematic of proposed ES. a 2ŷ where ω >. is progressed according to the adaptation law, and at can be progressed according to the DSAS or may simply be set to a constant, a = a min, where a min >. Meanwhile, the gradient estimator determines ˆy and ˆy for use in the adaptation law. A. Adaptation law Let =, and consider the Taylor expansions, y = y +O 2, y = y +O. 3 It then becomes apparent that the local small behaviour of a regular gradient descent law would yield a rate of change of which is proportional to y, whereas a Newton step would yield a rate of change which is proportional to. Therefore the local behaviour of an adaptation law using a Newton step will be independent of y unlike a regular gradient descent law. A more practical alternative to a Newton-step is, d /ˆy dt = k ω ˆy if ˆy < δa minˆy, 4 k ω δa min sgn ˆy otherwise, where δ,k > are dimensionless quantities. is progressed at a rate corresponding to an approximated Newton step only when δa minˆy > ˆy. Otherwise, is progressed at a rate corresponding to a sign-of-gradient descent. This latter behaviour has two purposes. First of all, because the Newton step seeks satisfying ˆy =, then it may seek a maximum or inflection point instead of a minimum. However, at a maximum or inflection point, y, so sign-of-gradient descent behaviour will instead be followed, giving a more desirable result. The second purpose of the sign-of-gradient descent behaviour is to saturate the rate of change of at ±k ω δa min. This avoids singular behaviour ˆy of the Newton step as, and also makes it possible to ensure changes slowly compared to the dither signal. As will become apparent in the next subsection, the latter of these properties is important for the estimation of y and y. B. Gradient estimator Consider a plant with no dynamics subject to the input as defined in 2. Taking the Taylor series expansion of y about = gives y = y +y asinω t+ 1 2 y a 2 sin 2 ω t+h, 5 384

3 where h is terms of third and higher order in a. Alternatively, the plant output may be represented in state-space: y = Cx+h, dx dt = ω Ax+ x d dt + x da a dt, where C = , y a2 y y as 1 x = y ac 1 y a 2 S 2, A = 1 1 2, y a 2 C 2 2 6a 6b y n = d n y/d n, S n = sinnω t and C n = cosnω t. This system can then be estimated with the state-space observer: ˆy = Cˆx, dˆx dt = ω Aˆx+ω Ly ˆy, 7a 7b where L R 5 is a non-dimensional gain vector. Thus y and y can be estimated from: aˆy = Cˆx, C = S 1 C 1, 7c a 2ˆy = Cˆx, C = S 2 C 2. 7d Since the observer does not explicitly account for the variation of a and with time, it is most effective when a and change slowly compared to the dither signal. Note that although this state-space observer is based on that used in [14], it has been extended to be capable of estimating y. C. Dither signal amplitude schedule As discussed in the previous subsection, the ability of the state-space observer to accurately estimate y and y relies on a and varying slowly compared to the dither signal. In order to increase the maximum allowable rates of change of a and, then one could simply increase ω. However, the maximum allowable value of ω depends upon plant noise and dynamics. If faster convergence of to was required than could be achieved by simply increasing ω, then a could be increased. However, even if the ES scheme was able to achieve perfect convergence of to, then the superposition of a large dither on the plant input would result in large fluctuations in y about its minimum. Instead, it would be ideal to have large a when the rate of change of is large, and small a when the rate of change of is small. The most simple way of achieving such behaviour would be to let a d /dt, however, this is impractical since: from 4 and 7c 7d, a circular algebraic relationship arises between a and d /dt; and noise in the measurement of y will prevent the use of an arbitrarily small dither signal amplitude. Instead, it is proposed that a be related to d /dt by the differential equation, da dt = k aω α a, where k a >. Therefore at any given instant, 8 attempts to drive a towards α, which is given by 1 d α = a min +a min χ, 9 a min ω k dt where χ: R R > and χz γz for all z and some γ. The quantities, a min, k and ω, are defined in 2 and 4. Remark 1: If there was no noise on y and was a constant, then a min could be made arbitrarily small, however, a min should be finite in any practical scenario. Generally χ should be chosen to be some function which increases with d /dt so that a scales, in some sense, with d /dt. A. Plant with no dynamics III. STABILITY RESULTS Assumption 1: The plant see Fig. 2 has no dynamics, and is given by a time-invariant function y : R R with y >. Furthermore, there exists a domain D containing the origin such that for all D : y + is twice continuously differentiable with respect to ; and sgny + = sgn. Assumption 2: A LC is Hurwitz. Remark 2: Assumption 2 can be satisfied by appropriate selection of L. Theorem 1: Under Assumptions 1 and 2, for any compact set D D R 5 containing the origin, there exist a min >, k > and ka >, such that for all a min,k,k a,a min,k,k a, the solution t, at of system 2, 4, 7a 9, with initial conditions a,, x [,γδa min ] D satisfies limsup t = O t limsup at = O t a 2 miny 3 1 /y a 2 miny 3 1 /y, 1a, 1b where a = a a min, x = ˆx x, and 1 is some input satisfying 1 a. Furthermore, if R satisfies the conditions placed on D in Assumption 1, then 1a and 1b can be satisfied for arbitrarily large and x. Proof: Consider the non-dimensional parameters, = a min, a = a a min, x = x a 2 min y, t = ω t. Letting A x = A LC, then system 2, 4, 7a 9 can be expressed in non-dimensional terms as d d t = k f, a, x, t, 11a d a d t = k a χ f, a, x, t a, 11b d x d t = A 1 x d x x+l h d t + x d a, 11c a d t

4 where h = h/a 2 min y, / f, a, x, t ˆy = δsgn ˆy a minˆy if ˆy < a min δˆy otherwise, ˆy = y +C x1+ a 1, ˆy = y +y C x1+ a 2, Under Assumption 2 there exists a symmetric, positive definite matrix, P, which satisfies the Lyapunov equation, PA x +A T xp = I. Let V = y y k a 2 +γ +1 min y, k + a k a +μ x T P x, 12 where μ >. The domain of interest for this proof is restricted to a min,k,k a, a,, x,a min,k,k a [,γδ] D, where D D R 5 is a compact set containing the origin and D = {z/a min : z D }. By Assumption 1, V is a positive definite function of, a, x and is, therefore, a suitable Lyapunov candidate function for system 2, 4, 7a 9. However, it is still necessary to explore the conditions under which dv/d t <. Letting denote the L 2 norm, where dv d t = μ h LT P x xt P x +ξ +ξa a+ξ x x, 13 ξ = g f, a,, t ξ a = μk a a+1y ξ x = g x 1 2 μ xt x x T P x, +μk C a ay P x y xt P x, [ 1 ] 2 S 2 C 2 P x 1, y xt P x and g x = g f, a, x, t g f, a,, t, x gz = g 1 z+g 2 z, [ y g 1 z = y a +γ +1sgn ] z +χz, min g 2 z = μ [ k z x +k a χz x ] T P x a xt P x. In the next part of this proof, it is demonstrated that there exists μ and sufficiently small k and k a such that ξ a,ξ,ξ x < over the entire domain of interest. It is quite obvious that ξ a < if μk a is sufficiently small, however, it is not so clear that ξ and ξ x can also be made negative. First, it is useful to note that sgnf, a,, t = sgn. Using this result and the fact that χz γz, then g 1 f, a,, t [ ] y f, a,, t y a +1 <. min Therefore ξ < if μk and μk a are sufficiently small. One can find a sufficiently large μ to ensure that ξ x < so long as g x is bounded over the domain of interest. This can be guaranteed if gf/ x n is bounded for n = 1,2,3,4,5. Note that gf/ x 1 =. When ˆy = and ˆy a minδˆy then gf/ x n = for n = 2,3,4,5. Furthermore, when ˆy < a minδˆy, it is a relatively simple matter to show { f ˆy x n y 1+ a 1 if n = 2,3, δ1+ a 2 14 if n = 4,5. Because f/ x n is bounded in all of these cases, then it looks promising that g x will also be bounded. However, there are two scenarios that require further discussion: gf/ x n contains a term N = f y x n a min y, 15 which is not obviously bounded as a min. However, the triangular inequality z 1 +z 2 z 1 +z 2 can be used to show that y a min y and when ˆy < a minδˆy, then y δˆy < y It follows that N is, in fact, bounded. ˆy + C x 1+ a, 16 + C x 1+ a. 17 = & ˆy. f/ x n is not bounded when ˆy Nonetheless, all of the terms in gf are bounded, most notably, f, a, x, t δ, and from 16, y /a min y C x/1 + a. It follows that g x is still bounded when ˆy = and ˆy unless such a situation can occur for infinitesimal x. This would require both y = and y. However, by Assumption 1, y = if and only if =, in which case y >. Therefore g x is bounded, and there exists sufficiently large μ such that ξ x <. Sinceξ,ξ a,ξ x < over the entire domain of interest, then dv/d t < unless, a, x = O h. Taking advantage of Assumption 1, then there exists an input, 1, such that h = a miny a 3 S1 3 6y, 18 and 1 a. Therefore, h is bounded and can be made arbitrarily small by decreasing a min. Thus, it is possible to ensure that dv/d t < except for a small domain containing the origin. Theorem 1 follows directly after transforming the system back into dimensional variables. Corollary 1: Under Assumptions 1 and 2, for any compact set D D R 5 containing the origin, there exist k > and a min >, such that for all a min,k,a min,k, the solution t of system 2, 4, 7a 7d with a = a min and initial conditions, x D satisfies 1a. Furthermore, if R satisfies the conditions placed on D in Assumption 1, then 1a can be satisfied for arbitrarily large and x. 3842

5 F i s i f F o s y Assumption 5: The dynamics F i s and F o s can be represented in state-space forms so that: Fig. 3. Plant Plant with LTI input and output dynamics. Remark 3: Theorem 1 demonstrates the stability of the proposed scheme with DSAS, whereas Corollary 1 demonstrates the stability of the scheme with a constant dither signal amplitude a = a min. In both cases, the influence of the ES scheme on the plant output may be found by taking the Taylor series expansion of y about = to gain limsupyt y = O y a 2 min. 19 t B. Plant with LTI input and output dynamics Up to this point, the analysis of the proposed ES scheme has considered the plant to be a static input-output map. If there were potentially nonlinear stable dynamics resulting in a time-invariant steady-state mapping of the input to the output, then it is expected that the dynamics will have only a small effect on the behaviour of the scheme provided that ω is sufficiently small. This choice of ω effectively ensures time-scale separation between the ES scheme and the plant dynamics a similar example of this can be seen in [2]. However, such a requirement limits the rate of convergence that can be achieved by the ES scheme. In this subsection, the local behaviour of the proposed scheme is studied when time-scale separation between the plant dynamics and the ES scheme is not guaranteed. To simplify matters, the influence of DSAS is not investigated. A slight modification of the proposed scheme is considered: lags φ 1 and φ 2 are introduced in the demodulation signals so that C and C appearing in 7c and 7d respectively are replaced with C = sinω t φ 1 cosω t φ 1, C = sin2ω t φ 2 cos2ω t φ 2. The purpose of these lags, as will become apparent in Theorem 2, is to compensate for lags due to the plant dynamics between the dither signal and the corresponding sinusoidal components of the plant output. Let s denote the Laplace variable and let the use of square brackets in the context Gs[ft] denote the time-domain output of the transfer function Gs when ft is its input. Assumption 3: The plant can be expressed as shown in Fig. 3, so that i = F i s[] and y = F o s[f i ], where F o s and F i s are LTI dynamics. Assumption 4: There exists a domain D R containing, such that for all z D, f z = f f z 2, with f >. 2 Furthermore, the dither amplitude, a >, is constant and sufficiently small to ensure that a, +a D. dx i dt = A dx o ix i +B i, = A o x o +B o f i, dt i = C i x i +D i, y = C o x o +D o f i, where A i and A o are Hurwitz and F i = F o = 1. Remark 4: Assumption 5 ensurees F i and F o have stable and proper LTI dynamics. The final part of Assumption 5 can be made without loss of generality. If F i = 1 and/or F o = 1, then it is a simple matter to transform Fi s F i s,f o s,f F i, F os F o,f if o f. Theorem 2: Let i = F i s[ ] and x = ˆx x where ] f a2 f F i iω f F o s [ 2 { ]} i af Im e iωt F i iω F o s+iω [ i { ]} x = af Re e iωt, F i iω F o s+iω [ i a 2 f Im { } e 2iωt F 2 a 2 f Re { } e 2iωt F 2 and are 2π/ω -periodic and F 2 = F 2 i iω F o 2iω. Also let x i = x i x 2π i x o = x o x 2π o solutions of d dt x2π d dt x2π Finally let where x 2π i and x 2π o i = A i x 2π i +B i +asinω t, o = A o x 2π o +B o f +Im { e iωt F i iω }. 1 Hs = s+k ω J sf i s, 21a J s = Re{ } F i iω F o s+iω e iφ1. 21b F 2 cosφ 2 +argf 2 Under Assumptions 2 5, i, x, x i, x o = is a locally exponentially stable equilibrium point of the system given by 2, 4 and 7a 7d provided that: s = ±iω are not zeroes of either F i s or F o s; s = ±2iω are not zeroes of F o s; cosφ 2 +argf 2 > ; k is sufficiently small; and the poles of Hs all have negative real parts. Proof: Due to space restrictions, only a sketch of a proof is provided here. A more complete proof may be found in [24]. Let i = i /a and x = x/a 2 f. Equations 7c and 7d can respectively be simplified to, ˆy af = C x+re { e iφ1 F i iω F o s+iω [ i ]}, 22a ˆy f = C x+f 2 cosφ 2 +argf 2. 22b If x =, then ˆy = f F 2 cosφ 2 + argf 2. Therefore, cosφ 2 + argf 2 must be positive for the adaptation law to locally follow a Newton step. Otherwise, the adaptation 3843

6 law would follow a sign-of-gradient descent and, at best, i would chatter about zero. Substituting 7a into 7b and 22a 22b into 4 and linearising each resulting equation about i, x, x i, x o = respectively yields d i d t = k F i J s [ i ] k F i s[c x] F 2 cosφ 2 +argf 2, 23 { [ ]} d x d t = A d i x x+lε x t Re ΓF o s+iω 24 d t where Γ = i 1 T F i iω expiω t and ε x t consists of terms that decay to zero independently of i and x i. Therefore, ε x t does not affect the local stability of the ES scheme and is ignored for the remainder of this analysis. If x = in 23, then the dynamics of i are given by Hs, and i converges to zero if the poles of Hs all have negative real parts. Similarly from 24, if i =, then x converges to zero under Assumption 2. Therefore 23 and 24 can be thought of as two interconnected systems which are independently stable. If the interconnections are sufficiently weak, then the ES scheme will remain stable. Theorem 9.2 in [25] quantifies the conditions under which the interconnections can be considered sufficiently weak. In this case, stability of the ES scheme can be guaranteed if k is sufficiently small. Theorem 2 follows directly. Remark 5: The local convergence of i to zero is independent of f since Hs is independent of f. Remark 6: Consider the quantity ˆk where ˆk k = J F i = F 1cosφ 1 +argf 1 F 2 cosφ 2 +argf 2, and F 1 = F i iω F o iω. If Ω := ˆk ω is sufficiently small, then Hs has a dominant pole at s = Ω +OΩ 2. This allows the final dot-point of Theorem 2 to be tested more easily, but effectively restricts the dynamics of i to be well-separated not only from the gradient estimator dynamics, but also from the plant dynamics. Remark 7: When φ 1 = φ 2 = and F i s = F o s = 1, Hs has a single pole at s = k ω. If Ω is small see Remark 6 and non-trivial dynamics are introduced, that pole is shifted to a location near s = ˆk ω. If ˆk <, then the dynamics will destabilise the system. This occurs because sgny = sgnˆy about the equilibrium point. It follows that the scheme will climb the plant map rather than descend it as intended. If ˆk < k, then slower convergence of i to zero will be observed than in the absence of dynamics. If ˆk > k, faster convergence will be observed, however, the region for which the adaptation law follows a Newton-step will decrease. For fixed ω, the influence of the pole shift from k ω to ˆk ω may be compensated through selection of a different value of k or through suitable selection of φ 1 and φ 2. IV. CONCLUSIONS An ES scheme using a Newton-like adaptation law and DSAS was developed and shown to achieve convergence of the control input to a small neighbourhood of the extremum from a potentially infinite domain of initial conditions. Furthermore, for a plant with LTI input and output dynamics, conditions for local stability of the ES scheme without DSAS were found and it was shown that local convergence of the control input is independent of y. In part II of this paper, the behaviour of the proposed ES scheme is further explored in both numerical and experimental tests. REFERENCES [1] K. B. Ariyur and M. Krstic, Real-Time Optimization by Extremum- Seeking Control. John Wiley & Sons, 23. [2] M. Krstic and H. Wang, Stability of extremum seeking feedback for general nonlinear dynamic systems, Automatica, vol. 36, pp , 2. [3] R. N. Banavar, D. F. Chichka, and J. L. Speyer, Convergence and synthesis issues in extremum-seeking control, in Proceedings of the American Control Conference, 2, pp [4] M. A. Rotea, Analysis of multivariable extremum seeking algorithms, in Proceedings of the American Control Conference, 2, pp [5] G. C. Walsh, On the application of multi-parameter extremum seeking control, in Proceedings of the American Control Conference, 2, pp [6] J.-Y. Choi, M. Krstic, K. Ariyur, and J. S. Lee, Extremum seeking control for discrete time systems, IEEE T. Automat. Contr., vol. 47, pp , 22. [7] Y. Tan, D. Nesic, and I. Mareels, On non-local stability properties of extremum seeking control, Automatica, vol. 42, pp , 26. [8] D. Nesic, Y. Tan, and I. Mareels, On the choice of dither in extremum seeking systems: a case study, in Proceedings of the IEEE Conference on Decision and Control, 26, pp [9] Y. Tan, D. Nesic, I. Mareels, and A. Astolfi, On global extremum seeking in the presence of local extrema, in Proceedings of the IEEE Conference on Decision and Control, 26, pp [1] C. Manzie and M. Krstic, Extremum seeking with stochastic perturbations, IEEE T. Automat. Contr., vol. 54, pp , 29. [11] M. Krstic, Performance improvement and limitations in extremum seeking control, Syst. Control Lett., vol. 39, pp , 2. [12] A. Banaszuk, K. B. Ariyur, M. Krstic, and C. A. Jacobsen, An adaptive algorithm for control of thermoacoustic instability, Automatica, vol. 4, pp , 24. [13] Y. Zhang, Stability and performance tradeoff with discrete time triangular search minimum seeking, in Proceedings of the American Control Conference, 2, pp [14] A. Banaszuk, Y. Zhang, and C. A. Jacobson, Adaptive control of combustion instability using extremum-seeking, in Proceedings of the American Control Conference, 2, pp [15] J. Kiefer and J. Wolfowitz, Stochastic estimation of a regression function, Ann. Math. Stat, vol. 23, pp , [16] J. R. Blum, Multidimensional stochastic approximation methods, Ann. Math. Stat, vol. 25, pp , [17] H. J. Kushner and D. S. Clark, Stochastic approximation methods for constrained and unconstrained systems. Springer-Verlag, [18] J. C. Spall, Multivariate stochastic approximation using a simultaneous perturbation gradient, IEEE T. Automat. Contr., vol. 37, no. 3, pp , [19] H. J. Kushner, Stochastic approximation algorithms for constrained optmization problems, Ann. Stat, vol. 2, no. 4, pp , [2] A. R. Teel, Lyapunov methods in nonsmooth optimization, part II: persistently exciting finite differences, in Proceedings of the IEEE Conference on Decision and Control, 2, pp [21] S. Yakowitz, P. L Ecuyer, and F. Vazquez-Abad, Global stochastic optmization with low-dispersion point sets, Oper. Res., vol. 48, no. 6, pp , 2. [22] V. Fabian, Stochastic approximation, in Optimizing Methods in Statistics. Academic Press, 1971, pp [23] J. C. Spall, Adaptive stochastic approximation by the simultaneous perturbation method, IEEE T. Automat. Contr., vol. 45, no. 1, pp , 2. [24] W. H. Moase, C. Manzie, and M. J. Brear, Newton-like extremumseeking for the control of thermoacoustic instability, IEEE T. Automat. Contr., In press. [25] H. Khalil, Nonlinear systems, 3rd ed. Prentice Hall,

Newton-Like Extremum-Seeking Part II: Simulations and Experiments

Newton-Like Extremum-Seeking Part II: Simulations and Experiments Joint 48th IEEE Conference on Decision and Control and 28th Chinese Control Conference Shanghai, P.R. China, December 16-18, 29 ThAIn1.1 Newton-Like Extremum-Seeking Part II: Simulations and Experiments

More information

A Systematic Approach to Extremum Seeking Based on Parameter Estimation

A Systematic Approach to Extremum Seeking Based on Parameter Estimation 49th IEEE Conference on Decision and Control December 15-17, 21 Hilton Atlanta Hotel, Atlanta, GA, USA A Systematic Approach to Extremum Seeking Based on Parameter Estimation Dragan Nešić, Alireza Mohammadi

More information

On sampled-data extremum seeking control via stochastic approximation methods

On sampled-data extremum seeking control via stochastic approximation methods On sampled-data extremum seeing control via stochastic approximation methods Sei Zhen Khong, Ying Tan, and Dragan Nešić Department of Electrical and Electronic Engineering The University of Melbourne,

More information

Extremum-Seeking for Adaptation of Urban Traffic Signal Control

Extremum-Seeking for Adaptation of Urban Traffic Signal Control Preprints of the 19th World Congress The International Federation of Automatic Control Extremum-Seeking for Adaptation of Urban Traffic Signal Control Ronny Kutadinata Will Moase Chris Manzie Lele Zhang

More information

Stochastic Optimization with Inequality Constraints Using Simultaneous Perturbations and Penalty Functions

Stochastic Optimization with Inequality Constraints Using Simultaneous Perturbations and Penalty Functions International Journal of Control Vol. 00, No. 00, January 2007, 1 10 Stochastic Optimization with Inequality Constraints Using Simultaneous Perturbations and Penalty Functions I-JENG WANG and JAMES C.

More information

Phasor Extremum Seeking Control with Adaptive Perturbation Amplitude

Phasor Extremum Seeking Control with Adaptive Perturbation Amplitude Phasor Extremum Seeking Control with Adaptive Perturbation Amplitude Khalid T. Atta, Roland Hostettler, Wolfgang Birk, and Andreas Johansson This is a post-print of a paper published in 55th IEEE Annual

More information

4. Complex Oscillations

4. Complex Oscillations 4. Complex Oscillations The most common use of complex numbers in physics is for analyzing oscillations and waves. We will illustrate this with a simple but crucially important model, the damped harmonic

More information

Semi-Global Stability Analysis of a Discrete-Time Extremum-Seeking Scheme using LDI Methods

Semi-Global Stability Analysis of a Discrete-Time Extremum-Seeking Scheme using LDI Methods 5nd IEEE Conerence on Decision and Control December 10-13, 013 Florence, Italy Semi-Global Stability Analysis o a Discrete-Time Extremum-Seeking Scheme using LDI Methods Rohan C Shekhar, William H Moase

More information

NONLINEAR SAMPLED DATA CONTROLLER REDESIGN VIA LYAPUNOV FUNCTIONS 1

NONLINEAR SAMPLED DATA CONTROLLER REDESIGN VIA LYAPUNOV FUNCTIONS 1 NONLINEAR SAMPLED DAA CONROLLER REDESIGN VIA LYAPUNOV FUNCIONS 1 Lars Grüne Dragan Nešić Mathematical Institute, University of Bayreuth, 9544 Bayreuth, Germany, lars.gruene@uni-bayreuth.de Department of

More information

Extremum Seeking for Dead-Zone Compensation and Its Application to a Two-Wheeled Robot

Extremum Seeking for Dead-Zone Compensation and Its Application to a Two-Wheeled Robot Extremum Seeking for Dead-Zone Compensation and Its Application to a Two-Wheeled Robot Dessy Novita Graduate School of Natural Science and Technology, Kanazawa University, Kakuma, Kanazawa, Ishikawa, Japan

More information

ADAPTIVE EXTREMUM SEEKING CONTROL OF CONTINUOUS STIRRED TANK BIOREACTORS 1

ADAPTIVE EXTREMUM SEEKING CONTROL OF CONTINUOUS STIRRED TANK BIOREACTORS 1 ADAPTIVE EXTREMUM SEEKING CONTROL OF CONTINUOUS STIRRED TANK BIOREACTORS M. Guay, D. Dochain M. Perrier Department of Chemical Engineering, Queen s University, Kingston, Ontario, Canada K7L 3N6 CESAME,

More information

A Novel Integral-Based Event Triggering Control for Linear Time-Invariant Systems

A Novel Integral-Based Event Triggering Control for Linear Time-Invariant Systems 53rd IEEE Conference on Decision and Control December 15-17, 2014. Los Angeles, California, USA A Novel Integral-Based Event Triggering Control for Linear Time-Invariant Systems Seyed Hossein Mousavi 1,

More information

Robust Stabilization of Non-Minimum Phase Nonlinear Systems Using Extended High Gain Observers

Robust Stabilization of Non-Minimum Phase Nonlinear Systems Using Extended High Gain Observers 28 American Control Conference Westin Seattle Hotel, Seattle, Washington, USA June 11-13, 28 WeC15.1 Robust Stabilization of Non-Minimum Phase Nonlinear Systems Using Extended High Gain Observers Shahid

More information

12.1. Exponential shift. The calculation (10.1)

12.1. Exponential shift. The calculation (10.1) 62 12. Resonance and the exponential shift law 12.1. Exponential shift. The calculation (10.1) (1) p(d)e = p(r)e extends to a formula for the effect of the operator p(d) on a product of the form e u, where

More information

Anti-synchronization of a new hyperchaotic system via small-gain theorem

Anti-synchronization of a new hyperchaotic system via small-gain theorem Anti-synchronization of a new hyperchaotic system via small-gain theorem Xiao Jian( ) College of Mathematics and Statistics, Chongqing University, Chongqing 400044, China (Received 8 February 2010; revised

More information

Prof. Krstic Nonlinear Systems MAE281A Homework set 1 Linearization & phase portrait

Prof. Krstic Nonlinear Systems MAE281A Homework set 1 Linearization & phase portrait Prof. Krstic Nonlinear Systems MAE28A Homework set Linearization & phase portrait. For each of the following systems, find all equilibrium points and determine the type of each isolated equilibrium. Use

More information

Multivariable Extremum Seeking Feedback: Analysis and Design 1

Multivariable Extremum Seeking Feedback: Analysis and Design 1 Multivariable Extremum Seeking Feedback: Analysis and Design Kartik B. Ariyur and Miroslav Krstic Department of Mechanical and Aerospace Engineering University of California, San Diego La Jolla, CA 92093-04

More information

Extremum Seeking with Drift

Extremum Seeking with Drift Extremum Seeking with Drift Jan aximilian ontenbruck, Hans-Bernd Dürr, Christian Ebenbauer, Frank Allgöwer Institute for Systems Theory and Automatic Control, University of Stuttgart Pfaffenwaldring 9,

More information

Multi-agent gradient climbing via extremum seeking control

Multi-agent gradient climbing via extremum seeking control Preprints of the 19th World Congress The International Federation of Automatic Control Multi-agent gradient climbing via extremum seeking control Sei Zhen Khong Chris Manzie Ying Tan Dragan Nešić Department

More information

A conjecture on sustained oscillations for a closed-loop heat equation

A conjecture on sustained oscillations for a closed-loop heat equation A conjecture on sustained oscillations for a closed-loop heat equation C.I. Byrnes, D.S. Gilliam Abstract The conjecture in this paper represents an initial step aimed toward understanding and shaping

More information

Indirect Model Reference Adaptive Control System Based on Dynamic Certainty Equivalence Principle and Recursive Identifier Scheme

Indirect Model Reference Adaptive Control System Based on Dynamic Certainty Equivalence Principle and Recursive Identifier Scheme Indirect Model Reference Adaptive Control System Based on Dynamic Certainty Equivalence Principle and Recursive Identifier Scheme Itamiya, K. *1, Sawada, M. 2 1 Dept. of Electrical and Electronic Eng.,

More information

The Rationale for Second Level Adaptation

The Rationale for Second Level Adaptation The Rationale for Second Level Adaptation Kumpati S. Narendra, Yu Wang and Wei Chen Center for Systems Science, Yale University arxiv:1510.04989v1 [cs.sy] 16 Oct 2015 Abstract Recently, a new approach

More information

Limit Cycles in High-Resolution Quantized Feedback Systems

Limit Cycles in High-Resolution Quantized Feedback Systems Limit Cycles in High-Resolution Quantized Feedback Systems Li Hong Idris Lim School of Engineering University of Glasgow Glasgow, United Kingdom LiHonIdris.Lim@glasgow.ac.uk Ai Poh Loh Department of Electrical

More information

Stability Analysis of the Simplest Takagi-Sugeno Fuzzy Control System Using Popov Criterion

Stability Analysis of the Simplest Takagi-Sugeno Fuzzy Control System Using Popov Criterion Stability Analysis of the Simplest Takagi-Sugeno Fuzzy Control System Using Popov Criterion Xiaojun Ban, X. Z. Gao, Xianlin Huang 3, and Hang Yin 4 Department of Control Theory and Engineering, Harbin

More information

DESIGN OF AN ON-LINE TITRATOR FOR NONLINEAR ph CONTROL

DESIGN OF AN ON-LINE TITRATOR FOR NONLINEAR ph CONTROL DESIGN OF AN ON-LINE TITRATOR FOR NONLINEAR CONTROL Alex D. Kalafatis Liuping Wang William R. Cluett AspenTech, Toronto, Canada School of Electrical & Computer Engineering, RMIT University, Melbourne,

More information

Stability of Hybrid Control Systems Based on Time-State Control Forms

Stability of Hybrid Control Systems Based on Time-State Control Forms Stability of Hybrid Control Systems Based on Time-State Control Forms Yoshikatsu HOSHI, Mitsuji SAMPEI, Shigeki NAKAURA Department of Mechanical and Control Engineering Tokyo Institute of Technology 2

More information

Extremum Seeking-based Indirect Adaptive Control for Nonlinear Systems

Extremum Seeking-based Indirect Adaptive Control for Nonlinear Systems MITSUBISHI ELECTRIC RESEARCH LABORATORIES http://www.merl.com Extremum Seeking-based Indirect Adaptive Control for Nonlinear Systems Benosman, M. TR14-85 August 14 Abstract We present in this paper a preliminary

More information

Event-Triggered Decentralized Dynamic Output Feedback Control for LTI Systems

Event-Triggered Decentralized Dynamic Output Feedback Control for LTI Systems Event-Triggered Decentralized Dynamic Output Feedback Control for LTI Systems Pavankumar Tallapragada Nikhil Chopra Department of Mechanical Engineering, University of Maryland, College Park, 2742 MD,

More information

Steady-state performance optimization for variable-gain motion control using extremum seeking*

Steady-state performance optimization for variable-gain motion control using extremum seeking* 51st IEEE Conference on Decision and Control December 1-13, 1. Maui, Hawaii, USA Steady-state performance optimization for variable-gain motion control using extremum seeking* B.G.B. Hunnekens 1, M.A.M.

More information

Self-Tuning Control for Synchronous Machine Stabilization

Self-Tuning Control for Synchronous Machine Stabilization http://dx.doi.org/.5755/j.eee.2.4.2773 ELEKTRONIKA IR ELEKTROTECHNIKA, ISSN 392-25, VOL. 2, NO. 4, 25 Self-Tuning Control for Synchronous Machine Stabilization Jozef Ritonja Faculty of Electrical Engineering

More information

State and Parameter Estimation Based on Filtered Transformation for a Class of Second-Order Systems

State and Parameter Estimation Based on Filtered Transformation for a Class of Second-Order Systems State and Parameter Estimation Based on Filtered Transformation for a Class of Second-Order Systems Mehdi Tavan, Kamel Sabahi, and Saeid Hoseinzadeh Abstract This paper addresses the problem of state and

More information

Prediction-based adaptive control of a class of discrete-time nonlinear systems with nonlinear growth rate

Prediction-based adaptive control of a class of discrete-time nonlinear systems with nonlinear growth rate www.scichina.com info.scichina.com www.springerlin.com Prediction-based adaptive control of a class of discrete-time nonlinear systems with nonlinear growth rate WEI Chen & CHEN ZongJi School of Automation

More information

EN Nonlinear Control and Planning in Robotics Lecture 3: Stability February 4, 2015

EN Nonlinear Control and Planning in Robotics Lecture 3: Stability February 4, 2015 EN530.678 Nonlinear Control and Planning in Robotics Lecture 3: Stability February 4, 2015 Prof: Marin Kobilarov 0.1 Model prerequisites Consider ẋ = f(t, x). We will make the following basic assumptions

More information

Systems Analysis and Control

Systems Analysis and Control Systems Analysis and Control Matthew M. Peet Illinois Institute of Technology Lecture 8: Response Characteristics Overview In this Lecture, you will learn: Characteristics of the Response Stability Real

More information

Lyapunov Stability of Linear Predictor Feedback for Distributed Input Delays

Lyapunov Stability of Linear Predictor Feedback for Distributed Input Delays IEEE TRANSACTIONS ON AUTOMATIC CONTROL VOL. 56 NO. 3 MARCH 2011 655 Lyapunov Stability of Linear Predictor Feedback for Distributed Input Delays Nikolaos Bekiaris-Liberis Miroslav Krstic In this case system

More information

On Extremum Seeking in Bioprocesses with Multivalued Cost Functions

On Extremum Seeking in Bioprocesses with Multivalued Cost Functions On Extremum Seeking in Bioprocesses with Multivalued Cost Functions Georges Bastin, Dragan Nešić, Ying Tan and Iven Mareels July 18, 2008 Abstract Finding optimal operating modes for bioprocesses has been,

More information

HIGHER ORDER SLIDING MODES AND ARBITRARY-ORDER EXACT ROBUST DIFFERENTIATION

HIGHER ORDER SLIDING MODES AND ARBITRARY-ORDER EXACT ROBUST DIFFERENTIATION HIGHER ORDER SLIDING MODES AND ARBITRARY-ORDER EXACT ROBUST DIFFERENTIATION A. Levant Institute for Industrial Mathematics, 4/24 Yehuda Ha-Nachtom St., Beer-Sheva 843, Israel Fax: +972-7-232 and E-mail:

More information

A Unifying Approach to Extremum Seeking: Adaptive Schemes Based on Estimation of Derivatives

A Unifying Approach to Extremum Seeking: Adaptive Schemes Based on Estimation of Derivatives 49th IEEE Conference on Decision and Control December 15-17, 2010 Hilton Atlanta Hotel, Atlanta, GA, USA A Unifying Approach to Etremum Seeking: Adaptive Schemes Based on Estimation of Derivatives D. Nešić,

More information

Converse Lyapunov theorem and Input-to-State Stability

Converse Lyapunov theorem and Input-to-State Stability Converse Lyapunov theorem and Input-to-State Stability April 6, 2014 1 Converse Lyapunov theorem In the previous lecture, we have discussed few examples of nonlinear control systems and stability concepts

More information

Passivity-based Stabilization of Non-Compact Sets

Passivity-based Stabilization of Non-Compact Sets Passivity-based Stabilization of Non-Compact Sets Mohamed I. El-Hawwary and Manfredi Maggiore Abstract We investigate the stabilization of closed sets for passive nonlinear systems which are contained

More information

1 Lyapunov theory of stability

1 Lyapunov theory of stability M.Kawski, APM 581 Diff Equns Intro to Lyapunov theory. November 15, 29 1 1 Lyapunov theory of stability Introduction. Lyapunov s second (or direct) method provides tools for studying (asymptotic) stability

More information

Robust Stabilization of Jet Engine Compressor in the Presence of Noise and Unmeasured States

Robust Stabilization of Jet Engine Compressor in the Presence of Noise and Unmeasured States obust Stabilization of Jet Engine Compressor in the Presence of Noise and Unmeasured States John A Akpobi, Member, IAENG and Aloagbaye I Momodu Abstract Compressors for jet engines in operation experience

More information

Analysis of different Lyapunov function constructions for interconnected hybrid systems

Analysis of different Lyapunov function constructions for interconnected hybrid systems Analysis of different Lyapunov function constructions for interconnected hybrid systems Guosong Yang 1 Daniel Liberzon 1 Andrii Mironchenko 2 1 Coordinated Science Laboratory University of Illinois at

More information

STABILITY ANALYSIS OF DAMPED SDOF SYSTEMS WITH TWO TIME DELAYS IN STATE FEEDBACK

STABILITY ANALYSIS OF DAMPED SDOF SYSTEMS WITH TWO TIME DELAYS IN STATE FEEDBACK Journal of Sound and Vibration (1998) 214(2), 213 225 Article No. sv971499 STABILITY ANALYSIS OF DAMPED SDOF SYSTEMS WITH TWO TIME DELAYS IN STATE FEEDBACK H. Y. HU ANDZ. H. WANG Institute of Vibration

More information

NONLINEAR SAMPLED-DATA OBSERVER DESIGN VIA APPROXIMATE DISCRETE-TIME MODELS AND EMULATION

NONLINEAR SAMPLED-DATA OBSERVER DESIGN VIA APPROXIMATE DISCRETE-TIME MODELS AND EMULATION NONLINEAR SAMPLED-DAA OBSERVER DESIGN VIA APPROXIMAE DISCREE-IME MODELS AND EMULAION Murat Arcak Dragan Nešić Department of Electrical, Computer, and Systems Engineering Rensselaer Polytechnic Institute

More information

Systems Analysis and Control

Systems Analysis and Control Systems Analysis and Control Matthew M. Peet Arizona State University Lecture 8: Response Characteristics Overview In this Lecture, you will learn: Characteristics of the Response Stability Real Poles

More information

The Harmonic Oscillator

The Harmonic Oscillator The Harmonic Oscillator Math 4: Ordinary Differential Equations Chris Meyer May 3, 008 Introduction The harmonic oscillator is a common model used in physics because of the wide range of problems it can

More information

Extremum-seeking in Singularly Perturbed Hybrid Systems

Extremum-seeking in Singularly Perturbed Hybrid Systems This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI.9/TAC.26.267282,

More information

ADAPTIVE FEEDBACK LINEARIZING CONTROL OF CHUA S CIRCUIT

ADAPTIVE FEEDBACK LINEARIZING CONTROL OF CHUA S CIRCUIT International Journal of Bifurcation and Chaos, Vol. 12, No. 7 (2002) 1599 1604 c World Scientific Publishing Company ADAPTIVE FEEDBACK LINEARIZING CONTROL OF CHUA S CIRCUIT KEVIN BARONE and SAHJENDRA

More information

Time Response of Systems

Time Response of Systems Chapter 0 Time Response of Systems 0. Some Standard Time Responses Let us try to get some impulse time responses just by inspection: Poles F (s) f(t) s-plane Time response p =0 s p =0,p 2 =0 s 2 t p =

More information

Linear and Nonlinear Oscillators (Lecture 2)

Linear and Nonlinear Oscillators (Lecture 2) Linear and Nonlinear Oscillators (Lecture 2) January 25, 2016 7/441 Lecture outline A simple model of a linear oscillator lies in the foundation of many physical phenomena in accelerator dynamics. A typical

More information

Stochastic Averaging in Continuous Time and Its Applications to Extremum Seeking

Stochastic Averaging in Continuous Time and Its Applications to Extremum Seeking IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL 55, NO 10, OCTOBER 2010 2235 Stochastic Averaging in Continuous Time and Its Applications to Extremum Seeking Shu-Jun Liu, Member, IEEE, and Miroslav Krstic,

More information

Course Summary. The course cannot be summarized in one lecture.

Course Summary. The course cannot be summarized in one lecture. Course Summary Unit 1: Introduction Unit 2: Modeling in the Frequency Domain Unit 3: Time Response Unit 4: Block Diagram Reduction Unit 5: Stability Unit 6: Steady-State Error Unit 7: Root Locus Techniques

More information

The Tuning of Robust Controllers for Stable Systems in the CD-Algebra: The Case of Sinusoidal and Polynomial Signals

The Tuning of Robust Controllers for Stable Systems in the CD-Algebra: The Case of Sinusoidal and Polynomial Signals The Tuning of Robust Controllers for Stable Systems in the CD-Algebra: The Case of Sinusoidal and Polynomial Signals Timo Hämäläinen Seppo Pohjolainen Tampere University of Technology Department of Mathematics

More information

Astanding assumption in extremum seeking is that the

Astanding assumption in extremum seeking is that the IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL 58, NO 2, FEBRUARY 2013 435 A Framework for Extremum Seeking Control of Systems With Parameter Uncertainties Dragan Nešić, Fellow, IEEE, Alireza Mohammadi, and

More information

Stability of Feedback Control Systems: Absolute and Relative

Stability of Feedback Control Systems: Absolute and Relative Stability of Feedback Control Systems: Absolute and Relative Dr. Kevin Craig Greenheck Chair in Engineering Design & Professor of Mechanical Engineering Marquette University Stability: Absolute and Relative

More information

L -Bounded Robust Control of Nonlinear Cascade Systems

L -Bounded Robust Control of Nonlinear Cascade Systems L -Bounded Robust Control of Nonlinear Cascade Systems Shoudong Huang M.R. James Z.P. Jiang August 19, 2004 Accepted by Systems & Control Letters Abstract In this paper, we consider the L -bounded robust

More information

RELAY CONTROL WITH PARALLEL COMPENSATOR FOR NONMINIMUM PHASE PLANTS. Ryszard Gessing

RELAY CONTROL WITH PARALLEL COMPENSATOR FOR NONMINIMUM PHASE PLANTS. Ryszard Gessing RELAY CONTROL WITH PARALLEL COMPENSATOR FOR NONMINIMUM PHASE PLANTS Ryszard Gessing Politechnika Śl aska Instytut Automatyki, ul. Akademicka 16, 44-101 Gliwice, Poland, fax: +4832 372127, email: gessing@ia.gliwice.edu.pl

More information

1 Simple Harmonic Oscillator

1 Simple Harmonic Oscillator Physics 1a Waves Lecture 3 Caltech, 10/09/18 1 Simple Harmonic Oscillator 1.4 General properties of Simple Harmonic Oscillator 1.4.4 Superposition of two independent SHO Suppose we have two SHOs described

More information

Adaptive and Robust Controls of Uncertain Systems With Nonlinear Parameterization

Adaptive and Robust Controls of Uncertain Systems With Nonlinear Parameterization IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 48, NO. 0, OCTOBER 003 87 Adaptive and Robust Controls of Uncertain Systems With Nonlinear Parameterization Zhihua Qu Abstract Two classes of partially known

More information

CONTROLLING IN BETWEEN THE LORENZ AND THE CHEN SYSTEMS

CONTROLLING IN BETWEEN THE LORENZ AND THE CHEN SYSTEMS International Journal of Bifurcation and Chaos, Vol. 12, No. 6 (22) 1417 1422 c World Scientific Publishing Company CONTROLLING IN BETWEEN THE LORENZ AND THE CHEN SYSTEMS JINHU LÜ Institute of Systems

More information

Supporting Information. Methods. Equations for four regimes

Supporting Information. Methods. Equations for four regimes Supporting Information A Methods All analytical expressions were obtained starting from quation 3, the tqssa approximation of the cycle, the derivation of which is discussed in Appendix C. The full mass

More information

Integrator Backstepping using Barrier Functions for Systems with Multiple State Constraints

Integrator Backstepping using Barrier Functions for Systems with Multiple State Constraints Integrator Backstepping using Barrier Functions for Systems with Multiple State Constraints Khoi Ngo Dep. Engineering, Australian National University, Australia Robert Mahony Dep. Engineering, Australian

More information

DISCRETE-TIME TIME-VARYING ROBUST STABILIZATION FOR SYSTEMS IN POWER FORM. Dina Shona Laila and Alessandro Astolfi

DISCRETE-TIME TIME-VARYING ROBUST STABILIZATION FOR SYSTEMS IN POWER FORM. Dina Shona Laila and Alessandro Astolfi DISCRETE-TIME TIME-VARYING ROBUST STABILIZATION FOR SYSTEMS IN POWER FORM Dina Shona Laila and Alessandro Astolfi Electrical and Electronic Engineering Department Imperial College, Exhibition Road, London

More information

D(s) G(s) A control system design definition

D(s) G(s) A control system design definition R E Compensation D(s) U Plant G(s) Y Figure 7. A control system design definition x x x 2 x 2 U 2 s s 7 2 Y Figure 7.2 A block diagram representing Eq. (7.) in control form z U 2 s z Y 4 z 2 s z 2 3 Figure

More information

Adaptive State Feedback Nash Strategies for Linear Quadratic Discrete-Time Games

Adaptive State Feedback Nash Strategies for Linear Quadratic Discrete-Time Games Adaptive State Feedbac Nash Strategies for Linear Quadratic Discrete-Time Games Dan Shen and Jose B. Cruz, Jr. Intelligent Automation Inc., Rocville, MD 2858 USA (email: dshen@i-a-i.com). The Ohio State

More information

Advanced Aerospace Control. Marco Lovera Dipartimento di Scienze e Tecnologie Aerospaziali, Politecnico di Milano

Advanced Aerospace Control. Marco Lovera Dipartimento di Scienze e Tecnologie Aerospaziali, Politecnico di Milano Advanced Aerospace Control Dipartimento di Scienze e Tecnologie Aerospaziali, Politecnico di Milano ICT for control systems engineering School of Industrial and Information Engineering Aeronautical Engineering

More information

Alexander Scheinker Miroslav Krstić. Model-Free Stabilization by Extremum Seeking

Alexander Scheinker Miroslav Krstić. Model-Free Stabilization by Extremum Seeking Alexander Scheinker Miroslav Krstić Model-Free Stabilization by Extremum Seeking 123 Preface Originating in 1922, in its 95-year history, extremum seeking has served as a tool for model-free real-time

More information

Target Localization and Circumnavigation Using Bearing Measurements in 2D

Target Localization and Circumnavigation Using Bearing Measurements in 2D Target Localization and Circumnavigation Using Bearing Measurements in D Mohammad Deghat, Iman Shames, Brian D. O. Anderson and Changbin Yu Abstract This paper considers the problem of localization and

More information

OUTPUT REGULATION OF THE SIMPLIFIED LORENZ CHAOTIC SYSTEM

OUTPUT REGULATION OF THE SIMPLIFIED LORENZ CHAOTIC SYSTEM OUTPUT REGULATION OF THE SIMPLIFIED LORENZ CHAOTIC SYSTEM Sundarapandian Vaidyanathan Research and Development Centre, Vel Tech Dr. RR & Dr. SR Technical University Avadi, Chennai-600 06, Tamil Nadu, INDIA

More information

10 Transfer Matrix Models

10 Transfer Matrix Models MIT EECS 6.241 (FALL 26) LECTURE NOTES BY A. MEGRETSKI 1 Transfer Matrix Models So far, transfer matrices were introduced for finite order state space LTI models, in which case they serve as an important

More information

Global stabilization of feedforward systems with exponentially unstable Jacobian linearization

Global stabilization of feedforward systems with exponentially unstable Jacobian linearization Global stabilization of feedforward systems with exponentially unstable Jacobian linearization F Grognard, R Sepulchre, G Bastin Center for Systems Engineering and Applied Mechanics Université catholique

More information

Disturbance Attenuation for a Class of Nonlinear Systems by Output Feedback

Disturbance Attenuation for a Class of Nonlinear Systems by Output Feedback Disturbance Attenuation for a Class of Nonlinear Systems by Output Feedback Wei in Chunjiang Qian and Xianqing Huang Submitted to Systems & Control etters /5/ Abstract This paper studies the problem of

More information

Damped Harmonic Oscillator

Damped Harmonic Oscillator Damped Harmonic Oscillator Wednesday, 23 October 213 A simple harmonic oscillator subject to linear damping may oscillate with exponential decay, or it may decay biexponentially without oscillating, or

More information

Professional Portfolio Selection Techniques: From Markowitz to Innovative Engineering

Professional Portfolio Selection Techniques: From Markowitz to Innovative Engineering Massachusetts Institute of Technology Sponsor: Electrical Engineering and Computer Science Cosponsor: Science Engineering and Business Club Professional Portfolio Selection Techniques: From Markowitz to

More information

Nonlinear L 2 -gain analysis via a cascade

Nonlinear L 2 -gain analysis via a cascade 9th IEEE Conference on Decision and Control December -7, Hilton Atlanta Hotel, Atlanta, GA, USA Nonlinear L -gain analysis via a cascade Peter M Dower, Huan Zhang and Christopher M Kellett Abstract A nonlinear

More information

An introduction to Birkhoff normal form

An introduction to Birkhoff normal form An introduction to Birkhoff normal form Dario Bambusi Dipartimento di Matematica, Universitá di Milano via Saldini 50, 0133 Milano (Italy) 19.11.14 1 Introduction The aim of this note is to present an

More information

Nonlinear System Analysis

Nonlinear System Analysis Nonlinear System Analysis Lyapunov Based Approach Lecture 4 Module 1 Dr. Laxmidhar Behera Department of Electrical Engineering, Indian Institute of Technology, Kanpur. January 4, 2003 Intelligent Control

More information

Distributed Adaptive Consensus Protocol with Decaying Gains on Directed Graphs

Distributed Adaptive Consensus Protocol with Decaying Gains on Directed Graphs Distributed Adaptive Consensus Protocol with Decaying Gains on Directed Graphs Štefan Knotek, Kristian Hengster-Movric and Michael Šebek Department of Control Engineering, Czech Technical University, Prague,

More information

Linear State Feedback Controller Design

Linear State Feedback Controller Design Assignment For EE5101 - Linear Systems Sem I AY2010/2011 Linear State Feedback Controller Design Phang Swee King A0033585A Email: king@nus.edu.sg NGS/ECE Dept. Faculty of Engineering National University

More information

AFAULT diagnosis procedure is typically divided into three

AFAULT diagnosis procedure is typically divided into three 576 IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 47, NO. 4, APRIL 2002 A Robust Detection and Isolation Scheme for Abrupt and Incipient Faults in Nonlinear Systems Xiaodong Zhang, Marios M. Polycarpou,

More information

Distributed Control of Multi-agent systems using Extremum Seeking

Distributed Control of Multi-agent systems using Extremum Seeking Distributed Control of Multi-agent systems using Extremum Seeking by Judith Ebegbulem A thesis submitted to the Department of Chemical Engineering in conformity with the requirements for the degree of

More information

CONTROL SYSTEMS, ROBOTICS AND AUTOMATION Vol. XI Stochastic Stability - H.J. Kushner

CONTROL SYSTEMS, ROBOTICS AND AUTOMATION Vol. XI Stochastic Stability - H.J. Kushner STOCHASTIC STABILITY H.J. Kushner Applied Mathematics, Brown University, Providence, RI, USA. Keywords: stability, stochastic stability, random perturbations, Markov systems, robustness, perturbed systems,

More information

OUTPUT REGULATION OF RÖSSLER PROTOTYPE-4 CHAOTIC SYSTEM BY STATE FEEDBACK CONTROL

OUTPUT REGULATION OF RÖSSLER PROTOTYPE-4 CHAOTIC SYSTEM BY STATE FEEDBACK CONTROL International Journal in Foundations of Computer Science & Technology (IJFCST),Vol., No., March 01 OUTPUT REGULATION OF RÖSSLER PROTOTYPE-4 CHAOTIC SYSTEM BY STATE FEEDBACK CONTROL Sundarapandian Vaidyanathan

More information

Problem 1: Lagrangians and Conserved Quantities. Consider the following action for a particle of mass m moving in one dimension

Problem 1: Lagrangians and Conserved Quantities. Consider the following action for a particle of mass m moving in one dimension 105A Practice Final Solutions March 13, 01 William Kelly Problem 1: Lagrangians and Conserved Quantities Consider the following action for a particle of mass m moving in one dimension S = dtl = mc dt 1

More information

Closed loop Identification of Four Tank Set up Using Direct Method

Closed loop Identification of Four Tank Set up Using Direct Method Closed loop Identification of Four Tan Set up Using Direct Method Mrs. Mugdha M. Salvi*, Dr.(Mrs) J. M. Nair** *(Department of Instrumentation Engg., Vidyavardhini s College of Engg. Tech., Vasai, Maharashtra,

More information

Laplace Transform Part 1: Introduction (I&N Chap 13)

Laplace Transform Part 1: Introduction (I&N Chap 13) Laplace Transform Part 1: Introduction (I&N Chap 13) Definition of the L.T. L.T. of Singularity Functions L.T. Pairs Properties of the L.T. Inverse L.T. Convolution IVT(initial value theorem) & FVT (final

More information

EE C128 / ME C134 Final Exam Fall 2014

EE C128 / ME C134 Final Exam Fall 2014 EE C128 / ME C134 Final Exam Fall 2014 December 19, 2014 Your PRINTED FULL NAME Your STUDENT ID NUMBER Number of additional sheets 1. No computers, no tablets, no connected device (phone etc.) 2. Pocket

More information

A LaSalle version of Matrosov theorem

A LaSalle version of Matrosov theorem 5th IEEE Conference on Decision Control European Control Conference (CDC-ECC) Orlo, FL, USA, December -5, A LaSalle version of Matrosov theorem Alessro Astolfi Laurent Praly Abstract A weak version of

More information

10. Operators and the Exponential Response Formula

10. Operators and the Exponential Response Formula 52 10. Operators and the Exponential Response Formula 10.1. Operators. Operators are to functions as functions are to numbers. An operator takes a function, does something to it, and returns this modified

More information

Chapter 2 Optimal Control Problem

Chapter 2 Optimal Control Problem Chapter 2 Optimal Control Problem Optimal control of any process can be achieved either in open or closed loop. In the following two chapters we concentrate mainly on the first class. The first chapter

More information

Multi-Robotic Systems

Multi-Robotic Systems CHAPTER 9 Multi-Robotic Systems The topic of multi-robotic systems is quite popular now. It is believed that such systems can have the following benefits: Improved performance ( winning by numbers ) Distributed

More information

Additive resonances of a controlled van der Pol-Duffing oscillator

Additive resonances of a controlled van der Pol-Duffing oscillator Additive resonances of a controlled van der Pol-Duffing oscillator This paper has been published in Journal of Sound and Vibration vol. 5 issue - 8 pp.-. J.C. Ji N. Zhang Faculty of Engineering University

More information

Basic Concepts in Data Reconciliation. Chapter 6: Steady-State Data Reconciliation with Model Uncertainties

Basic Concepts in Data Reconciliation. Chapter 6: Steady-State Data Reconciliation with Model Uncertainties Chapter 6: Steady-State Data with Model Uncertainties CHAPTER 6 Steady-State Data with Model Uncertainties 6.1 Models with Uncertainties In the previous chapters, the models employed in the DR were considered

More information

Optimization of Phase-Locked Loops With Guaranteed Stability. C. M. Kwan, H. Xu, C. Lin, and L. Haynes

Optimization of Phase-Locked Loops With Guaranteed Stability. C. M. Kwan, H. Xu, C. Lin, and L. Haynes Optimization of Phase-Locked Loops With Guaranteed Stability C. M. Kwan, H. Xu, C. Lin, and L. Haynes ntelligent Automation nc. 2 Research Place Suite 202 Rockville, MD 20850 phone: (301) - 590-3155, fax:

More information

EECE Adaptive Control

EECE Adaptive Control EECE 574 - Adaptive Control Overview Guy Dumont Department of Electrical and Computer Engineering University of British Columbia Lectures: Thursday 09h00-12h00 Location: PPC 101 Guy Dumont (UBC) EECE 574

More information

Hybrid Systems Course Lyapunov stability

Hybrid Systems Course Lyapunov stability Hybrid Systems Course Lyapunov stability OUTLINE Focus: stability of an equilibrium point continuous systems decribed by ordinary differential equations (brief review) hybrid automata OUTLINE Focus: stability

More information

Iterative Controller Tuning Using Bode s Integrals

Iterative Controller Tuning Using Bode s Integrals Iterative Controller Tuning Using Bode s Integrals A. Karimi, D. Garcia and R. Longchamp Laboratoire d automatique, École Polytechnique Fédérale de Lausanne (EPFL), 05 Lausanne, Switzerland. email: alireza.karimi@epfl.ch

More information

Review of Linear Time-Invariant Network Analysis

Review of Linear Time-Invariant Network Analysis D1 APPENDIX D Review of Linear Time-Invariant Network Analysis Consider a network with input x(t) and output y(t) as shown in Figure D-1. If an input x 1 (t) produces an output y 1 (t), and an input x

More information

QUANTITATIVE L P STABILITY ANALYSIS OF A CLASS OF LINEAR TIME-VARYING FEEDBACK SYSTEMS

QUANTITATIVE L P STABILITY ANALYSIS OF A CLASS OF LINEAR TIME-VARYING FEEDBACK SYSTEMS Int. J. Appl. Math. Comput. Sci., 2003, Vol. 13, No. 2, 179 184 QUANTITATIVE L P STABILITY ANALYSIS OF A CLASS OF LINEAR TIME-VARYING FEEDBACK SYSTEMS PINI GURFIL Department of Mechanical and Aerospace

More information