Adaptive input-to-state stable synchronization for uncertain time-delay Lur'e systems
本站小编 Free考研考试/2022-01-02
Zhi Zhang(章枝)1, Zhilian Yan(闫志莲),2,*, Jianping Zhou(周建平)1,3,4, Yebin Chen(陈业斌)11School of Computer Science & Technology, Anhui University of Technology, Ma'anshan 243032, China 2College of Electrical Engineering & Automation, Shandong University of Science and Technology, Qingdao 266590, China 3Research Institute of Information Technology, Anhui University of Technology, Ma'anshan 243000, China 4Anhui Province Key Laboratory of Special Heavy Load Robot, Anhui University of Technology, Ma'anshan 243032, China
First author contact:*Author to whom any correspondence should be addressed. Received:2021-02-22Revised:2021-04-7Accepted:2021-05-17Online:2021-06-25
Abstract This paper is dedicated to the study of adaptive input-to-state stable synchronization of uncertain time-delay Lur'e systems with exterior interference. With the help of the Lyapunov function approach, a sufficient condition for the input-to-state stability of the synchronization-error system is derived, which is theoretically less conservative than a previously reported criterion in the absence of parameter uncertainties. On the basis of the present condition, a co-design of the feedback gain and estimates of the uncertain parameters is given to determine the desired adaptive synchronization controller. Finally, an example with simulations is provided to demonstrate the applicability and superiority of the analysis and design strategies. Keywords：input-to-state stability;Lur'e systems;parameter uncertainty;adaptive control
PDF (1032KB)MetadataMetricsRelated articlesExportEndNote|Ris|BibtexFavorite Cite this article Zhi Zhang(章枝), Zhilian Yan(闫志莲), Jianping Zhou(周建平), Yebin Chen(陈业斌). Adaptive input-to-state stable synchronization for uncertain time-delay Lur'e systems. Communications in Theoretical Physics, 2021, 73(8): 085003- doi:10.1088/1572-9494/ac01dc
1. Introduction
Lur'e systems are a class of typical nonlinear dynamic systems that include a number of different chaotic models, such as Chua's circuits, Hopfield neural networks, Lorenz systems, and genetic oscillators, as their special cases. Over the past few decades, synchronization of Lur'e systems without and with time-delay has attracted increasing attention from researchers in various disciplines owing to their applications in image encryption as well as secure communication [1, 2]. Corresponding, plenty of research results regarding the synchronization of Lur'e systems have been proposed in the literature; see, e.g. [3-7] and the references therein.
An actual dynamic system inevitably suffers from exterior interference of varying degrees, which may have a great influence on the synchronization behaviors. Considering that the existing interference is bounded in most cases, it is of necessity to make sure that the state of the synchronization-error system (SES) is also bounded. This issue, called the input-to-state stable (ISS) synchronization problem [8-11], has drawn the interest of researchers in the control community. Nevertheless, it seems that there are few results available on ISS synchronization of time-delay Lur'e systems despite their importance.
On the other hand, structure and parameter uncertainties often exist owing to modeling errors. It is often the case that one has very little information on the bound of these uncertainties, which had led in the past decades to a focus on the adaptive control scheme, with the belief that such a control scheme not only is easy to be implemented, but also can relatively smoothly and accurately track the system trajectory during normal operation [12]. For time-delay Lur'e systems, although the adaptive control approach has been considered in [13-15], how to use it to achieve ISS synchronization of uncertain time-delay Lur'e systems with exterior interference is still unclear and deserves further detailed investigation.
In this paper, we attempt to study the issue of adaptive ISS synchronization for uncertain time-delay Lur'e systems with exterior interference. The main contributions of our work are twofold:A sufficient condition on the ISS of the SES is derived with the help of the Lyapunov function approach. It is theoretically less conservative than the criterion presented in [16] when there are no parameter uncertainties in the Lur'e system; A co-design of the feedback gain and estimates of the uncertain parameters is given to determine the desired adaptive synchronization controller. It depends on the feasibility of few linear matrix inequalities (LMIs) and, therefore, is numerically tractable.
Notation: let us use ${{\mathbb{R}}}^{m}$, ${{\mathbb{R}}}^{+}$, ${{\mathbb{S}}}_{+}^{m}$, and ${{\mathbb{R}}}^{m\times l}$ to denote the m-dimensional Euclidean space equipped with the Euclidean norm $\parallel \cdot \parallel $, the space of nonnegative real numbers, the space of m-dimensional positive-definite real matrices, and the set of all m × l real matrices, respectively; use ${\mathscr{C}}\left({\rm{\Omega }},{{\mathbb{R}}}^{n}\right)$ to denote the family of all continuous ${{\mathbb{R}}}^{n}$-valued functions on set Ω; use λM(P), λm(P), and PT to stand for the largest eigenvalue, the smallest eigenvalue, and the transpose of a matrix P, respectively; and denote further by ${\mathfrak{H}}e(M)$ the sum of M and MT. Moreover, define $\gamma (\rho ):{{\mathbb{R}}}^{+}\to {{\mathbb{R}}}^{+}$ as a ${\mathscr{K}}$ function provided that the function is not only continuous and strictly nondecreasing, but also satisfies γ(0) = 0.
2. Preliminaries
Consider a type of Lur'e system modeled as$\begin{eqnarray}\left\{\begin{array}{l}{\dot{\zeta }}_{1}(t)={ \mathcal A }{\zeta }_{1}(t)+{{ \mathcal A }}_{\delta }{\zeta }_{1}(t-\delta )+{ \mathcal B }F\left({ \mathcal H }{\zeta }_{1}(t)\right)\\ +{{ \mathcal B }}_{\delta }G\left({ \mathcal H }{\zeta }_{1}(t-\delta )\right)\\ +\sum _{j=1}^{p}{{\rm{\Phi }}}_{j}({\zeta }_{1}(t)){\phi }_{j}+\sum _{j=1}^{q}{{\rm{\Psi }}}_{j}({\zeta }_{1}(t-\delta )){\psi }_{j}+{{ \mathcal E }}_{1}\omega (t)\\ {\eta }_{1}(t)={ \mathcal C }{\zeta }_{1}(t),\end{array}\right.\end{eqnarray}$where ${\zeta }_{1}(t)\in {{\mathbb{R}}}^{m}$ represents the state vector; ${\eta }_{1}(t)\in {{\mathbb{R}}}^{n}$ represents the output vector; $\omega (t)\in {{\mathbb{R}}}^{l}$ represents the exterior interference; $F(\cdot )={[\begin{array}{ccc}{F}_{1}(\cdot ) & \cdots & {F}_{m}(\cdot )\end{array}]}^{{\rm{T}}}$, $G(\cdot )={[\begin{array}{c}{G}_{1}(\cdot )\cdots {G}_{m}(\cdot )\end{array}]}^{{\rm{T}}}$, Øj(∙), ${{\rm{\Psi }}}_{j}(\cdot )\in {\mathscr{C}}({{\mathbb{R}}}^{n},{{\mathbb{R}}}^{n})$ are continuous nonlinear vector functions; ${ \mathcal A }$, ${{ \mathcal A }}_{\delta }$, ${ \mathcal B }$, ${{ \mathcal B }}_{\delta }$, ${ \mathcal C }$, ${{ \mathcal E }}_{1}$, and ${ \mathcal H }$ are the system parameter matrices; Øj, ${\psi }_{j}\in {\mathbb{R}}$ denote uncertain parameters; δ > 0 stands for a time-delay.
Lur'e system is a common type of nonlinear system formed by a linear time-invariant system and a feedback loop involving nonlinear dynamics. Such systems are capable of generating strange chaotic attractors and, thus, of great application potential in image encryption and secure communication. It is worthy to point out that the Lur'e system model considered herein includes many famous chaotic models, such as the Lorenz systems, Chua's circuits, and chaotic Hopfield neural networks as its special cases.
In the past decades, chaotic synchronization has aroused great interest in various fields. The existing schemes of synchronization can be roughly divided into two types: the coupled synchronization [17-19] and drive-response (or master-slave) synchronization [20-22]. In this paper, the drive-response scheme will be utilized. Let us set (1) as a drive (master) system and then set the following as the corresponding response (slave) system:$\begin{eqnarray}\left\{\begin{array}{l}{\dot{\zeta }}_{2}(t)={ \mathcal A }{\zeta }_{2}(t)+{{ \mathcal A }}_{\delta }{\zeta }_{2}(t-\delta )+{ \mathcal B }F\left({ \mathcal H }{\zeta }_{2}(t)\right)\\ +\ {{ \mathcal B }}_{\delta }G\left({ \mathcal H }{\zeta }_{2}(t-\delta )\right)\\ +u(t)+{{ \mathcal E }}_{2}\omega (t)\\ {\eta }_{2}(t)={ \mathcal C }{\zeta }_{2}(t),\end{array}\right.\end{eqnarray}$in which ${\zeta }_{2}(t)\in {{\mathbb{R}}}^{m}$, $u(t)\in {{\mathbb{R}}}^{m}$, and ${\eta }_{2}(t)\in {{\mathbb{R}}}^{n}$ stand for the state, input, and output vectors, respectively; ${{ \mathcal E }}_{2}\in {{\mathbb{R}}}^{n\times l}$ is a constant matrix.
Define ∈(t) = ζ2(t) − ζ1(t). Then we are able to establish the following SES:$\begin{eqnarray}\displaystyle \begin{array}{rcl}\dot{\epsilon }(t) & = & { \mathcal A }\epsilon (t)+{{ \mathcal A }}_{\delta }\epsilon (t-\delta )\\ & & +\ { \mathcal B }f\left({ \mathcal H }\epsilon (t)\right)+{{ \mathcal B }}_{\delta }g\left({ \mathcal H }\epsilon (t-\delta )\right)\\ & & +u(t)-\sum _{j=1}^{p}{{\rm{\Phi }}}_{j}({\zeta }_{1}(t)){\phi }_{j}\\ & & -\ \sum _{j=1}^{q}{{\rm{\Psi }}}_{j}({\zeta }_{1}(t-\delta )){\psi }_{j}+{ \mathcal E }\omega (t),\end{array}\end{eqnarray}$in which$\begin{eqnarray*}\begin{array}{rcl}f\left({ \mathcal H }\epsilon (t)\right) & = & F\left({ \mathcal H }\left(\epsilon (t)+{\zeta }_{1}(t)\right)\right)\\ & & -\ F({ \mathcal H }{\zeta }_{1}(t)),{ \mathcal E }={{ \mathcal E }}_{2}-{{ \mathcal E }}_{1},\\ g\left({ \mathcal H }\epsilon (t-\delta )\right) & = & G\left({ \mathcal H }\left(\epsilon (t-\delta )+{\zeta }_{1}(t-\delta )\right)\right)\\ & & -G({ \mathcal H }{\zeta }_{1}(t-\delta )).\end{array}\end{eqnarray*}$
In the paper, the adaptive control approach will be adopted. The form of the controller is as follows:$\begin{eqnarray}\displaystyle \begin{array}{rcl}u(t) & = & K\eta (t)+\sum _{j=1}^{p}{{\rm{\Phi }}}_{j}({\zeta }_{1}(t)){\breve{\phi }}_{j}(t)\\ & & +\sum _{j=1}^{q}{{\rm{\Psi }}}_{j}({\zeta }_{1}(t-\delta )){\breve{\psi }}_{j}(t),\end{array}\end{eqnarray}$in which η(t) = η2(t) − η1(t), K is the controller gain, and ${\breve{\phi }}_{j}(t)$ and ${\breve{\psi }}_{j}(t)$ are the estimates of Øj and ψj, respectively.
In order to achieve synchronization of Lur'e systems, a lot of control strategies including the PD control [6], the adaptive control [13-15], the sampled-state feedback control [23], the time-delayed feedback control [24], the output-feedback control [25], and the sliding mode control [26] etc, have been utilized. Among these control strategies, the adaptive control method not only is easy to be implemented, but also can relatively smoothly and accurately track the system trajectory during normal operation [12]. When the drive system under consideration is subject to structure and parameter uncertainties, such a control method serves as an ideal candidate.
Substituting (4) into (3), one has$\begin{eqnarray}\displaystyle \begin{array}{rcl}\dot{\epsilon }(t) & = & \left({ \mathcal A }+K{ \mathcal C }\right)\epsilon (t)+{{ \mathcal A }}_{\delta }\epsilon (t-\delta )\\ & & +\ { \mathcal B }f\left({ \mathcal H }\epsilon (t)\right)+{{ \mathcal B }}_{\delta }g\left({ \mathcal H }\epsilon (t-\delta )\right)\\ & & +\sum _{j=1}^{p}{{\rm{\Phi }}}_{j}({\zeta }_{1}(t))\left({\breve{\phi }}_{j}(t)-{\phi }_{j}\right)\\ & & +\sum _{j=1}^{q}{{\rm{\Psi }}}_{j}({\zeta }_{1}(t-\delta ))\left({\breve{\psi }}_{j}(t)-{\psi }_{j}\right)\\ & & +{ \mathcal E }\omega (t).\end{array}\end{eqnarray}$
In what follows we give the definitions of ISS and ISS synchronization:
Suppose that there are a ${\mathscr{K}}$ function γ(ρ) and a strictly decreasing function β(t) that converges to 0 as $t\to \infty $. Then, system (5) is called to be ISS if$\begin{eqnarray}\parallel \epsilon (t)\parallel \leqslant \beta \left(t\right)+\gamma \left(\mathop{\sup }\limits_{0\leqslant \rho \leqslant t}\parallel \omega (\rho )\parallel \right)\end{eqnarray}$holds for each t ≥ 0.
Systems (1) and (2) are said to be ISS synchronized if their SES (5) is ISS.
In a physical system, exterior interference is usually unavoidable, which is able to make the system oscillating, unstable or difficult to synchronize. Generally, the interference is not known a priori but is bounded in most cases. It can be observed from definition 1 that the ISS means that the system not only has bounded state under any bounded interference input but also is Lyapunov asymptotically stable in the absence of the interference. Thus the ISS may serve as a bridge among the input−output stability and the Lyapunov stability [27]. Especially, if $\beta \left(t\right)$ is chosen to be $\beta {{\rm{e}}}^{-\varepsilon t}$ [28], where β and ϵ are positive constants, then the ISS implies the exponential stability for the zero interference input case. Therefore, it is of great significance to study the ISS synchronization problem of Lur'e systems with exterior interference.
In this paper, the aim is to give a design of a adaptive controller in the form of (4) to make sure that the drive and response systems (1) and (2) to be ISS synchronized.
To accomplish the aim, let us introduce the following assumption, which is frequently applied in the literature [29-33]:Functions ${F}_{1}(\cdot ),\cdots ,{F}_{m}(\cdot )$, ${G}_{1}(\cdot ),\cdots ,$ and ${G}_{m}(\cdot )$ are global Lipschitz continuous; that is, there exist constants ${L}_{{fi}}\geqslant 0$ and ${L}_{{gj}}\geqslant 0$, $i=1,\cdots ,m$, $j=1,\cdots ,m$ such that$\begin{eqnarray*}\begin{array}{l}\left|{F}_{i}(\alpha )-{F}_{i}(\beta )\right|\leqslant {L}_{{fi}}\left|\alpha -\beta \right|,\\ \left|{G}_{j}(\alpha )-{G}_{j}(\beta )\right|\leqslant {L}_{{gj}}\left|\alpha -\beta \right|,\end{array}\end{eqnarray*}$for any α, $\beta \in {\mathbb{R}}$.
Besides, we prepare two necessary lemmas, where the proof of lemma 2 is not difficult and is omitted herein for brevity:[34] Given real matrices P and Q of suitable size, for any real constant θ > 0 we have$\begin{eqnarray*}{\mathfrak{H}}e\left({P}^{{\rm{T}}}Q\right)\leqslant {\theta }^{-1}{P}^{{\rm{T}}}P+\theta {Q}^{{\rm{T}}}Q.\end{eqnarray*}$
For any $\omega (\rho )\in {\mathscr{C}}\left([a,b],{{\mathbb{R}}}^{n}\right)$ (a, $b\in {\mathbb{R}})$, the following equality holds true:$\begin{eqnarray*}\mathop{\sup }\limits_{\rho \in [a,b]}{\parallel \omega (\rho )\parallel }^{2}={\left(\mathop{\sup }\limits_{\rho \in [a,b]}\parallel \omega (\rho )\parallel \right)}^{2}.\end{eqnarray*}$
3. Main results
This section is focused on the ISS analysis and the corresponding adaptive synchronization controller design. The following criterion regarding the ISS analysis of the SES in (5) can be established:Suppose that there are scalars θ1 > 0, θ2 > 0, and matrices P, Q, R, $S\in {{\mathbb{S}}}_{+}^{n}$ such that$\begin{eqnarray}\left[\begin{array}{ccccc}{{\rm{\Xi }}}_{1} & P{{ \mathcal A }}_{\delta } & P{ \mathcal E } & P{ \mathcal B } & P{{ \mathcal B }}_{\delta }\\ * & {\theta }_{2}{{ \mathcal H }}^{{\rm{T}}}{L}_{g}^{{\rm{T}}}{L}_{g}{ \mathcal H }-Q & 0 & 0 & 0\\ * & * & -R & 0 & 0\\ * & * & * & -{\theta }_{1}I & 0\\ * & * & * & * & -{\theta }_{2}I\end{array}\right]\lt 0,\end{eqnarray}$where$\begin{eqnarray*}\begin{array}{rcl}{{\rm{\Xi }}}_{1} & = & {\mathfrak{H}}e\left(P({ \mathcal A }+K{ \mathcal C })\right)+{\theta }_{1}{{ \mathcal H }}^{{\rm{T}}}{L}_{f}^{{\rm{T}}}{L}_{f}{ \mathcal H }+Q+S,\\ {L}_{f} & = & {\rm{diag}}\left\{{L}_{f1},\cdots ,{L}_{{\rm{fm}}}\right\},\\ {L}_{g} & = & {\rm{diag}}\left\{{L}_{g1},\cdots ,{L}_{{\rm{gm}}}\right\}.\end{array}\end{eqnarray*}$Then, system (5) is ISS if$\begin{eqnarray}{\mathop{\breve{\phi }}\limits^{\cdot }}_{j}(t)=-(1+2\varepsilon ){\breve{\phi }}_{j}(t)-\mu {{\rm{\Phi }}}_{j}^{{\rm{T}}}({\zeta }_{1}(t))P\epsilon (t)+(1+2\varepsilon ){\phi }_{j},\end{eqnarray}$$\begin{eqnarray}\begin{array}{rcl}{\mathop{\breve{\psi }}\limits^{\cdot }}_{j}(t) & = & -(1+2\varepsilon ){\breve{\psi }}_{j}(t)-\nu {{\rm{\Psi }}}_{j}^{{\rm{T}}}({\zeta }_{1}(t-\delta ))P\epsilon (t)\\ & & +\ (1+2\varepsilon ){\psi }_{j},\end{array}\end{eqnarray}$where $\varepsilon $ is a scalar satisfying$\begin{eqnarray}2\varepsilon {\lambda }_{M}(P)+({{\rm{e}}}^{2\varepsilon \delta }-1){\lambda }_{M}(Q)-{\lambda }_{m}(S)\leqslant 0,\end{eqnarray}$and $\mu $, ν are any positive scalars.
Proof. Let us select a Lyapunov function as$\begin{eqnarray}\begin{array}{rcl}V(t) & = & {\epsilon }^{{\rm{T}}}(t)P\epsilon (t)+{\int }_{t-\delta }^{t}{\epsilon }^{{\rm{T}}}(\rho )Q\epsilon (\rho ){\rm{d}}\rho \\ & & +\ {\mu }^{-1}\displaystyle \sum _{j=1}^{p}{\left({\breve{\phi }}_{j}(t)-{\phi }_{j}\right)}^{2}\\ & & +\ {\nu }^{-1}\displaystyle \sum _{j=1}^{q}{\left({\breve{\psi }}_{j}(t)-{\psi }_{j}\right)}^{2}.\end{array}\end{eqnarray}$We have along system (5) that$\begin{eqnarray}\displaystyle \begin{array}{l}\dot{V}(t)=2{\epsilon }^{{\rm{T}}}(t)P\dot{\epsilon }(t)+{\epsilon }^{{\rm{T}}}(t)Q\epsilon (t)\\ -\,{\epsilon }^{{\rm{T}}}(t-\delta )Q\epsilon (t-\delta )+2{\mu }^{-1}\sum _{j=1}^{p}\left({\breve{\phi }}_{j}(t)-{\phi }_{j}\right){\mathop{\breve{\phi }}\limits^{\cdot }}_{j}(t)\\ +\,2{\nu }^{-1}\sum _{j=1}^{q}\left({\breve{\psi }}_{j}(t)-{\psi }_{j}\right){\mathop{\breve{\psi }}\limits^{\cdot }}_{j}(t)\\ =\,2{\epsilon }^{{\rm{T}}}(t)P({ \mathcal A }+K{ \mathcal C })\epsilon (t)+2{\epsilon }^{{\rm{T}}}(t)P{{ \mathcal A }}_{\delta }\epsilon (t-\delta )\\ +\,2{\epsilon }^{{\rm{T}}}(t)P{ \mathcal B }f\left({ \mathcal H }\epsilon (t)\right)+2{\epsilon }^{{\rm{T}}}(t)P{{ \mathcal B }}_{\delta }g\left({ \mathcal H }\epsilon (t-\delta )\right)\\ +\,2{\epsilon }^{{\rm{T}}}(t)P{ \mathcal E }\omega (t)+2\sum _{j=1}^{p}\left({\breve{\phi }}_{j}(t)-{\phi }_{j}\right){{\rm{\Phi }}}_{j}^{{\rm{T}}}({\zeta }_{1}(t))P\epsilon (t)\\ +\,2\sum _{j=1}^{q}\left({\breve{\psi }}_{j}(t)-{\psi }_{j}\right){{\rm{\Psi }}}_{j}^{{\rm{T}}}({\zeta }_{1}(t-\delta ))P\epsilon (t)+{\epsilon }^{{\rm{T}}}(t)Q\epsilon (t)\\ -\,{\epsilon }^{{\rm{T}}}(t-\delta )Q\epsilon (t-\delta )+2{\mu }^{-1}\sum _{j=1}^{p}\left({\breve{\phi }}_{j}(t)-{\phi }_{j}\right){\mathop{\breve{\phi }}\limits^{\cdot }}_{j}(t)\\ +\,2{\nu }^{-1}\sum _{j=1}^{q}\left({\breve{\psi }}_{j}(t)-{\psi }_{j}\right){\mathop{\breve{\psi }}\limits^{\cdot }}_{j}(t).\end{array}\end{eqnarray}$With Assumption 1, the following inequalities hold true:$\begin{eqnarray*}\begin{array}{l}{f}^{{\rm{T}}}\left({ \mathcal H }\epsilon (t)\right)f\left({ \mathcal H }\epsilon (t)\right)\leqslant {\epsilon }^{{\rm{T}}}(t){{ \mathcal H }}^{{\rm{T}}}{L}_{f}^{{\rm{T}}}{L}_{f}{ \mathcal H }\epsilon (t),\\ {g}^{{\rm{T}}}\left({ \mathcal H }\epsilon (t-\delta )\right)g\left({ \mathcal H }\epsilon (t-\delta )\right)\\ \leqslant {\epsilon }^{{\rm{T}}}(t-\delta ){{ \mathcal H }}^{{\rm{T}}}{L}_{g}^{{\rm{T}}}{L}_{g}{ \mathcal H }\epsilon (t-\delta ),\end{array}\end{eqnarray*}$which together with lemma 1 result in$\begin{eqnarray}\begin{array}{l}{\epsilon }^{{\rm{T}}}(t)P{ \mathcal B }f\left({ \mathcal H }\epsilon (t)\right)+{f}^{{\rm{T}}}\left({ \mathcal H }\epsilon (t)\right){{ \mathcal B }}^{{\rm{T}}}P\epsilon (t)\\ \leqslant \,\frac{1}{{\theta }_{1}}{\epsilon }^{{\rm{T}}}(t)P{ \mathcal B }{{ \mathcal B }}^{{\rm{T}}}P\epsilon (t)+{\theta }_{1}{f}^{{\rm{T}}}\left({ \mathcal H }\epsilon (t)\right)f\left({ \mathcal H }\epsilon (t)\right)\\ \leqslant \,{\epsilon }^{{\rm{T}}}(t)\left(\frac{1}{{\theta }_{1}}P{ \mathcal B }{{ \mathcal B }}^{{\rm{T}}}P+{\theta }_{1}{{ \mathcal H }}^{{\rm{T}}}{L}_{f}^{{\rm{T}}}{L}_{f}{ \mathcal H }\right)\epsilon (t),\end{array}\end{eqnarray}$$\begin{eqnarray}\begin{array}{l}{\epsilon }^{{\rm{T}}}(t)P{{ \mathcal B }}_{\delta }g\left({ \mathcal H }\epsilon (t-\delta )\right)+{g}^{{\rm{T}}}\left({ \mathcal H }\epsilon (t-\delta )\right){{ \mathcal B }}_{\delta }^{{\rm{T}}}P\epsilon (t)\\ \leqslant \,\frac{1}{{\theta }_{2}}{\epsilon }^{{\rm{T}}}(t)P{{ \mathcal B }}_{\delta }{{ \mathcal B }}_{\delta }^{{\rm{T}}}P\epsilon (t)+{\theta }_{2}{g}^{{\rm{T}}}\left({ \mathcal H }\epsilon (t-\delta )\right)g\left({ \mathcal H }\epsilon (t-\delta )\right)\\ \leqslant \,\frac{1}{{\theta }_{2}}{\epsilon }^{{\rm{T}}}(t)P{{ \mathcal B }}_{\delta }{{ \mathcal B }}_{\delta }^{{\rm{T}}}P\epsilon (t)+{\theta }_{2}{\epsilon }^{{\rm{T}}}(t-\delta ){{ \mathcal H }}^{{\rm{T}}}{L}_{g}^{{\rm{T}}}{L}_{g}{ \mathcal H }\epsilon (t-\delta ).\end{array}\end{eqnarray}$Noting (13) and (14), it follows from (12) that$\begin{eqnarray}\begin{array}{l}\dot{V}(t)\leqslant {\left[\begin{array}{c}\epsilon (t)\\ \epsilon (t-\delta )\\ \omega (t)\end{array}\right]}^{{\rm{T}}}\left[\begin{array}{ccc}{{\rm{\Xi }}}_{2} & P{{ \mathcal A }}_{\delta } & P{ \mathcal E }\\ * & {\theta }_{2}{{ \mathcal H }}^{{\rm{T}}}{L}_{g}^{{\rm{T}}}{L}_{g}{ \mathcal H }-Q & 0\\ * & * & -R\end{array}\right]\\ \left[\begin{array}{c}\epsilon (t)\\ \epsilon (t-\delta )\\ \omega (t)\end{array}\right]-{\epsilon }^{{\rm{T}}}(t)S\epsilon (t)+{\omega }^{{\rm{T}}}(t)R\omega (t)\\ +\,2\displaystyle \sum _{j=1}^{p}\left({\breve{\phi }}_{j}(t)-{\phi }_{j}\right){{\rm{\Phi }}}_{j}^{{\rm{T}}}({\zeta }_{1}(t))P\epsilon (t)\\ +\,2\displaystyle \sum _{j=1}^{q}\left({\breve{\psi }}_{j}(t)-{\psi }_{j}\right){{\rm{\Psi }}}_{j}^{{\rm{T}}}({\zeta }_{1}(t-\delta ))P\epsilon (t)\\ +\,2{\mu }^{-1}\displaystyle \sum _{j=1}^{p}\left({\breve{\phi }}_{j}(t)-{\phi }_{j}\right){\mathop{\breve{\phi }}\limits^{\cdot }}_{j}(t)\\ +\,2{\nu }^{-1}\displaystyle \sum _{j=1}^{q}\left({\breve{\psi }}_{j}(t)-{\psi }_{j}\right){\mathop{\breve{\psi }}\limits^{\cdot }}_{j}(t),\end{array}\end{eqnarray}$where$\begin{eqnarray*}\begin{array}{rcl}{{\rm{\Xi }}}_{2} & = & {\mathfrak{H}}e\left(P({ \mathcal A }+K{ \mathcal C })\right)+\frac{1}{{\theta }_{1}}P{ \mathcal B }{{ \mathcal B }}^{{\rm{T}}}P\\ & & +\ {\theta }_{1}{{ \mathcal H }}^{{\rm{T}}}{L}_{f}^{{\rm{T}}}{L}_{f}{ \mathcal H }+\frac{1}{{\theta }_{2}}P{{ \mathcal B }}_{\delta }{{ \mathcal B }}_{\delta }^{{\rm{T}}}P+Q+S.\end{array}\end{eqnarray*}$According to Schur's complement, (7) ensures that$\begin{eqnarray}\left[\begin{array}{ccc}{{\rm{\Xi }}}_{2} & P{{ \mathcal A }}_{\delta } & P{ \mathcal E }\\ * & {\theta }_{2}{{ \mathcal H }}^{{\rm{T}}}{L}_{g}^{{\rm{T}}}{L}_{g}{ \mathcal H }-Q & 0\\ * & * & -R\end{array}\right]\lt 0.\end{eqnarray}$Using (8), (9), and (16), we have from (15) that$\begin{eqnarray}\begin{array}{l}\dot{V}(t)\leqslant -{\epsilon }^{{\rm{T}}}(t)S\epsilon (t)+{\omega }^{{\rm{T}}}(t)R\omega (t)\\ -\,2(1+2\varepsilon ){\mu }^{-1}\displaystyle \sum _{j=1}^{p}{\left({\breve{\phi }}_{j}(t)-{\phi }_{j}\right)}^{2}\\ -\,2(1+2\varepsilon ){\nu }^{-1}\displaystyle \sum _{j=1}^{q}{\left({\breve{\psi }}_{j}(t)-{\psi }_{j}\right)}^{2}\\ \leqslant -{\epsilon }^{{\rm{T}}}(t)S\epsilon (t)+{\omega }^{{\rm{T}}}(t)R\omega (t)-4\varepsilon {\mu }^{-1}\displaystyle \sum _{j=1}^{p}\\ {\left({\breve{\phi }}_{j}(t)-{\phi }_{j}\right)}^{2}-4\varepsilon {\nu }^{-1}\displaystyle \sum _{j=1}^{q}{\left({\breve{\psi }}_{j}(t)-{\psi }_{j}\right)}^{2}.\end{array}\end{eqnarray}$Deriving the derivative of e2ϵtV(t) and using (11) and (17), one gains$\begin{eqnarray}\begin{array}{l}\frac{{\rm{d}}\left({{\rm{e}}}^{2\varepsilon t}V(t)\right)}{{dt}}=2\varepsilon {{\rm{e}}}^{2\varepsilon t}V(t)+{{\rm{e}}}^{2\varepsilon t}\dot{V}(t)\\ \leqslant \,2\varepsilon {{\rm{e}}}^{2\varepsilon t}{\epsilon }^{{\rm{T}}}(t)P\epsilon (t)+2\varepsilon {{\rm{e}}}^{2\varepsilon t}{\int }_{t-\delta }^{t}{\epsilon }^{{\rm{T}}}(\rho )Q\epsilon (\rho ){\rm{d}}\rho \\ -\,{{\rm{e}}}^{2\varepsilon t}{\epsilon }^{{\rm{T}}}(t)S\epsilon (t)+{{\rm{e}}}^{2\varepsilon t}{\omega }^{{\rm{T}}}(t)R\omega (t).\end{array}\end{eqnarray}$By integrating both sides of (18) from 0 to t, one has$\begin{eqnarray}\begin{array}{l}{{\rm{e}}}^{2\varepsilon t}V(t)-V(0)\leqslant {\int }_{0}^{t}2\varepsilon {{\rm{e}}}^{2\varepsilon r}{\epsilon }^{{\rm{T}}}(r)P\epsilon (r){\rm{d}}r\\ +{\int }_{0}^{t}\left(2\varepsilon {{\rm{e}}}^{2\varepsilon r}{\int }_{r-\delta }^{r}{\epsilon }^{{\rm{T}}}(\rho )Q\epsilon (\rho ){\rm{d}}\rho \right){\rm{d}}r\\ -{\int }_{0}^{t}{{\rm{e}}}^{2\varepsilon r}{\epsilon }^{{\rm{T}}}(r)S\epsilon (r){\rm{d}}r+{\int }_{0}^{t}{{\rm{e}}}^{2\varepsilon r}{\omega }^{{\rm{T}}}(r)R\omega (r){\rm{d}}r.\end{array}\end{eqnarray}$From the second term on the right hand side of (19), it can be calculated that$\begin{eqnarray*}\begin{array}{l}{\int }_{0}^{t}\left({\int }_{r-\delta }^{r}2\varepsilon {{\rm{e}}}^{2\varepsilon r}{\epsilon }^{{\rm{T}}}(\rho )Q\epsilon (\rho ){\rm{d}}\rho \right){\rm{d}}r\\ \leqslant \,\,{\int }_{-\delta }^{t}\left({\int }_{\rho }^{\rho +\delta }2\varepsilon {{\rm{e}}}^{2\varepsilon r}{\epsilon }^{{\rm{T}}}(\rho )Q\epsilon (\rho ){\rm{d}}r\right){\rm{d}}\rho \\ \leqslant \,{\int }_{-\delta }^{0}({{\rm{e}}}^{2\varepsilon \delta }-1){\epsilon }^{{\rm{T}}}(\rho )Q\epsilon (\rho ){\rm{d}}\rho \\ +\,{\int }_{0}^{t}({{\rm{e}}}^{2\varepsilon \delta }-1){{\rm{e}}}^{2\varepsilon \rho }{\epsilon }^{{\rm{T}}}(\rho )Q\epsilon (\rho ){\rm{d}}\rho .\end{array}\end{eqnarray*}$Substituting the above inequality into (19), we can obtain$\begin{eqnarray}\begin{array}{l}V(t)\leqslant \left[{\int }_{0}^{t}\left(2\varepsilon {{\rm{e}}}^{2\varepsilon \rho }{\epsilon }^{{\rm{T}}}\right.\right.\\ \left.(\rho )P\epsilon (\rho )+({{\rm{e}}}^{2\varepsilon \delta }-1){{\rm{e}}}^{2\varepsilon \rho }{\epsilon }^{{\rm{T}}}(\rho )Q\epsilon (\rho )\\ \left.-\ {{\rm{e}}}^{2\varepsilon \rho }{\epsilon }^{{\rm{T}}}(\rho )S\epsilon (\rho )\right){\rm{d}}\rho \\ +\,{{\rm{e}}}^{2\varepsilon \delta }{\int }_{-\delta }^{0}{\epsilon }^{{\rm{T}}}(\rho )Q\epsilon (\rho ){\rm{d}}\rho +{\epsilon }^{{\rm{T}}}(0)P\epsilon (0)\\ \left.\ +\,{\mu }^{-1}\displaystyle \sum _{j=1}^{p}{\left({\breve{\phi }}_{j}(0)-{\phi }_{j}\right)}^{2}+{\nu }^{-1}\displaystyle \sum _{j=1}^{q}{\left({\breve{\psi }}_{j}(0)-{\psi }_{j}\right)}^{2}\right]\\ {{\rm{e}}}^{-2\varepsilon t}+\frac{1-{{\rm{e}}}^{-2\varepsilon t}}{2\varepsilon }{\lambda }_{M}(R)\mathop{\sup }\limits_{\rho \in [0,t]}{\parallel \omega (\rho )\parallel }^{2}\\ \leqslant \,\left[,[,\left(2\varepsilon {\lambda }_{M}(P)+({{\rm{e}}}^{2\varepsilon \delta }-1){\lambda }_{M}(Q)-{\lambda }_{m}(S)\right)\right.\\ {\int }_{0}^{t}{{\rm{e}}}^{2\varepsilon \rho }{\epsilon }^{{\rm{T}}}(\rho )\epsilon (\rho ){\rm{d}}\rho +{{\rm{e}}}^{2\varepsilon \delta }{\int }_{-\delta }^{0}{\epsilon }^{{\rm{T}}}(\rho )Q\epsilon (\rho ){\rm{d}}\rho \\ +\,{\lambda }_{M}(P){\parallel \epsilon (0)\parallel }^{2}+{\mu }^{-1}\displaystyle \sum _{j=1}^{p}{\left({\breve{\phi }}_{j}(0)-{\phi }_{j}\right)}^{2}\\ +\,{\nu }^{-1}\displaystyle \sum _{j=1}^{q}{\left({\breve{\psi }}_{j}(0)-{\psi }_{j}\right)}^{2}\right]{{\rm{e}}}^{-2\varepsilon t}\\ +\,\frac{{\lambda }_{M}(R)}{2\varepsilon }\mathop{\sup }\limits_{\rho \in [0,t]}{\parallel \omega (\rho )\parallel }^{2}.\end{array}\end{eqnarray}$On the basis of (10), it follows by (20) that$\begin{eqnarray}\begin{array}{l}V(t)\leqslant \left[{{\rm{e}}}^{2\varepsilon \delta }{\int }_{-\delta }^{0}{\epsilon }^{{\rm{T}}}(\rho )Q\epsilon (\rho ){\rm{d}}\rho +{\lambda }_{M}(P){\parallel \epsilon (0)\parallel }^{2}\right.\\ +\,\left.{\mu }^{-1}\displaystyle \sum _{j=1}^{p}{\left({\breve{\phi }}_{j}(0)-{\phi }_{j}\right)}^{2}+{\nu }^{-1}\displaystyle \sum _{j=1}^{q}{\left({\breve{\psi }}_{j}(0)-{\psi }_{j}\right)}^{2}\right]\\ {{\rm{e}}}^{-2\varepsilon t}+\frac{{\lambda }_{M}(R)}{2\varepsilon }\mathop{\sup }\limits_{\rho \in [0,t]}{\parallel \omega (\rho )\parallel }^{2}.\end{array}\end{eqnarray}$On the other hand, it is effortless to get$\begin{eqnarray}{\lambda }_{m}(P){\parallel \epsilon (t)\parallel }^{2}\leqslant V(t).\end{eqnarray}$Now, using (21), (22) and lemma 2, we can get (6), where$\begin{eqnarray*}\beta \left(t\right)=\sqrt{\displaystyle \frac{{{ \mathcal C }}_{0}}{{\lambda }_{m}(P)}}{{\rm{e}}}^{-\varepsilon t},\gamma \left(\rho \right)=\sqrt{\displaystyle \frac{{\lambda }_{M}(R)}{2\varepsilon {\lambda }_{m}(P)}}\mathop{\sup }\limits_{\rho \in [0,t]}\parallel \omega (\rho )\parallel \end{eqnarray*}$with$\begin{eqnarray*}\begin{array}{l}{{ \mathcal C }}_{0}=\left({{\rm{e}}}^{2\varepsilon \delta }\delta {\lambda }_{\max }(Q)+{\lambda }_{\max }(P)\right){\left(\mathop{\sup }\limits_{\rho \in [-\delta ,0]}\parallel \epsilon (\rho )\parallel \right)}^{2}\\ +{\mu }^{-1}\sum _{j=1}^{p}{\left({\breve{\phi }}_{j}(0)-{\phi }_{j}\right)}^{2}+{\nu }^{-1}\sum _{j=1}^{q}{\left({\breve{\psi }}_{j}(0)-{\psi }_{j}\right)}^{2}.\end{array}\end{eqnarray*}$Therefore, system (5) is ISS and the proof is finished.When Ω(t) ≡ 0, it can be seen from the above proof that system (5) is exponentially stable with decay rate ϵ under the conditions of theorem 1.
Next, attention will be concentrated on the no parameter uncertainties case (i.e. Øj ≡ ψj ≡ 0). In this case, system (5) becomes$\begin{eqnarray}\begin{array}{rcl}\dot{\epsilon }(t) & = & ({ \mathcal A }+K{ \mathcal C })\epsilon (t)+{{ \mathcal A }}_{\delta }\epsilon (t-\delta )+{ \mathcal B }f\left({ \mathcal H }\epsilon (t)\right)\\ & & +{{ \mathcal B }}_{\delta }g\left({ \mathcal H }\epsilon (t-\delta )\right)+{ \mathcal E }\omega (t).\end{array}\end{eqnarray}$And one can write the following results:
System (23) is ISS if there exist scalars θ1 > 0, θ2 > 0 and matrices P, Q, R, $S\in {{\mathbb{S}}}_{+}^{n}$ such that$\begin{eqnarray}\left[\begin{array}{ccccc}{{\rm{\Xi }}}_{1} & P{{ \mathcal A }}_{\delta } & P{ \mathcal E } & P{ \mathcal B } & P{{ \mathcal B }}_{\delta }\\ * & {\theta }_{2}{{ \mathcal H }}^{{\rm{T}}}{L}_{g}^{{\rm{T}}}{L}_{g}{ \mathcal H }-Q & 0 & 0 & 0\\ * & * & -R & 0 & 0\\ * & * & * & -{\theta }_{1}I & 0\\ * & * & * & * & -{\theta }_{2}I\end{array}\right]\lt 0,\end{eqnarray}$where$\begin{eqnarray*}{{\rm{\Xi }}}_{1}={\mathfrak{H}}e\left(P{ \mathcal A }+{PK}{ \mathcal C }\right)+{\theta }_{1}{{ \mathcal H }}^{{\rm{T}}}{L}_{f}^{{\rm{T}}}{L}_{f}{ \mathcal H }+Q+S.\end{eqnarray*}$
Proof. Let us select a Lyapunov function as$\begin{eqnarray}V(t)={\epsilon }^{{\rm{T}}}(t)P\epsilon (t)+{\int }_{t-\delta }^{t}{\epsilon }^{{\rm{T}}}(\rho )Q\epsilon (\rho ){\rm{d}}\rho .\end{eqnarray}$Then, along parallel lines as the proof of theorem 1, one can readily reveal that system (23) is ISS for any bounded interference Ω(t).
In [16], the ISS filtering issue for delayed nonlinear systems was considered. With the application of a more complicated Lyapunov function as$\begin{eqnarray*}\begin{array}{rcl}V(t) & = & {\epsilon }^{{\rm{T}}}(t)P\epsilon (t)+{\int }_{t-\delta }^{t}{\epsilon }^{{\rm{T}}}(\rho )Q\epsilon (\rho ){\rm{d}}\rho \\ & & +\ {\int }_{t-\delta }^{t}\left(\rho -t+\delta \right){\epsilon }^{{\rm{T}}}(\rho )R\epsilon (\rho ){\rm{d}}\rho \\ & & +{\left[{\int }_{t-\delta }^{t}\epsilon (\rho ){\rm{d}}\rho \right]}^{{\rm{T}}}U\left[{\int }_{t-\delta }^{t}\epsilon (\rho ){\rm{d}}\rho \right],\end{array}\end{eqnarray*}$together with the analysis approach therein, the following criterion concerning ISS of system (23) was obtained:[16] System (23) is ISS if there exist matrices P, Q, R, S1, S2, $U\in {{\mathbb{S}}}_{+}^{n}$ such that$\begin{eqnarray}\left[\begin{array}{cccccc}{{\rm{\Lambda }}}_{1} & P{{ \mathcal A }}_{\delta } & U & P{ \mathcal E } & P{ \mathcal B } & P{{ \mathcal B }}_{\delta }\\ * & {{ \mathcal H }}^{{\rm{T}}}{L}_{g}^{{\rm{T}}}{L}_{g}{ \mathcal H }-Q & -U & 0 & 0 & 0\\ * & * & -\frac{1}{2\delta }{S}_{2} & 0 & 0 & 0\\ * & * & * & -R & 0 & 0\\ * & * & * & * & -I & 0\\ * & * & * & * & * & -I\end{array}\right]\lt 0,\end{eqnarray}$where$\begin{eqnarray*}{{\rm{\Lambda }}}_{1}={\mathfrak{H}}e\left(P{ \mathcal A }+{PK}{ \mathcal C }\right)+{{ \mathcal H }}^{{\rm{T}}}{L}_{f}^{{\rm{T}}}{L}_{f}{ \mathcal H }+\delta {S}_{2}+Q+{S}_{1}.\end{eqnarray*}$
With the aid of Schur's complement, it is not difficult to see that (24) is a necessary condition for establishing (26), which means that the present ISS analysis approach is less conservative than that proposed in [16].
Now we are in a position to present our design method for the adaptive synchronization controller in (4) via the following theorem.Suppose that there are scalars θ1 > 0, θ2 > 0, matrices P, Q, R, $S\in {{\mathbb{S}}}_{+}^{n}$, and a matrix $M\in {{\mathbb{R}}}^{n\times m}$ such that$\begin{eqnarray}\left[\begin{array}{ccccc}{{\rm{\Xi }}}_{1} & P{{ \mathcal A }}_{\delta } & P{ \mathcal E } & P{ \mathcal B } & P{{ \mathcal B }}_{\delta }\\ * & {\theta }_{2}{{ \mathcal H }}^{{\rm{T}}}{L}_{g}^{{\rm{T}}}{L}_{g}{ \mathcal H }-Q & 0 & 0 & 0\\ * & * & -R & 0 & 0\\ * & * & * & -{\theta }_{1}I & 0\\ * & * & * & * & -{\theta }_{2}I\end{array}\right]\lt 0,\end{eqnarray}$where$\begin{eqnarray*}\begin{array}{rcl}{{\rm{\Xi }}}_{1} & = & {\mathfrak{H}}e\left(P{ \mathcal A }+M{ \mathcal C }\right)+{\theta }_{1}{{ \mathcal H }}^{{\rm{T}}}{L}_{f}^{{\rm{T}}}{L}_{f}{ \mathcal H }+Q+S,\\ {L}_{f} & = & \mathrm{diag}\left\{{L}_{f1},\cdots ,{L}_{\mathrm{fm}}\right\},\,\,{L}_{g}=\mathrm{diag}\left\{{L}_{g1},\cdots ,{L}_{\mathrm{gm}}\right\}.\end{array}\end{eqnarray*}$Then, under the adaptive controller in (4) with gain K = P−1M as well as updating laws (8) and (9), systems (1) and (2) are ISS synchronized.
Proof. With PK=M, the inequality of (7) can be rewritten as that of (27). Thus, systems (1 ) and (2) are ISS synchronized according to definition 2.
Once the LMIs in theorem 2 are feasible, one can obtain the feedback gain K as well as the estimates of the uncertain parameters ${\phi }_{j}$ and ψj simultaneously. Note that there are efficient numerical algorithms available to check the feasibility of LMIs, such as alternating projection method, the ellipsoid method, and the interior-point method. Thus, the present design method of adaptive controller in (4) for ISS synchronization is numerically tractable.
4. Example
The section provides an example with simulations to show the applicability and superiority of the analysis and design approach.Consider the drive and response Lur'e systems in (1) and (2) subject to parameters$\begin{eqnarray*}\begin{array}{rcl}{ \mathcal A } & = & \left[\begin{array}{ccc}-\displaystyle \frac{18}{7} & 9 & 0\\ 1 & -1 & 1\\ 0 & -14.286 & 0\end{array}\right],\\ {{ \mathcal A }}_{\delta } & = & \left[\begin{array}{ccc}-0.1 & 0 & 0\\ -0.1 & 0 & 0\\ 0.2 & 0 & -0.1\end{array}\right],\,{ \mathcal B }=\left[\begin{array}{ccc}\displaystyle \frac{27}{7} & 0 & 0\\ 0 & 0 & 0\\ 0 & 0 & 0\end{array}\right],\\ {{ \mathcal B }}_{\delta } & = & \left[\begin{array}{ccc}0 & 0 & 0\\ 0 & 0 & 0\\ 0 & 0 & 0\end{array}\right],\,{{ \mathcal E }}_{1}=\left[\begin{array}{c}-1\\ 0\\ 0\end{array}\right],\,{{ \mathcal E }}_{2}=\left[\begin{array}{c}0\\ 1\\ 0\end{array}\right],\\ { \mathcal C } & = & {\left[\begin{array}{c}1\\ 0\\ 0\end{array}\right]}^{{\rm{T}}},\,{ \mathcal H }=\left[\begin{array}{ccc}1 & 0 & 0\\ 0 & 0 & 0\\ 0 & 0 & 0\end{array}\right],\,\\ {\zeta }_{1}(t) & = & \left[\begin{array}{c}{\zeta }_{11}(t)\\ {\zeta }_{12}(t)\\ {\zeta }_{13}(t)\end{array}\right],\,{{\rm{\Phi }}}_{1}({\zeta }_{1}(t))=\left[\begin{array}{c}F\left({ \mathcal H }{\zeta }_{11}(t)\right)\\ 0\\ 0\end{array}\right],\\ {{\rm{\Psi }}}_{1}({\zeta }_{1}(t)) & = & \left[\begin{array}{c}G\left({ \mathcal H }{\zeta }_{11}(t)\right)\\ 0\\ 0\end{array}\right]\\ F\left({ \mathcal H }{\zeta }_{11}(t)\right) & = & G\left({ \mathcal H }{\zeta }_{11}(t)\right)=\displaystyle \frac{1}{2}\left(\left|{\zeta }_{11}(t)+1\right|-\left|{\zeta }_{11}(t)-1\right|\right),\\ F\left({ \mathcal H }{\zeta }_{12}(t)\right) & = & F\left({ \mathcal H }{\zeta }_{13}(t)\right)=G\left({ \mathcal H }{\zeta }_{12}(t)\right)=G\left({ \mathcal H }{\zeta }_{13}(t)\right)=0,\\ \omega (t) & = & 0.2\sin (10t),p\,=\,1,\,q\,=\,1.\end{array}\end{eqnarray*}$Notice Assumption 1 is satisfied with ${L}_{f}={L}_{g}={\rm{diag}}\{1,0,0\}.$
When Øj ≡ ψj ≡ 0, (i.e. there are no parameter uncertainties), then the system considered in this example can be viewed as a time-delay Chua's circuit [35]. For this situation, both proposition 1 and corollary 1 can be used to check the ISS of the SES. However, as theoretically explained in remark 5, the former is much conservative than the latter. To illustrate this, let us consider the case that $K=k{[\begin{array}{ccc}1 & 1 & 1\end{array}]}^{{\rm{T}}}$. Then, it can be verified that the LMIs in proposition 1 hold true for k ≤ −3.062, while, by corollary 1, it is observed that the ISS can be ensured even for k ∈ [ −3.062, −1.651].
Next, let us show the applicability of our design approach. Solving the LMIs in theorem 2, we can obtain a feasible solution, where$\begin{eqnarray*}\begin{array}{rcl}P & = & \left[\begin{array}{ccc}16.4135 & -21.0625 & 2.4124\\ -21.0625 & 78.6244 & -10.4966\\ 2.4124 & -10.4966 & 7.1752\end{array}\right],\\ M & = & \left[\begin{array}{c}-128.7534\\ -212.3368\\ 31.7822\end{array}\right],\\ Q & = & \left[\begin{array}{ccc}109.8950 & -0.0001 & -0.1741\\ -0.0001 & 51.7207 & 1.1786\\ -0.1741 & 1.1786 & 9.7692\end{array}\right],\,R=\begin{array}{c}95.3604\end{array},\\ S & = & \left[\begin{array}{ccc}67.6844 & 0.0000 & 0.0000\\ 0.0000 & 41.8412 & 1.0287\\ 0.0000 & 1.0287 & 5.3038\end{array}\right],\\ {\theta }_{1} & = & 99.2800,\,{\theta }_{2}=61.2565.\end{array}\end{eqnarray*}$Therefore, according to theorem 2, the ISS synchronization of the drive and response systems can be ensured under the controller in (4) with feedback gain$\begin{eqnarray}K={P}^{-1}M={\left[\begin{array}{ccc}-17.2568 & -7.4035 & -0.5991\end{array}\right]}^{{\rm{T}}}\end{eqnarray}$as well as updating laws in (8) and (9).
The simulations are carried out in MATLAB R2020b. As in [36], we consider the case when the uncertain parameters have different orders of magnitude. More specifically, the following two cases will be examined: Case A. Ø1 = 0.02, ψ1 = − 0.005; Case B. Ø1 = − 0.75, ψ1 = − 0.075. Under initial condition$\begin{eqnarray*}{\zeta }_{1}(\rho )={\left[-0.2-0.30.2\right]}^{{\rm{T}}},\,\rho \in [-\delta ,0],\,\,\delta =1,\end{eqnarray*}$figure 1 shows the chaotic behavior of drive system (1) for Case A. For Case B, the chaotic behavior is similar to figure 1 and therefore omitted herein for brevity.
Next, set μ = 0.1, ν = 0.1, $\varepsilon =0.001,\,{\breve{\phi }}_{1}(0)=-0.5$, ${\breve{\psi }}_{1}(0)=0.3$, and randomly choose the initial value of response system (2) between −1 and 1 as follows:$\begin{eqnarray*}{\zeta }_{2}(\rho )={\left[\begin{array}{ccc}-0.7613 & -0.1478 & 0.5349\end{array}\right]}^{{\rm{T}}},\,\rho \in [-1,0].\end{eqnarray*}$Figures 2 and 3 show the state of the SES and the estimates ${\breve{\phi }}_{1}(t)$, ${\breve{\psi }}_{1}(t)$ of Ø1, ψ1 with exterior interference for Case A and Case B, respectively; figures 4 and 5 further show the trajectories in the absence of exterior interference. It is found from these figures that the SES based on the designed adaptive controller is state-bounded under the bounded interference and quickly convergent in the absence of interference. Therefore, the simulations coincide with the present theoretical results.
Figure 2.
New window|Download| PPT slide Figure 2.Trajectories of ∈i(t)(i = 1, 2, 3), ${\breve{\phi }}_{1}(t)$, and ${\breve{\psi }}_{1}(t)$ when (Ø1, ψ1) = (0.02, −0.005).
Figure 3.
New window|Download| PPT slide Figure 3.Trajectories of ∈i(t)(i = 1, 2, 3), ${\breve{\phi }}_{1}(t)$, and ${\breve{\psi }}_{1}(t)$ when (Ø1, ψ1) = ( − 0.75, −0.075).
Figure 4.
New window|Download| PPT slide Figure 4.Trajectories of ∈i(t)(i = 1, 2, 3), ${\breve{\phi }}_{1}(t)$, and ${\breve{\psi }}_{1}(t)$ when (Ø1, ψ1) = (0.02, −0.005) and Ω(t) ≡ 0.
Figure 5.
New window|Download| PPT slide Figure 5.Trajectories of ∈i(t)(i = 1, 2, 3), ${\breve{\phi }}_{1}(t)$, and ${\breve{\psi }}_{1}(t)$ when (Ø1, ψ1) = ( − 0.75, −0.075) and Ω(t) ≡ 0.
5. Conclusions
In this paper, the issue of adaptive ISS synchronization of time-delay Lur'e systems has been studied, where the exterior interference and parameter uncertainties have been considered simultaneously. With the aid of a Lyapunov function and some inequality techniques, a co-design of the feedback gain and estimates of the uncertain parameters has been proposed to determine the desired adaptive synchronization controller. For showing the applicability and superiority of the analysis and design approaches, an example with simulations has been given.
Acknowledgements
This work was supported by the Natural Science Foundation of the Anhui Higher Education Institutions (Grant No. KJ2020A0248), the National Natural Science Foundation of China (Grant Nos. 61806004 and 61503002), and the Open Project of Anhui Province Key Laboratory of Special and Heavy Load Robot (Grant No. TZJQR005-2020).
BowongSKakmeniF M MKoinaR2004 A new synchronization principle for a class of Lur'e systems with applications in secure communication Int. J. Bifurcation Chaos 14 24772491 DOI:10.1142/S0218127404010758 [Cited within: 1]
VolosC KKyprianidisI MStouboulosI N2013 Image encryption process based on chaotic synchronization phenomena Signal Process. 93 13281340 DOI:10.1016/j.sigpro.2012.11.008 [Cited within: 1]
BouridahM SBoudenTYalçinM E2020 Chaos synchronization of fractional-order Lur'e systems Int. J. Bifurcation Chaos 30 2050206 DOI:10.1142/S0218127420502065 [Cited within: 1]
HuangYBaoH2020 Master-slave synchronization of complex-valued delayed chaotic Lur'e systems with sampled-data control Appl. Math. Comput. 379 125261 DOI:10.1016/j.amc.2020.125261
ParkJ HLeeT HLiuYChenJ2019Dynamic Systems with Time Delays: Stability and Control SingaporeSpringer DOI:10.1007/978-981-13-9254-2
KwonWKooBLeeS M2018 Integral-based event-triggered synchronization criteria for chaotic Lur'e systems with networked PD control Nonlinear Dyn. 94 9911002 DOI:10.1007/s11071-018-4405-9 [Cited within: 1]
DongYXianJ2016 Finite-time quasi-synchronization of two nonidentical chaotic systems via intermittent control Commun. Theor. Phys. 66 306 DOI:10.1088/0253-6102/66/3/306 [Cited within: 1]
AguilarLOrlovYPisanoA2020 Leader-follower synchronization and ISS analysis for a network of boundary-controlled wave PDEs IEEE Control Syst. Lett. 5 683688 DOI:10.1109/LCSYS.2020.3004505 [Cited within: 1]
LiCZhangJ2016 Synchronisation of a fractional-order chaotic system using finite-time input-to-state stability Int. J. Syst. Sci. 47 24402448 DOI:10.1080/00207721.2014.998741
ChenDYangGHanZ2012 Impulsive observer for input-to-state stability based synchronization of Lur'e differential inclusion system Commun. Nonlinear Sci. Numer. Simul. 17 29902996 DOI:10.1016/j.cnsns.2011.10.023
LiTYuanRFeiSDingZ2018 Sampled-data synchronization of chaotic Lur'e systems via an adaptive event-triggered approach Inf. Sci. 462 4054 DOI:10.1016/j.ins.2018.06.012 [Cited within: 2]
TangZParkJ HFengJ2018 Novel approaches to pin cluster synchronization on complex dynamical networks in Lur'e forms Commun. Nonlinear Sci. Numer. Simul. 57 422438 DOI:10.1016/j.cnsns.2017.10.010
AbdM HTahirF RAl-SuhailG APhamV T2017 An adaptive observer synchronization using chaotic time-delay system for secure communication Nonlinear Dyn. 90 25832598 DOI:10.1007/s11071-017-3825-2 [Cited within: 2]
AhnC K2013 Input-to-state stable nonlinear filtering for a class of continuous-time delayed nonlinear systems Int. J. Control 86 11791185 DOI:10.1080/00207179.2013.787547 [Cited within: 4]
LiQLiuS2018 Switching event-triggered network-synchronization for chaotic systems with different dimensions Neurocomputing 311 3240 DOI:10.1016/j.neucom.2018.05.039
ParkJ MParkP G2018 ${{ \mathcal H }}_{\infty }$ sampled-state feedback control for synchronization of chaotic Lur'e systems with time delays J. Franklin Inst. 355 80058026 DOI:10.1016/j.jfranklin.2018.08.019 [Cited within: 1]
DuanWLiYSunYChenJYangX2020 Enhanced master-slave synchronization criteria for chaotic Lur'e systems based on time-delayed feedback control Math. Comput. Simul. 177 276294 DOI:10.1016/j.matcom.2020.04.010 [Cited within: 1]
RaoYTongDChenQZhouWXuY2021 Synchronization of chaotic Lur'e systems with time-delays via quantized output feedback control Trans. Inst. Meas. Control 43 933944 DOI:10.1177/0142331220950864 [Cited within: 1]
ZhaoMLiuRGaoY2017 Dissipative lag synchronization of chaotic Lur'e systems with unknown disturbances IMA J. Math. Control Inf. 34 123138 DOI:10.1093/imamci/dnv034 [Cited within: 1]
YangZZhouWHuangT2014 Exponential input-to-state stability of recurrent neural networks with multiple time-varying delays Cogn. Neurodyn. 8 4754 DOI:10.1007/s11571-013-9258-9 [Cited within: 1]
ZhouYLiuYZhouJWangZ2021 Quantized passive filtering for switched delayed neural networks Nonlinear Anal.: Modelling Control 26 93112 DOI:10.15388/namc.2021.26.20562 [Cited within: 1]
ArikS2020 New criteria for stability of neutral-type neural networks with multiple time delays IEEE Trans. Neural Netw. Learn. Syst. 31 15041513 DOI:10.1109/TNNLS.2019.2920672
HuangXZhouYKongQZhouJFangM2020 ${{ \mathcal H }}_{\infty }$ synchronization of chaotic Hopfield networks with time-varying delay: a resilient DOF control approach Commun. Theor. Phys. 72 015003 DOI:10.1088/1572-9494/ab5452
TaiWZuoDXuanZZhouJWangZ2021 Non-fragile ${{ \mathcal L }}_{2}-{{ \mathcal L }}_{\infty }$ filtering for a class of switched neural networks Math. Comput. Simul. 185 629645 DOI:10.1016/j.matcom.2021.01.014
SakthivelRWangCSantraSKaviarasanB2018 Non-fragile reliable sampled-data controller for nonlinear switched time-varying systems Nonlinear Anal. Hybrid Syst. 27 6276 DOI:10.1016/j.nahs.2017.08.005 [Cited within: 1]
ZhouKKhargonekarP P1988 Robust stabilization of linear systems with norm-bounded time-varying uncertainty Syst. Control Lett. 10 1720 DOI:10.1016/0167-6911(88)90034-5 [Cited within: 1]
LuJCaoJHoD W C2008 Adaptive stabilization and synchronization for chaotic Lur'e systems with time-varying delay IEEE Trans. Circuits Syst. I: Regular Papers 55 13471356 DOI:10.1109/TCSI.2008.916462 [Cited within: 1]
,2,*, Jianping Zhou(周建平)1,3,4, Yebin Chen(陈业斌)11School of Computer Science & Technology, 
New window|Download| PPT slide
New window|Download| PPT slide
New window|Download| PPT slide
New window|Download| PPT slide
New window|Download| PPT slide
