Next Article in Journal
Machine Learning (ML) in Medicine: Review, Applications, and Challenges
Previous Article in Journal
Inviscid Modes within the Boundary-Layer Flow of a Rotating Disk with Wall Suction and in an External Free-Stream
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Observer-Based Control for Nonlinear Time-Delayed Asynchronously Switching Systems: A New LMI Approach

1
Faculty of Electrical and Computer Engineering, University of Tabriz, Tabriz 5166616471, Iran
2
Electrical Engineering Department, University of Bonab, Bonab 5551395133, Iran
3
Department of Electrical Engineering, Ilam University, Ilam 69315516, Iran
4
Future Technology Research Center, National Yunlin University of Science and Technology, Douliu 64002, Taiwan
5
Advanced Research Center for Computational Simulation, Department of Mathematics, Faculty of Science, Chiang Mai University, Chiang Mai 50200, Thailand
6
Department of Physics, Faculty of Applied Sciences, Palestine Technical University, Tulkarm P.O. Box 7, Palestine
7
Department of Cyber-Physical Systems, St. Petersburg State Marine Technical University, 190121 Saint-Petersburg, Russia
*
Authors to whom correspondence should be addressed.
Submission received: 24 October 2021 / Revised: 14 November 2021 / Accepted: 18 November 2021 / Published: 21 November 2021

Abstract

:
This paper designs an observer-based controller for switched systems (SSs) with nonlinear dynamics, exogenous disturbances, parametric uncertainties, and time-delay. Based on the multiple Lyapunov–Krasovskii and average dwell time (DT) approaches, some conditions are presented to ensure the robustness and investigate the effect of time-delay, uncertainties, and lag issues between switching times. The control parameters are determined through solving the established linear matrix inequalities (LMIs) under asynchronous switching. A novel LMI-based conditions are suggested to guarantee the H performance. Finally, the accuracy of the designed observer-based controller is examined by simulations on practical case-study plants.

1. Introduction

Switched systems (SSs) are a class of hybrid systems due to their strong potential applications [1]. Various physical problems can be modeled and represented by SSs. Moreover, SSs consist of different modes with a switching signal to determine the active mode within any time intervals. The main challenges are the effect of time-delay, uncertain parameters, and the effect of switching instants [2].
In recent years, controller design of SSs has been one of the most significant aspects of the research in the literature. For example, by the use of the DT method, the state-output feedback control systems are studied in [3]. The Lyapunov–Krasovskii approach is developed in [4,5], to investigate the effect of time-delay. As is well known, parameter variations or an error in the measurement of system parameters may cause uncertainty in many dynamical systems. Constructing a piecewise linear Lyapunov function, the robust stability conditions are obtained for linear SSs subject to polytopic uncertainties [6]. In [7], both state/output feedback schemes are developed in the framework of LMI to analyze the uncertain singular SS. The robust H control approach is investigated in [8] to stabilize a discrete-time SS against polytopic uncertainties. In [9], the stability of SSs with time-delayed switchings and bounded uncertainties is investigated based on generalized polyhedral cells. The problem of polytopic uncertainties under time-delay condition has been studied in [10,11]. The effect of affine parametric uncertainties is investigated in [12,13] and the upper bounds of uncertainties are determined via computational algorithms. It should be noted that, in polytopic uncertain problems, a great number of LMIs should be solved to cope with uncertain parameters. Therefore, the computational cost of the problem is high and needs more studies. The computational characteristics are studied in [14], and it is shown that less conservative analysis is required in comparison with upper bound computations of uncertainties. The stabilization of SSs in a discrete-time form is investigated in [15], and the effect of uncertainties and time-delay is analyzed by the use of output feedback control system. Similarly, the state feedback control system is developed in [16] for discrete-time SSs and the stabilization conditions are acquired. The predictive control approach is suggested in [17] to ensure the robustness of discrete-time SSs with variable switching laws and time-delays.
The asynchronous switching (AS) problem is another challenging issue in SSs. The lag among the switching times and associated controllers commonly cause the asynchronous switching problem [18]. This problem has been rarely studied for SSs under time-delay condition. For instance, a Lyapunov method is suggested in [19] to compute L 2 gains of SSs with exogenous perturbations and AS problem. An asynchronous dynamic control technique is addressed for time-delayed SSs in [20]. The AS problem is taken into account in [21], and the finite-time stability criteria and state feedback control scheme are suggested to investigate the stabilization problem. The tracking accuracy of linear SSs subject to time-delays and AS is studied in [22].
Utilizing the average DT approach (ADT), the boundedness of nonlinear SSs is studied in [23], and the input delay and AS effects are investigated. The H problem is studied in [24] for discrete-time SSs, and the AS problem is analyzed using non-fragile controllers. The Lyapunov–Krasovskii approach is formulated in [25], and the asynchronous stabilization of neutral SSs is analyzed. The free-weighting matrices technique on the basis of the ADT method is developed in [26] to design a stable state feedback controller for linear SSs subject to the AS problem. The zonotope method is employed in [27] to acquire the appropriate estimations and then the computational complexity of the robust H scheme is analyzed.
The robust control of nonlinear SSs under unmeasurable states, time-delay, uncertainties, and the AS problem has not been completely studied. The main contributions are:
  • Unlike the reviewed research, besides the time-delay and asynchronous switching, the parametric uncertainties are also considered in this study to ensure the robust stabilization problem.
  • Due to the inaccessibility of all state variables in many actual operations, an observer-based control system is developed to reconstruct the state variables of SSs under asynchronous switching.
  • The asynchronous H problem is investigated and novel LMI-based conditions are presented as a feasibility problem to design the observer-based controller and to compute the prescribed performance index in the under exogenous disturbances.
  • The AS problem is addressed by a simple observer-based method such that the computational complexity is reduced.
  • The average DT technique is developed by the Lyapunov–Krasovskii method, and some stabilization conditions are derived.
  • The robustness against time-delay, AS, and uncertainties is analyzed via LMIs and the singular-value decomposition (SVD) approach.

2. Problem Formulation and Preliminaries

Consider the SSs as (1):
χ ˙ ( t ) = ( A σ ( t ) n + Δ A σ ( t ) ) χ ( t ) + ( B σ ( t ) n + Δ B σ ( t ) ) χ ( t ι τ ) + ( D σ ( t ) n + Δ D σ ( t ) ) u σ ˜ ( t ) ( t ) + σ ( t ) ( t , χ ( t ) ) + ( W σ ( t ) n + Δ W σ ( t ) ) d ˜ ( t ) y ( t ) = C χ ( t ) χ ( s ) = φ ( s ) , s [ ι τ , 0 ]
where χ ( t ) R n χ denotes state vector, u σ ˜ ( t ) ( t ) R n u represents control signal, y ( t ) R n y is output, σ ( t ) ( t , χ ( t ) ) is the nonlinear function, d ˜ ( t ) is the exogenous disturbance which belongs to l 2 [ 0 , ) , ι τ denotes the time-delay, and φ ( s ) is a function which specifies the initial state. The switching signal σ ( t ) : [ t 0 , ) M = { 1 , 2 , , l } is a piecewise function, where l designates the number of system modes. Associated to the function σ ( t ) , the switching sequence σ ( t ) : { ( t 0 , σ ( t 0 ) ) , ( t 1 , σ ( t 1 ) ) , , ( t k , σ ( t k ) ) , | σ ( t k ) M , k = 1 , 2 , , N } is determined, and k denotes the switching number. Ideally, the observers and controllers alter simultaneously with the system modes, which we can write σ ( t ) = σ ˜ ( t ) , but in actual operation, since it takes some times to perceive the active subsystem and apply the matched observer and control signals, the switchings of the observers and control signals σ ˜ ( t ) lag behind system modes, which means that σ ˜ ( t ) : { ( t 0 , σ ( t 0 ) ) , ( t 1 + ω 1 , σ ( t 1 ) ) , , ( t k + ω k , σ ( t k ) ) , | σ ( t k ) M } . For each subsystem, the length of the lag time ϖ k is unknown, but it is assumed that ω k < ϖ , k = 1 , 2 , , N , where ϖ is a known constant.
On the other hand, since all of the states are not measurable in some actual operations, the following switched observer is suggested (1):
χ ^ ˙ ( t ) = A σ ( t ) n χ ^ ( t ) + B σ ( t ) n χ ^ ( t ι τ ) + D σ ( t ) n u σ ˜ ( t ) ( t ) + σ ( t ) ( t , χ ^ ( t ) ) + L σ ˜ ( t ) ( y ( t ) y ^ ( t ) ) y ^ ( t ) = C χ ^ ( t ) χ ^ ( s ) = 0 , s [ ι τ , 0 ]
where χ ^ ( t ) R n χ is observer states, and L σ ˜ ( t ) is the observer gain that has to be designed. Moreover, a controller is updated, i.e., u σ ˜ ( t ) ( t ) = K σ ˜ ( t ) χ ^ ( t ) . To analyze the AS problem, the entire operation time of the system is written as [ t k 1 + ω k 1 , t k ) , k = 1 , 2 , , N , with ω 0 = 0 and mismatched intervals [ t k , t k + ω k ) , k = 1 , 2 , , N . Without loss of generality, assume that the ith subsystem is active at t k 1 , σ ( t k 1 ) = i , and the jth subsystem is active at t k , σ ( t k ) = j , then the corresponding observer and control signals are identified at t k 1 + ω k 1 and t k + ω k . Moreover, the state estimation error is defined as e ( t ) = χ ( t ) χ ^ ( t ) , and the switched observer dynamic (2) can be rewritten as
χ ^ ˙ ( t ) = ( A i n + D i n K i ) χ ^ ( t ) + B i n χ ^ ( t ι τ ) + i ( t , χ ^ ( t ) ) + L i C e ( t ) , t [ t k 1 + ω k 1 , t k ) χ ^ ˙ ( t ) = ( A j n + D j n K i ) χ ^ ( t ) + B j n χ ^ ( t ι τ ) + j ( t , χ ^ ( t ) ) + L i C e ( t ) , t [ t k , t k + ω k )
Additionally, estimation error dynamics are written as:
e ˙ ( t ) = ( A i n L i C + Δ A i ) e ( t ) + ( Δ A i + Δ D i K i ) χ ^ ( t ) + ( B i n + Δ B i ) e ( t ι τ ) + Δ B i χ ^ ( t ι τ ) + i ( t , χ ( t ) ) i ( t , χ ^ ( t ) ) + ( W i n + Δ W i ) d ˜ ( t ) , t [ t k 1 + ω k 1 , t k ) e ˙ ( t ) = ( A j n L i C + Δ A j ) e ( t ) + ( Δ A j + Δ D j K i ) χ ^ ( t ) + ( B j n + Δ B j ) e ( t ι τ ) + Δ B j χ ^ ( t ι τ ) + j ( t , χ ( t ) ) j ( t , χ ^ ( t ) ) + ( W j n + Δ W j ) d ˜ ( t ) , t [ t k , t k + ω k )
From (3) and (4), the augmented switched system under AS is achieved as follows:
ξ ˙ ( t ) = A ¯ i i ξ ( t ) + B ¯ i ξ ( t ι τ ) + ¯ i ( t ) + W ¯ i n d ˜ ( t ) , t [ t k 1 + ω k 1 , t k ) ξ ˙ ( t ) = A ¯ j i ξ ( t ) + B ¯ j ξ ( t ι τ ) + ¯ j ( t ) + W ¯ j n d ˜ ( t ) , t [ t k , t k + ω k )
where ξ ( t ) = [ χ ^ T ( t ) , e T ( t ) ] T , and
A ¯ i i = A i n + D i n K i L i C Δ A i + Δ D i K i A i n L i C + Δ A i A ¯ j i = A j n + D j n K i L i C Δ A j + Δ D j K i A j n L i C + Δ A j B ¯ i = B i n 0 Δ B i B i n + Δ B i , B ¯ j = B j n 0 Δ B j B j n + Δ B j ¯ i ( t ) = i ( t , χ ^ ( t ) ) i ( t , χ ( t ) ) i ( t , χ ^ ( t ) ) , ¯ j ( t ) = j ( t , χ ^ ( t ) ) j ( t , χ ( t ) ) j ( t , χ ^ ( t ) ) W ¯ i ( t ) = 0 W i n + Δ W i , W ¯ j ( t ) = 0 W j n + Δ W j
Note that parametric uncertain matrices Δ A i , Δ B i , Δ D i , and Δ W i i M have the following structure:
Δ i : Δ A i = = 1 q δ θ E i Δ B i = = 1 q δ θ F i Δ D i = = 1 q δ θ G i Δ W i = = 1 q δ θ T i
where δ θ [ δ ˜ θ , δ ˜ θ ] , = 1 , , q are the uncertain parameters. E i , F i , G i and T i are the uncertainty matrices to determine the dependency of Δ A i , Δ B i , Δ D i , Δ W i on δ θ .
Assumption 1.
Consider a full-row rank matrix Ψ n y × n χ ; From this assumption, the SVD of Ψ ; is written as
Ψ n y × n χ = Ξ n y × n y Ψ 0 n y × n y 0 Υ n χ × n χ T
where Ξ / Υ and Ψ 0 are unitary and diagonal matrices, respectively.
Lemma 1
([28]). Consider a full rank matrix Ψ n y × n χ and symmetric matrix X n χ × n χ ; There exists Ω n y × n y such that Ψ X = Ω Ψ , if and only if:
X = Υ X 1 0 0 X 2 Υ T
where dimensions of X 1 , X 2 are as n y × n y and ( n χ n y ) × ( n χ n y ) , respectively. Υ is defined in (7).
Assumption 2.
Consider that, for each i M and any vectors η 1 ( t ) , η 2 ( t ) , the nonlinear vector-valued function i ( . ) is Lipschitz such that
i ( t , η 1 ( t ) ) i ( t , η 2 ( t ) ) π ¯ H i η 1 ( t ) η 2 ( t )
where M is defined in (1), H i ’s are real weighting Lipschitz matrices, and π ¯ is a Lipschitz constant.
Definition 1.
(a) The system (5) is exponentially stable in the presence of switchings σ ( . ) and when d ˜ ( t ) = 0 , if the following condition for ε 1 and λ > 0 is satisfied:
ξ ( t ) ε ξ ( t 0 ) e λ ( t t 0 ) , t t 0 ξ ( t 0 ) = sup ι τ ϕ 0 ξ ( t 0 + ϕ )
(b) If the system (5) satisfies the following condition for d ˜ ( t ) 0 , d ˜ ( t ) l 2 [ 0 , ) , and positive constants λ α , ψ
t 0 e λ α s t 0 y T s y s d s t 0 ψ 2 d ˜ T s d ˜ s d s
Then, (5) has H performance.
Definition 2
([29]). For any scalars k 1 , k 2 , let N σ ( k 1 , k 2 ) denote the switchings σ ( . ) over ( k 1 , k 2 ) . If
N σ ( k 1 , k 2 ) N 0 + k 2 k 1 τ a
holds for τ a > 0 , N 0 0 , then τ a is called the average dwell time.
Lemma 2
([15]). For a scalar ϵ > 0 , any matrices X i R x × y , i = 1 , , q , and positive-definite matrices P R x × x , the following inequality holds:
( X 1 + + X q ) T P ( X 1 + + X q ) ϵ ( X 1 T P X 1 + + X q T P X q )
Lemma 3
([30]). let ϑ , D , F , E , ζ be some vectors or matrices with appropriate dimensions. As a result, the inequality
2 ϑ T D F E ζ ϵ ϑ T D D T ϑ + ϵ 1 ζ T E T E ζ
holds for F T F I and a real positive parameter ϵ.

3. Main Results

In this section, an observer-based control technique is designed for the switched system (5) in the presence of the uncertain parameters δ θ [ δ ˜ θ , δ ˜ θ ] , = 1 , , q . The design methods for the augmented switched system (5) with d ˜ ( t ) = 0 and d ˜ ( t ) 0 are investigated in Theorems 1 and 2, respectively.
Theorem 1.
Consider system (5) under uncertain parameters δ θ [ δ ˜ θ , δ ˜ θ ] . For given positive scalars ϵ 1 , ϵ 2 , ϵ 3 , ϵ 4 , ϵ 5 , ϵ 6 , and parameters ι τ , ϖ , λ α , λ β , π ¯ , μ 1 , suppose that there exist symmetric matrices X χ ^ i > 0 , X e i > 0 , Y χ ^ i > 0 , Y e i > 0 and matrices Z χ ^ i , Z e i , for any i , j M , i j , such that:
X χ ^ j μ X χ ^ i , X e j μ X e i , Y χ ^ j μ Y χ ^ i , Y e j μ Y e i
Ψ i = Ψ i 1 Ψ i 2 Ψ i 3 Ψ i 4 Ψ i 5 Ψ i 6 Ψ i 7 Ψ i 8 Ψ i 9 * I 0 0 0 0 0 0 0 * * I 0 0 0 0 0 0 * * * I 0 0 0 0 0 * * * * I 0 0 0 0 * * * * * I 0 0 0 * * * * * * Y χ ^ i 0 0 * * * * * * * I 0 * * * * * * * * Y e i < 0
Θ i = Θ i 1 Θ i 2 Θ i 3 Θ i 4 Θ i 5 Θ i 6 Θ i 7 Θ i 8 Θ i 9 * I 0 0 0 0 0 0 0 * * I 0 0 0 0 0 0 * * * I 0 0 0 0 0 * * * * I 0 0 0 0 * * * * * I 0 0 0 * * * * * * Y χ ^ i 0 0 * * * * * * * I 0 * * * * * * * * Y e i < 0
where
Ψ i 1 = Ψ i 1 ( 1 , 1 ) B i n Y χ ^ i Z e i C 0 * e λ α ι τ Y χ ^ i 0 0 * * Ψ i 1 ( 3 , 3 ) B i n Y e i * * * e λ α ι τ Y e i Ψ i 3 = 0 0 ϵ ¯ 5 Y χ ^ i ( F ¯ 1 i ) T ϵ ¯ 5 Y χ ^ i ( F ¯ q i ) T 0 0 0 0 Ψ i 2 = ϵ ¯ 3 X χ ^ i ( E ¯ 1 i ) T + Z χ ^ i T ( G ¯ 1 i ) T ϵ ¯ 3 X χ ^ i ( E ¯ q i ) T + Z χ ^ i T ( G ¯ q i ) T 0 0 0 0 0 0 Ψ i 4 = 0 0 0 0 ϵ ¯ 2 X e i ( E ¯ 1 i ) T ϵ ¯ 2 X e i ( E ¯ q i ) T 0 0 , Ψ i 5 = 0 0 0 0 0 0 ϵ ¯ 4 Y e i ( F ¯ 1 i ) T ϵ ¯ 4 Y e i ( F ¯ q i ) T Ψ i 6 = ϵ ¯ 1 γ X χ ^ i H i T 0 0 0 , Ψ i 7 = X χ ^ i 0 0 0 , Ψ i 8 = 0 0 ϵ ¯ 6 γ X e i H i T 0 , Ψ i 9 = 0 0 X e i 0 Ψ i 1 ( 1 , 1 ) = A i n X χ ^ i + X χ ^ i A i n T + D i n Z χ ^ i + Z χ ^ i T D i n T + ϵ 1 1 I + λ α X χ ^ i Ψ i 1 ( 3 , 3 ) = A i n X e i + X e i A i n T Z e i C C T Z e i T + ϵ 2 1 + ϵ 3 1 + ϵ 4 1 + ϵ 5 1 + ϵ 6 1 I + λ α X e i ϵ ¯ 1 = ϵ 1 , ϵ ¯ 2 = ϵ 2 × q , ϵ ¯ 3 = ϵ 3 × q , ϵ ¯ 4 = ϵ 4 × q , ϵ ¯ 5 = ϵ 5 × q , ϵ ¯ 6 = ϵ 6 E ¯ i = δ ˜ θ E i , F ¯ i = δ ˜ θ F i , G ¯ i = δ ˜ θ G i
and
Θ i 1 = Θ i 1 ( 1 , 1 ) B j n Y χ ^ i Z e i C 0 * Y χ ^ i 0 0 * * Θ i 1 ( 3 , 3 ) B j n Y e i * * * Y e i , Θ i 3 = 0 0 ϵ ¯ 5 Y χ ^ i ( F ¯ 1 j ) T ϵ ¯ 5 Y χ ^ i ( F ¯ q j ) T 0 0 0 0 Θ i 2 = ϵ ¯ 3 X χ ^ i ( E ¯ 1 j ) T + Z χ ^ i T ( G ¯ 1 j ) T ϵ ¯ 3 X χ ^ i ( E ¯ q j ) T + Z χ ^ i T ( G ¯ q j ) T 0 0 0 0 0 0
Θ i 4 = 0 0 0 0 ϵ ¯ 2 X e i ( E ¯ 1 j ) T ϵ ¯ 2 X e i ( E ¯ q j ) T 0 0 , Θ i 5 = 0 0 0 0 0 0 ϵ ¯ 4 Y e i ( F ¯ 1 j ) T ϵ ¯ 4 Y e i ( F ¯ q j ) T Θ i 6 = ϵ ¯ 1 γ X χ ^ i H j T 0 0 0 , Θ i 7 = X χ ^ i 0 0 0 , Θ i 8 = 0 0 ϵ ¯ 6 γ X e i H j T 0 , Θ i 9 = 0 0 X e i 0 Θ i 1 ( 1 , 1 ) = A j n X χ ^ i + X χ ^ i A j n T + D j n Z χ ^ i + Z χ ^ i T D j n T + ϵ 1 1 I λ β X χ ^ i Θ i 1 ( 3 , 3 ) = A j n X e i + X e i A j n T Z e i C C T Z e i T + ϵ 2 1 + ϵ 3 1 + ϵ 4 1 + ϵ 5 1 + ϵ 6 1 I λ β X e i
Then, system (5) is exponentially stable with switchings that satisfy τ a > τ a * = ln μ + λ α + λ β ι τ + ϖ λ α . Moreover, the gains K i , i M and L i , i M are achieved as follows:
K i = Z χ ^ i X χ ^ i 1
L i = Z e i U C 0 X e 1 i 1 C 0 1 U T
Proof. 
From (5), when at time t [ t k 1 + ω k 1 , t k ) , the ith subsystem is activated and the corresponding switching controller K i and observer L i are identified. During t [ t k 1 + ω k 1 , t k ) , consider:
V 1 i ( t ) = ξ T ( t ) P i ξ ( t ) + t ι τ t e λ α t s ξ T s Q i ξ ( s ) d s
where ξ ( t ) = [ χ ^ T ( t ) , e T ( t ) ] T , P i = diag { P χ ^ i , P e i } , and Q i = diag { Q χ ^ i , Q e i } . The (20) can be rewritten as
V 1 i ( t ) = χ ^ T t P χ ^ i χ ^ t + e T t P e i e t + t ι τ t e λ α t s χ ^ T s Q χ ^ i χ ^ s d s + t ι τ t e λ α t s e T s Q e i e s d s
Considering (3), the derivative of V 1 i leads to
V ˙ 1 i ( t ) = 2 χ ^ T ( t ) P χ ^ i χ ^ ˙ ( t ) + 2 e T ( t ) P e i e ˙ ( t ) + χ ^ T ( t ) Q χ ^ i χ ^ ( t ) e λ α ι τ χ ^ T ( t ι τ ) Q χ ^ i χ ^ ( t ι τ ) λ α t ι τ t e λ α ( t s ) χ ^ T ( s ) Q χ ^ i χ ^ ( s ) d s + e T ( t ) Q e i e ( t ) e λ α ι τ e T ( t ι τ ) Q e i e ( t ι τ ) λ α t ι τ t e λ α ( t s ) e T ( s ) Q e i e ( s ) d s
According to (3), it can be concluded that
2 χ ^ T ( t ) P χ ^ i χ ^ ˙ ( t ) = χ ^ T ( t ) P χ ^ i A i n + D i n K i + A i n + D i n K i T P χ ^ i χ ^ ( t ) + 2 χ ^ T ( t ) P χ ^ i B i n χ ^ ( t ι τ ) + 2 χ ^ T ( t ) P χ ^ i L i C e ( t ) + 2 χ ^ T ( t ) P χ ^ i i ( t , χ ^ ( t ) )
From Lemma 3, one can attain
2 χ ^ T ( t ) P χ ^ i χ ^ ˙ ( t ) χ ^ T ( t ) P χ ^ i A i n + D i n K i + A i n + D i n K i T P χ ^ i χ ^ ( t ) + 2 χ ^ T ( t ) P χ ^ i B i n χ ^ ( t ι τ ) + ϵ 1 1 χ ^ T ( t ) P χ ^ i P χ ^ i χ ^ ( t ) + ϵ 1 i ( t , χ ^ ( t ) ) T i ( t , χ ^ ( t ) )
Considering the estimation error dynamic (4) and parametric uncertain matrices (6), one has
2 e T ( t ) P e i e ˙ ( t ) = e T ( t ) P e i A i n L i C + A i n L i C T P e i e ( t ) + 2 e T ( t ) P e i = 1 q δ θ E i e ( t ) + 2 e T ( t ) P e i = 1 q δ θ E i + δ θ G i K i χ ^ ( t ) + 2 e T ( t ) P e i B i n e ( t ι τ ) + 2 e T ( t ) P e i = 1 q δ θ F i e ( t ι τ ) + 2 e T ( t ) P e i = 1 q δ θ F i χ ^ ( t ι τ ) + 2 e T ( t ) P e i i ( t , χ ( t ) ) i ( t , χ ^ ( t ) )
Regarding Lemmas 2 and 3, and considering δ θ [ δ ˜ θ , δ ˜ θ ] , = 1 , , q , it is derived that
2 e T ( t ) P e i e ˙ ( t ) e T ( t ) P e i A i n L i C + A i n L i C T P e i e ( t ) + 2 e T ( t ) P e i B i n e ( t ι τ ) + ϵ 2 1 e T ( t ) P e i P e i e ( t ) + ( ϵ 2 × q ) e T ( t ) = 1 q δ ˜ θ 2 ( E i ) T E i e ( t ) + ϵ 3 1 e T ( t ) P e i P e i e ( t ) + ( ϵ 3 × q ) χ ^ T ( t ) = 1 q δ ˜ θ 2 ( E i + G i K i ) T ( E i + G i K i ) χ ^ ( t ) + ϵ 4 1 e T ( t ) P e i P e i e ( t ) + ( ϵ 4 × q ) e T ( t ι τ ) = 1 q δ ˜ θ 2 ( F i ) T F i e ( t ι τ ) + ϵ 5 1 e T ( t ) P e i P e i e ( t ) + ( ϵ 5 × q ) χ ^ T ( t ι τ ) = 1 q δ ˜ θ 2 ( F i ) T F i χ ^ ( t ι τ ) + ϵ 6 1 e T ( t ) P e i P e i e ( t ) + ϵ 6 i ( t , χ ( t ) ) i ( t , χ ^ ( t ) ) T i ( t , χ ( t ) ) i ( t , χ ^ ( t ) )
Now, from (22), (24), (26), considering Assumption 2, and defining E ¯ i = δ ˜ θ E i , F ¯ i = δ ˜ θ F i , G ¯ i = δ ˜ θ G i , we have:
V ˙ 1 i ( t ) + λ α V 1 i ( t ) ζ T ( t ) Π i ζ ( t )
where ζ ( t ) = [ χ ^ T ( t ) , χ ^ T ( t ι τ ) , e T ( t ) , e T ( t ι τ ) ] T , and
Π i = Π i ( 1 , 1 ) P χ ^ i B i n P χ ^ i L i C 0 * Π i ( 2 , 2 ) 0 0 * * Π i ( 3 , 3 ) P e i B i n * * * Π i ( 4 , 4 ) < 0
with
Π i ( 1 , 1 ) = P χ ^ i A i n + D i n K i + A i n + D i n K i T P χ ^ i + ϵ 1 1 P χ ^ i P χ ^ i + ϵ 1 π ¯ 2 H i T H i + Q χ ^ i + λ α P χ ^ i + ( ϵ 3 × q ) = 1 q E ¯ i + G ¯ i K i T E ¯ i + G ¯ i K i Π i ( 2 , 2 ) = e λ α ι τ Q χ ^ i + ( ϵ 5 × q ) = 1 q ( F ¯ i ) T F ¯ i
Π i ( 3 , 3 ) = P e i A i n L i C + A i n L i C T P e i ϵ 2 1 + ϵ 3 1 + ϵ 4 1 + ϵ 5 1 + ϵ 6 1 P e i P e i + ϵ 6 π ¯ 2 H i T H i + Q e i + λ α P e i + ( ϵ 2 × q ) = 1 q ( E ¯ i ) T E ¯ i Π i ( 4 , 4 ) = e λ α ι τ Q e i + ( ϵ 4 × q ) = 1 q ( F ¯ i ) T F ¯ i
Using the Schur lemma, (28) is written as:
Γ i = Γ i 1 Γ i 2 Γ i 3 Γ i 4 Γ i 5 * I 0 0 0 * * I 0 0 * * * I 0 * * * * I < 0
in which
Γ i 1 = Γ i 1 ( 1 , 1 ) P χ ^ i B i n P χ ^ i L i C 0 * e λ α ι τ Q χ ^ i 0 0 * * Γ i 1 ( 3 , 3 ) P e i B i n * * * e λ α ι τ Q e i Γ i 2 = ϵ ¯ 3 E ¯ 1 i + G ¯ 1 i K i T ϵ ¯ 3 E ¯ q i + G ¯ q i K i T 0 0 0 0 0 0 Γ i 3 = 0 0 ϵ ¯ 5 ( F ¯ 1 i ) T ϵ ¯ 5 ( F ¯ q i ) T 0 0 0 0 , Γ i 4 = 0 0 0 0 ϵ ¯ 2 ( E ¯ 1 i ) T ϵ ¯ 2 ( E ¯ q i ) T 0 0 Γ i 5 = 0 0 0 0 0 0 ϵ ¯ 4 ( F ¯ 1 i ) T ϵ ¯ 4 ( F ¯ q i ) T ϵ ¯ 2 = ϵ 2 × q , ϵ ¯ 3 = ϵ 3 × q , ϵ ¯ 4 = ϵ 4 × q , ϵ ¯ 5 = ϵ 5 × q Γ i 1 ( 1 , 1 ) = P χ ^ i A i n + D i n K i + A i n + D i n K i T P χ ^ i + ϵ 1 1 P χ ^ i P χ ^ i + ϵ 1 π ¯ 2 H i T H i + Q χ ^ i + λ α P χ ^ i Γ i 1 ( 3 , 3 ) = P e i A i n L i C + A i n L i C T P e i ϵ 2 1 + ϵ 3 1 + ϵ 4 1 + ϵ 5 1 + ϵ 6 1 P e i P e i + ϵ 6 π ¯ 2 H i T H i + Q e i + λ α P e i
From Lemma 1, if we could find matrix V such that
X e i = V X e 1 i 0 0 X e 2 i V T
could be established, the condition C X e i = R i C holds. Now, let P χ ^ i 1 = X χ ^ i , P e i 1 = X e i , Q χ ^ i 1 = Y χ ^ i , Q e i 1 = Y e i , K i X χ ^ i = Z χ ^ i , L i R i = Z e i , applying the congruent transformation
diag X χ ^ i , Y x ^ i , X e i , Y e i , I , , I r , I , , I r , I , , I r , I , , I r
to the LMI (29), results in the LMI (16). Furthermore, based on Assumption 1, Equation (30), and the condition C X e i = R i C , the matrices R i can be computed as follows:
U C 0 0 V T V X e 1 i 0 0 X e 2 i V T = R i U C 0 0 V T U C 0 X e 1 i = R i U C 0 R i = U C 0 X e 1 i ( U C 0 ) 1
Therefore, observer gains can be computed through L i = Z e i U C 0 X e 1 i 1 C 0 1 U T . Furthermore, according to K i X χ ^ i = Z χ ^ i , it can be deduced that K i = Z χ ^ i X χ ^ i 1 . As a result, when t [ t k 1 + ω k 1 , t k ) , one can achieve that
V ˙ 1 i ( t ) + λ α V 1 i ( t ) ζ T ( t ) Ψ i ζ ( t )
From LMI (16), V ˙ 1 i ( t ) + λ α V 1 i ( t ) 0 holds, and it means that, when t [ t k 1 + ω k 1 , t k ) , one has
V 1 i ( t ) V 1 i ( t k 1 + ω k 1 ) e λ α ( t t k 1 ω k 1 )
Furthermore, according to (5), it is obvious that, at time t [ t k , t k + ω k ) , the jth subsystem is activated, but the control signal K i and the observer L i are not varied. Therefore, consider:
V 2 i t = ξ T ( t ) P i ξ ( t ) + t ι τ t e λ β t s ξ T s Q i ξ ( s ) d s = χ ^ T t P χ ^ i χ ^ t + e T t P e i e t + t ι τ t e λ β t s χ ^ T s Q χ ^ i χ ^ s d s + t ι τ t e λ β t s e T s Q e i e s d s
Similarly, we can write:
V ˙ 2 i ( t ) λ β V 2 i ( t ) ζ T ( t ) Θ i ζ ( t )
Moreover, LMI (17) implies that V ˙ 2 i ( t ) λ β V 2 i ( t ) 0 . Integrating both sides of this inequality, one can acquire that
V 2 i ( t ) V 2 i ( t k ) e λ β ( t t k )
On the other hand, one can obtain the following inequalities:
t ι τ t e λ β t s ξ T s Q i ξ ( s ) d s e λ β ι τ t ι τ t ξ T s Q i ξ ( s ) d s e ( λ α + λ β ) ι τ t ι τ t e λ α t s ξ T s Q i ξ ( s ) d s
From (15), (33), (36), (37), and similar to [31], one can derive the following results:
V ( t ) [ e ( λ α + λ β ) ι τ ] N σ ( t k , t ) μ N σ ˜ ( t k , t ) V ( t k ) e λ β T + ( t k , t ) λ α T ( t k , t ) [ e ( λ α + λ β ) ι τ ] N σ ( t k 1 , t ) μ N σ ˜ ( t k 1 , t ) V ( t k 1 ) e λ β T + ( t k 1 , t ) λ α T ( t k 1 , t ) [ e ( λ α + λ β ) ι τ ] N σ ( t 0 , t ) μ N σ ˜ ( t 0 , t ) V ( t 0 ) e λ β T + ( t 0 , t ) λ α T ( t 0 , t ) [ e ( λ α + λ β ) ι τ ] N σ ( t 0 , t ) μ N σ ˜ ( t 0 , t ) V ( t 0 ) e ( λ α + λ β ) T + ( t 0 , t ) λ α ( t t 0 ) [ e ( λ α + λ β ) ι τ ] N σ ( t 0 , t ) μ N σ ˜ ( t 0 , t ) V ( t 0 ) e ϖ ( λ α + λ β ) N σ ( t 0 , t ) λ α ( t t 0 )
where T ( t 0 , t ) indicates the total matched intervals and T + ( t 0 , t ) stands for ( t 0 , t ) , and N σ ( t 0 , t ) is written as:
N σ ( t 0 , t ) = N σ ˜ ( t 0 , t ) , t [ t k 1 + ω k 1 , t k ) , k = 1 , 2 , , N . N σ ( t 0 , t ) = N σ ˜ ( t 0 , t ) + 1 , t [ t k , t k + ω k ) , k = 1 , 2 , , N .
Furthermore, from Definition 2 and (39), we have:
V ( t ) μ e λ α + λ β ι τ t t 0 τ a V ( t 0 ) e ϖ λ α + λ β t t 0 τ a λ α ( t t 0 ) = V ( t 0 ) e λ α ln μ e ( λ α + λ β ) ι τ + ϖ λ α + λ β τ a ( t t 0 )
In addition, from (20) and (34), one can further achieve that
V ( t ) κ 1 ξ ( t ) 2 , V ( t 0 ) κ 2 ξ ( t 0 ) 2 .
Then, the solution of (5) exists globally and is satisfied as follows:
ξ ( t ) 2 κ 2 κ 1 ξ ( t 0 ) 2 e λ α ln μ e ( λ α + λ β ) ι τ + ϖ λ α + λ β τ a ( t t 0 )
where
κ 1 = min i M λ min ( P i ) κ 2 = max i M λ max ( P i ) + ι τ max i M λ max ( Q i )
Remark 1.
It is worth nothing that the results of Theorem 1 can be extended to investigate the asynchronous H observer-based control problem and the prescribed performance index for the system (5) subject to the parametric uncertainties and the exogenous disturbance d ˜ ( t ) can be obtained using the same method in [22] and Definition 1. In this regard, the following results can be achieved:
V ˙ 1 i ( t ) + λ α V 1 i ( t ) y T ( t ) y ( t ) + ϑ 2 d ˜ T ( t ) d ˜ ( t ) , t [ t k 1 + ω k 1 , t k ) V ˙ 2 i ( t ) λ β V 2 i ( t ) y T ( t ) y ( t ) + ϑ 2 d ˜ T ( t ) d ˜ ( t ) , t [ t k , t k + ω k )
Now, defining Υ ( t ) = y T ( t ) y ( t ) ϑ 2 d ˜ T ( t ) d ˜ ( t ) and integrating from t k 1 + ω k 1 to t and t k to t lead to the following inequalities:
V 1 i ( t ) V 1 i ( t k 1 + ω k 1 ) e λ α t t k 1 ω k 1 t k 1 + ω k 1 t Υ ( s ) e λ α ( t s ) d s , t [ t k 1 + ω k 1 , t k ) V 2 i ( t ) V 2 i ( t k ) e λ β t t k t k t Υ ( s ) e λ β ( t s ) d s , t [ t k , t k + ω k )
Considering (38) and (46), the following results can be achieved:
V ( t ) [ e ( λ α + λ β ) ι τ ] N σ ( t k , t ) μ N σ ˜ ( t k , t ) V ( t k ) e λ β T + ( t k , t ) λ α T ( t k , t ) t k t μ N σ ˜ ( s , t ) [ e ( λ α + λ β ) ι τ ] N σ ( s , t ) Υ ( s ) e λ β T + ( s , t ) λ α T ( s , t ) d s [ e ( λ α + λ β ) ι τ ] N σ ( t k 1 , t ) μ N σ ˜ ( t k 1 , t ) V ( t k 1 ) e λ β T + ( t k 1 , t ) λ α T ( t k 1 , t ) t k 1 t μ N σ ˜ ( s , t ) [ e ( λ α + λ β ) ι τ ] N σ ( s , t ) Υ ( s ) e λ β T + ( s , t ) λ α T ( s , t ) d s [ e ( λ α + λ β ) ι τ ] N σ ( t 0 , t ) μ N σ ˜ ( t 0 , t ) V ( t 0 ) e λ β T + ( t 0 , t ) λ α T ( t 0 , t ) t 0 t μ N σ ˜ ( s , t ) [ e ( λ α + λ β ) ι τ ] N σ ( s , t ) Υ ( s ) e λ β T + ( s , t ) λ α T ( s , t ) d s = V ( t 0 ) e λ β T + ( t 0 , t ) λ α T ( t 0 , t ) + N σ ˜ ( t 0 , t ) ln μ + N σ ( t 0 , t ) ln [ e ( λ α + λ β ) ι τ ] t 0 t Υ ( s ) e λ β T + ( s , t ) λ α T ( s , t ) + N σ ˜ ( s , t ) ln μ + N σ ( s , t ) ln [ e ( λ α + λ β ) ι τ ] d s
Furthermore, utilizing some simplifications, for all d ˜ ( t ) l 2 [ 0 , ) , one can get
t 0 e λ α s t 0 y T s y s d s t 0 ψ 2 d ˜ T s d ˜ s d s
in which the prescribed performance index ψ is computed as
ψ = μ ϑ 2 1 λ α + λ β ϖ ln μ + ι τ + ϖ λ α + λ β
Furthermore, ψ can be minimized via searching the optimal value for the scalar ϑ in the elicited stabilization conditions. The following Theorem gives the sufficient conditions to design the observer-based controller and the prescribed performance index ψ for the system (5).
Theorem 2.
Consider system (5) under uncertain parameters δ θ [ δ ˜ θ , δ ˜ θ ] . For given positive scalars ϵ 1 , ϵ 2 , ϵ 3 , ϵ 4 , ϵ 5 , ϵ 6 , ϵ 7 , and parameters ι τ , ϖ , λ α , λ β , π ¯ , ϑ , μ 1 , suppose that there exist symmetric matrices X χ ^ i > 0 , X e i > 0 , Y χ ^ i > 0 , Y e i > 0 and matrices Z χ ^ i , Z e i , for any i , j M , i j , such that:
X χ ^ j μ X χ ^ i , X e j μ X e i , Y χ ^ j μ Y χ ^ i , Y e j μ Y e i
Ψ i = Ψ i 1 Ψ i 2 Ψ i 3 Ψ i 4 Ψ i 5 Ψ i 6 Ψ i 7 Ψ i 8 Ψ i 9 Ψ i 11 * I 0 0 0 0 0 0 0 0 * * I 0 0 0 0 0 0 0 * * * I 0 0 0 0 0 0 * * * * I 0 0 0 0 0 * * * * * I 0 0 0 0 * * * * * * I 0 0 0 * * * * * * * I 0 0 * * * * * * * * Ψ i 10 0 * * * * * * * * * Ψ i 12 < 0
Θ i = Θ i 1 Θ i 2 Θ i 3 Θ i 4 Θ i 5 Θ i 6 Θ i 7 Θ i 8 Θ i 9 Θ i 11 * I 0 0 0 0 0 0 0 0 * * I 0 0 0 0 0 0 0 * * * I 0 0 0 0 0 0 * * * * I 0 0 0 0 0 * * * * * I 0 0 0 0 * * * * * * I 0 0 0 * * * * * * * I 0 0 * * * * * * * * Θ i 10 0 * * * * * * * * * Θ i 12 < 0
where
Ψ i 1 = Ψ i 1 ( 1 , 1 ) B i n Y χ ^ i Z e i C 0 0 * e λ α ι τ Y χ ^ i 0 0 0 * * Ψ i 1 ( 3 , 3 ) B i n Y e i W i n * * * e λ α ι τ Y e i 0 * * * * ϑ 2 I Ψ i 2 = ϵ ¯ 3 X χ ^ i ( E ¯ 1 i ) T + Z χ ^ i T ( G ¯ 1 i ) T ϵ ¯ 3 X χ ^ i ( E ¯ q i ) T + Z χ ^ i T ( G ¯ q i ) T 0 0 0 0 0 0 0 0 Ψ i 3 = 0 0 ϵ ¯ 5 Y χ ^ i ( F ¯ 1 i ) T ϵ ¯ 5 Y χ ^ i ( F ¯ q i ) T 0 0 0 0 0 0 , Ψ i 4 = 0 0 0 0 ϵ ¯ 2 X e i ( E ¯ 1 i ) T ϵ ¯ 2 X e i ( E ¯ q i ) T 0 0 0 0 Ψ i 5 = 0 0 0 0 0 0 ϵ ¯ 4 Y e i ( F ¯ 1 i ) T ϵ ¯ 4 Y e i ( F ¯ q i ) T 0 0 , Ψ i 6 = 0 0 0 0 0 0 0 0 ϵ ¯ 7 T ¯ 1 i T ϵ ¯ 7 T ¯ q i T Ψ i 7 = ϵ ¯ 1 γ X χ ^ i H i T 0 0 0 0 , Ψ i 8 = 0 0 ϵ ¯ 6 γ X e i H i T 0 0 , Ψ i 9 = X χ ^ i 0 0 0 0 X e i 0 0 0 0
Ψ i 10 = Y χ ^ i 0 * Y e i , Ψ i 11 = 3 X χ ^ i C T 0 0 0 0 3 X e i C T 0 0 0 0 , Ψ i 12 = I 0 * I Ψ i 1 ( 1 , 1 ) = A i n X χ ^ i + X χ ^ i A i n T + D i n Z χ ^ i + Z χ ^ i T D i n T + ϵ 1 1 I + λ α X χ ^ i Ψ i 1 ( 3 , 3 ) = A i n X e i + X e i A i n T Z e i C C T Z e i T + ϵ 2 1 + ϵ 3 1 + ϵ 4 1 + ϵ 5 1 + ϵ 6 1 + ϵ 7 1 I + λ α X e i ϵ ¯ 1 = ϵ 1 , ϵ ¯ 2 = ϵ 2 × q , ϵ ¯ 3 = ϵ 3 × q , ϵ ¯ 4 = ϵ 4 × q , ϵ ¯ 5 = ϵ 5 × q , ϵ ¯ 6 = ϵ 6 , ϵ ¯ 7 = ϵ 7 × q , E ¯ i = δ ˜ θ E i , F ¯ i = δ ˜ θ F i , G ¯ i = δ ˜ θ G i , T ¯ i = δ ˜ θ T i
and
Θ i 1 = Θ i 1 ( 1 , 1 ) B j n Y χ ^ i Z e i C 0 0 * Y χ ^ i 0 0 0 * * Θ i 1 ( 3 , 3 ) B j n Y e i W j n * * * Y e i 0 * * * * ϑ 2 I Θ i 2 = ϵ ¯ 3 X χ ^ i ( E ¯ 1 j ) T + Z χ ^ i T ( G ¯ 1 j ) T ϵ ¯ 3 X χ ^ i ( E ¯ q j ) T + Z χ ^ i T ( G ¯ q j ) T 0 0 0 0 0 0 0 0 Θ i 3 = 0 0 ϵ ¯ 5 Y χ ^ i ( F ¯ 1 j ) T ϵ ¯ 5 Y χ ^ i ( F ¯ q j ) T 0 0 0 0 0 0 , Θ i 4 = 0 0 0 0 ϵ ¯ 2 X e i ( E ¯ 1 j ) T ϵ ¯ 2 X e i ( E ¯ q j ) T 0 0 0 0 Θ i 5 = 0 0 0 0 0 0 ϵ ¯ 4 Y e i ( F ¯ 1 j ) T ϵ ¯ 4 Y e i ( F ¯ q j ) T 0 0 , Θ i 6 = 0 0 0 0 0 0 0 0 ϵ ¯ 7 T ¯ 1 j T ϵ ¯ 7 T ¯ q j T Θ i 7 = ϵ ¯ 1 γ X χ ^ i H j T 0 0 0 0 , Θ i 8 = 0 0 ϵ ¯ 6 γ X e i H j T 0 0 , Θ i 9 = X χ ^ i 0 0 0 0 X e i 0 0 0 0 Θ i 10 = Y χ ^ i 0 * Y e i , Θ i 11 = 3 X χ ^ i C T 0 0 0 0 3 X e i C T 0 0 0 0 , Θ i 12 = I 0 * I Θ i 1 ( 1 , 1 ) = A j n X χ ^ i + X χ ^ i A j n T + D j n Z χ ^ i + Z χ ^ i T D j n T + ϵ 1 1 I λ β X χ ^ i Θ i 1 ( 3 , 3 ) = A j n X e i + X e i A j n T Z e i C C T Z e i T + ϵ 2 1 + ϵ 3 1 + ϵ 4 1 + ϵ 5 1 + ϵ 6 1 + ϵ 7 1 I λ β X e i
Then, system (5) is exponentially stable under switchings τ a > τ a * = ln μ + λ α + λ β ι τ + ϖ λ α . Moreover, the controller gains, the observer gains, and the performance index ψ can be designed via (18), (19), and (47), respectively.
Proof. 
Considering the conditions (43) and utilizing the same method in Theorem 1, LMIs (48)–(50) can be achieved. □
Remark 2.
In this paper, the robustness against time-delay, AS, and uncertainties are studied. For future studies, the designed scheme can be can be extended by employing the concept of impulsive stabilization [32,33,34].
Remark 3.
Although static output feedback control method has been presented in [31] to control SSs, the switched observer is designed in this paper to estimate the states of the system. The proposed switched observer-based controller enables us to reconstruct the system states and steer them to zero in the presence of the AS problem and uncertain parameters of the system. Note that AS among system/observer modes is a challenging problem investigated in this paper. In this regard, the gains of the observer have switched asynchronously with the modes of the system; therefore, the suggested switched observer can tackle the lag between the switching instants of the system/observer. Furthermore, unlike the results of [31], the effects of external disturbances on the system are studied in this paper. For this purpose, a prescribed H performance level is considered, and a novel set of LMI-based conditions is achieved based on the multiple Lyapunov–Krasovskii functionals and an ADT approach. Therefore, the robustness of the system is guaranteed in the presence of external disturbances.

4. Simulation Results

The numerical/practical examples are given to evaluate theoretical accomplishments in Theorems 1 and 2. In particular, the mass-springer plant with a switching dynamic [23,31] and the F 18 aircraft system [31,35] are provided to examine the designed controller under AS problem, affine parametric uncertainty, time-delay, and exogenous disturbances.
Example 1.
Considering the SS of the form (1) subject to the parametric uncertain matrices (6) and time-delay AS problem, the matrices are
A 1 = 1.2 + δ θ 1 0.4 1 1 + δ θ 1 , A 2 = 1.6 + δ θ 1 0.1 0.6 0.5 + δ θ 1 B 1 = 0.2 + δ θ 1 0.1 0.1 0.1 + δ θ 1 , B 2 = 0.2 + δ θ 1 0.1 0.1 0.1 + δ θ 1 D 1 = 0.3 + δ θ 1 0.6 0.5 0.9 + δ θ 1 , D 2 = 0.1 + δ θ 1 1.1 0.8 0.3 + δ θ 1 W 1 = 0.2 + δ θ 1 0.1 0.1 0.2 + δ θ 1 , W 2 = 0.1 + δ θ 1 0.2 0.2 0.2 + δ θ 1 1 ( t , χ ( t ) ) = 0.3 χ 1 ( t ) sin ( χ 2 ( t ) ) 0.3 χ 1 ( t ) cos ( χ 2 ( t ) ) , 2 ( t , χ ( t ) ) = 0.5 χ 2 ( t ) cos χ 1 ( t ) 0.5 χ 2 ( t ) sin ( χ 1 ( t ) ) d ˜ ( t ) = 2 sin ( 4 π t ) 1 1 + t , C = 0.4 0.6 0.6 0.4
The uncertainty of the system belongs to the interval as [ 0.1 , 0.1 ] . Then,
A 1 n = 1.2 0.4 1 1 , A 2 n = 1.6 1.1 0.6 0.5 B 1 n = 0.2 0.1 0.1 0.1 , B 2 n = 0.2 0.1 0.1 0.1 D 1 n = 0.3 0.6 0.5 0.9 , D 2 n = 0.1 1.1 0.8 0.3 W 1 n = 0.2 0.1 0.1 0.2 , W 2 n = 0.1 0.2 0.2 0.2
Furthermore, the uncertain matrices are E 1 1 = E 1 2 = F 1 1 = F 1 2 = G 1 1 = G 1 2 = T 1 1 = T 1 2 = I 2 × 2 . In addition, the time-delay and the upper-bound of lag time are selected as ι τ = 1 ( s ) and ϖ = 500 ( ms ) , respectively. Let λ α = 0.3 , λ β = 0.2 , ε 1 : ε 7 = 1 , and μ = 1.1 , then, the ADT is computed via τ a > τ a * = ln μ + λ α + λ β ι τ + ϖ λ α = 2.8177 , and, from (47) the given value for the prescribed performance index is ψ = 0.25 . In the case of the observer-based controller, using LMI Toolbox of MATLAB to solve the LMIs (48)–(50), the controller and observer gains are achieved as follows:
K 1 = 2.0316 7.0247 15.5929 5.4244 , K 2 = 0.7716 7.2377 14.7551 4.7179
L 1 = 8.3063 28.7817 7.3309 7.3658 , L 2 = 8.1690 28.9192 7.6978 6.9693
Figure 1 displays the switchings of the simulations. The time responses of the first and the second state variables with their estimations are represented in Figure 2 and Figure 3, respectively. Figure 4 shows the time history of the estimation error while the system’s output is demonstrated in Figure 5. Furthermore, the time response of control law is depicted in Figure 6. It can be viewed from simulations that utilizing the suggested observer-based control method leads to the stable estimates and states of the system with robust performance against the time-delay, parametric uncertainty, and exogenous disturbances. Moreover, variation of the designed control signal is acceptable.
Example 2.
Consider the mass-springer mechanical system (see [31] for more detail) with the following parameters:
A 1 = 1 1 1 2 , A 2 = 1 1 2 δ θ 1 3 + δ θ 1 B 1 = 0 0 δ θ 1 1 + δ θ 1 , B 2 = 0 0 1 δ θ 1 2 + δ θ 1 D 1 = 0 1 + δ θ 1 , D 2 = 0 1 + δ θ 1 W 1 = 0.2 + δ θ 1 0.1 0.1 0.2 + δ θ 1 , W 2 = 0.1 + δ θ 1 0.2 0.2 0.2 + δ θ 1 1 ( t , χ ( t ) ) = 0.1 sin ( χ 1 ( t ) ) 0.1 cos ( χ 2 ( t ) ) , 2 ( t , χ ( t ) ) = 0.1 sin ( χ 2 ( t ) ) 0.1 cos ( χ 1 ( t ) ) d ˜ ( t ) = 0.7 sin ( 6 π t ) 0.9 sin ( 2 π t ) , C = 0.1 0.1 0.1 0.2
The nominal matrices are
A 1 n = 1 1 1 2 , A 2 n = 1 1 2 3 B 1 n = 0 0 0 1 , B 2 n = 0 0 1 2 D 1 n = 0 1 , D 2 n = 0 1 W 1 n = 0.1 0.2 0.2 0.1 , W 2 n = 0.1 0.1 0 0.1
Accordingly, we have:
E 1 1 = 0 0 0 0 , E 1 2 = 0 0 1 1 F 1 1 = 0 0 1 1 , F 1 2 = 0 0 1 1 G 1 1 = 0 1 , G 1 2 = 0 1 T 1 1 = 1 0 0 1 , T 1 2 = 1 0 0 1
Let ι τ = 2 ( s ) and ϖ = 1 ( s ) . For simulations, choose λ α = 0.4 , λ β = 0.2 , ε 1 : ε 7 = 1 , and μ = 1.1 . As a result, the ADT is τ a > τ a * = ln μ + λ α + λ β ι τ + ϖ α = 4.7383 and the given value for the prescribed index is ψ = 0.2537 . Furthermore, by the use of LMI Toolbox of MATLAB to solve (48)–(50), the observer and controller parameters are acquired as follows:
K 1 = 2.1619 25.7037 , K 2 = 2.8005 37.8457
L 1 = 17.2409 4.5831 214.1548 196.4140 , L 2 = 27.5308 13.8381 334.9659 301.0917
Figure 7 depicts the switchings. The time responses and their approximations are illustrated in Figure 8 and Figure 9. The estimation error is demonstrated in Figure 10 while the system’s output is demonstrated in Figure 11. From the simulations, it is clear that the suggested observer-based control system can ensure the robustness under the parametric uncertainty, time-delay, exogenous disturbance, and the AS problem. Therefore, the designed controller, which is displayed in Figure 12, is persuasive.
Example 3.
The designed observer-based control scheme is examined to an F 18 aircraft (see [31] for more details) with the dynamic model of the form (3) and the parameters as follows:
A 1 = A long m 5 h 40 = 0.2423 + 0.1 δ θ 1 0.9964 + 0.5 δ θ 1 2.342 + δ θ 1 0.1737 + 0.1 δ θ 1 A 2 = A long m 6 h 30 = 0.0416 + 0.05 δ θ 1 0.01141 + 0.05 δ θ 1 2.595 + δ θ 1 0.8161 + 0.5 δ θ 1 B 1 = B long m 5 h 40 = 0.161 + 0.1 δ θ 1 0.387 + 0.1 δ θ 1 1.144 + δ θ 1 0.06 + 0.05 δ θ 1 B 2 = B long m 6 h 30 = 0.017 + 0.05 δ θ 1 0.001 + 0.05 δ θ 1 1.817 + δ θ 1 0.336 + 0.1 δ θ 1 D 1 = D long m 5 h 40 = 0.2423 + 0.1 δ θ 1 0.4978 + 0.1 δ θ 1 1.8420 + 0.1 δ θ 1 0.0877 D 2 = D long m 6 h 30 = 0.5088 + 0.1 δ θ 1 0.0107 + 0.05 δ θ 1 0.1310 + 0.1 δ θ 1 0.6219 + 0.1 δ θ 1 W 1 = 0.2 + 0.2 δ θ 1 0.1 0.1 0.2 + 0.2 δ θ 1 , W 2 = 0.2 + 0.2 δ θ 1 0.1 0.1 0.2 + 0.2 δ θ 1 1 ( t , χ ( t ) ) = 0.1 sin ( χ 2 ( t ) ) 0.1 sin ( χ 1 ( t ) ) , 2 ( t , χ ( t ) ) = 0.1 sin ( χ 1 ( t ) ) 0.1 sin ( χ 2 ( t ) ) d ˜ ( t ) = e ( 0.1 t ) sin ( 0.6 π t ) sin ( 0.6 π t ) , C = 0.5 0.5 0.5 0.6
where the uncertain parameter is δ θ 1 = 0.6 sin ( t ) , and one can obtain
A 1 n = 0.2423 0.9964 2.342 0.1737 , A 2 n = 0.0416 0.01141 2.595 0.8161 B 1 n = 0.161 0.387 1.144 0.06 , B 2 n = 0.017 0.001 1.817 0.336 D 1 n = 0.2423 0.4978 1.8420 0.0877 , D 2 n = 0.5088 0.0107 0.1310 0.6219 W 1 n = 0.2 0.1 0.1 0.2 , W 2 n = 0.2 0.1 0.1 0.2
The uncertainty matrices are
E 1 1 = 0.1 0.5 1 0.1 , E 1 2 = 0.05 0.05 1 0.5 F 1 1 = 0.1 0.1 1 0.05 , F 1 2 = 0.05 0.05 1 0.1 G 1 1 = 0.1 0.1 0.1 0 , G 1 2 = 0.1 0.05 0.1 0.1 T 1 1 = 0.2 0 0 0.2 , T 1 2 = 0.2 0 0 0.2
Furthermore, it is presumed that ι τ = 2 ( s ) and ϖ = 500 ( ms ) . With λ α = 0.4 , λ β = 0.2 , ε 1 : ε 7 = 1 , and μ = 1.1 , the ADT is 3.9883 and the prescribed performance index is ψ = 0.2328 . Furthermore, solving (48)–(50) yields:
K 1 = 13.0352 9.7596 49.4119 31.3771 , K 2 = 17.2132 6.7025 53.4946 38.5073
L 1 = 198.4849 158.3565 130.5456 63.4251 , L 2 = 198.3110 159.4725 72.5886 24.3458
The switching signal is shown in Figure 13, and the estimated signals and trajectories controlled by the designed observer-based controller are illustrated in Figure 14 and Figure 15. Moreover, the estimation error and system’s output are represented in Figure 16 and Figure 17, respectively. It is perceivable that the observer-based controller is properly operating under the AS problem and with respect to time-delay, uncertainties, and external perturbations.

5. Conclusions

In this study, a control technique was presented for nonlinear SSs under time-delay, uncertainties, and AS problems. In this regard, switched Lyapunov–Krasovskii techniques and the ADT approach were utilized to obtain sufficient stabilization conditions. To derive the observer/controller gains, the obtained conditions were converted into LMIs, and an observer-based control policy was developed to reconstruct the system states and stabilize the closed-loop system. Stabilization conditions were proposed as a feasibility problem that depends on the value of time-delay, upper bounds of the uncertainties and lag time, and Lipschitz constants. Furthermore, proposing a novel set of LMI-based conditions, the H observer-based control problem was investigated for the AS problem against exogenous disturbances. Finally, simulations have demonstrated the superiority of the suggested observer-based control law.

Author Contributions

Conceptualization, A.T., A.M., J.T., S.M., T.R., J.H.A. and A.Z.; Formal analysis, A.T., A.M., J.T., S.M., T.R., J.H.A. and A.Z.; Funding acquisition, A.Z.; Investigation, A.T., A.M., J.T., S.M., T.R., J.H.A. and A.Z.; Methodology, A.T., A.M., J.T., S.M., T.R., J.H.A. and A.Z.; Writing—original draft, A.T. and A.M.; All authors have read and agreed to the published version of the manuscript.

Funding

The research is partially funded by the Ministry of Science and Higher Education of the Russian Federation as part of World-class Research Center program: Advanced Digital Technologies (contract No. 075-15-2020-903 dated 16.11.2020).

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

The study do not report any data.

Acknowledgments

We thank Chiang Mai University for partial support.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Liberzon, D. Finite data-rate feedback stabilization of switched and hybrid linear systems. Automatica 2014, 50, 409–420. [Google Scholar] [CrossRef]
  2. Qi, W.; Park, J.H.; Zong, G.; Cao, J.; Cheng, J. Filter for positive stochastic nonlinear switching systems with phase-type semi-Markov parameters and application. IEEE Trans. Syst. Man Cybern. Syst. 2021. [Google Scholar] [CrossRef]
  3. Gholami, H.; Shafiei, M.H. Finite-time boundedness and stabilization of switched nonlinear systems using auxiliary matrices and average dwell time method. Trans. Inst. Meas. Control 2020, 42, 1406–1416. [Google Scholar] [CrossRef]
  4. Choi, Y.H.; Yoo, S.J. Neural-networks-based adaptive quantized feedback tracking of uncertain nonlinear strict-feedback systems with unknown time delays. J. Frankl. Inst. 2020, 357, 10691–10715. [Google Scholar] [CrossRef]
  5. Yoo, S.J.; Park, B.S. Quantized feedback control strategy for tracking performance guarantee of nonholonomic mobile robots with uncertain nonlinear dynamics. Appl. Math. Comput. 2021, 407, 126349. [Google Scholar] [CrossRef]
  6. Allerhand, L.I.; Shaked, U. Robust Stability and Stabilization of Linear Switched Systems With Dwell Time. IEEE Trans. Autom. Control 2011, 56, 381–386. [Google Scholar] [CrossRef]
  7. Chadli, M.; Darouach, M. Robust admissibility of uncertain switched singular systems. Int. J. Control 2011, 84, 1587–1600. [Google Scholar] [CrossRef]
  8. Ban, J.; Kwon, W.; Won, S.; Kim, S. Robust H finite-time control for discrete-time polytopic uncertain switched linear systems. Nonlinear Anal. Hybrid Syst. 2018, 29, 348–362. [Google Scholar] [CrossRef]
  9. Kani, M.H.H.; Yazdanpanah, M.J.; Markazi, A.H. Stability analysis of a class of uncertain switched time-delay systems with sliding modes. Int. J. Robust Nonlinear Control 2019, 29, 19–42. [Google Scholar] [CrossRef] [Green Version]
  10. Yan, Y.; Zhang, L. Robust Model Predictive Control with Almost Zero Online Computation. Mathematics 2021, 9, 242. [Google Scholar] [CrossRef]
  11. Gholami, H.; Shafiei, M.H. Finite-time H static and dynamic output feedback control for a class of switched nonlinear time-delay systems. Appl. Math. Comput. 2021, 389, 125557. [Google Scholar] [CrossRef]
  12. Baleghi, N.A.; Shafiei, M.H. Stability analysis for discrete-time switched systems with uncertain time delay and affine parametric uncertainties. Trans. Inst. Meas. Control 2018, 40, 1150–1157. [Google Scholar] [CrossRef]
  13. Bagherzadeh, M.A.; Askari, J.; Ghaisari, J.; Mojiri, M. Robust asymptotic stability of parametric switched linear systems with dwell time. IET Control Theory Appl. 2018, 12, 477–483. [Google Scholar] [CrossRef]
  14. Baleghi, N.A.; Shafiei, M.H. Delay-dependent stability analysis for discrete-time switched linear systems with parametric uncertainties. J. Vib. Control 2018, 24, 4921–4930. [Google Scholar] [CrossRef]
  15. Baleghi, N.A.; Shafiei, M.H. Design of static and dynamic output feedback controllers for a discrete-time switched nonlinear system with time-varying delay and parametric uncertainty. IET Control Theory Appl. 2018, 12, 1635–1643. [Google Scholar] [CrossRef]
  16. Baleghi, N.A.; Shafiei, M.H. Stability analysis and stabilization of a class of discrete-time nonlinear switched systems with time-delay and affine parametric uncertainty. J. Vib. Control 2019, 25, 1326–1340. [Google Scholar] [CrossRef]
  17. Taghieh, A.; Shafiei, M.H. Observer-based robust model predictive control of switched nonlinear systems with time delay and parametric uncertainties. J. Vib. Control 2021, 27, 1939–1955. [Google Scholar] [CrossRef]
  18. Zhang, L.; Gao, H. Asynchronously switched control of switched linear systems with average dwell time. Automatica 2010, 46, 953–958. [Google Scholar] [CrossRef]
  19. Yuan, S.; Zhang, L.; Schutter, B.D.; Baldi, S. A novel Lyapunov function for a non-weighted L2 gain of asynchronously switched linear systems. Automatica 2018, 87, 310–317. [Google Scholar] [CrossRef] [Green Version]
  20. Wang, X.; Zong, G.; Sun, H. Asynchronous finite-time dynamic output feedback control for switched time-delay systems with nonlinear disturbances. IET Control Theory Appl. 2016, 10, 1142–1150. [Google Scholar] [CrossRef]
  21. Wang, R.; Xing, J.; Xiang, Z. Finite-time stability and stabilization of switched nonlinear systems with asynchronous switching. Appl. Math. Comput. 2018, 316, 229–244. [Google Scholar] [CrossRef]
  22. Lian, J.; Ge, Y. Robust H output tracking control for switched systems under asynchronous switching. Nonlinear Anal. Hybrid Syst. 2013, 8, 57–68. [Google Scholar] [CrossRef]
  23. Wang, Y.E.; Sun, X.M.; Wu, B.W. Lyapunov–Krasovskii functionals for switched nonlinear input delay systems under asynchronous switching. Automatica 2015, 61, 126–133. [Google Scholar] [CrossRef]
  24. Shi, S.; Fei, Z.; Shi, Z.; Ren, S. Stability and stabilization for discrete-time switched systems with asynchronism. Appl. Math. Comput. 2018, 338, 520–536. [Google Scholar] [CrossRef]
  25. Liu, G.; Hua, C.; Guan, X. Asynchronous stabilization of switched neutral systems: A cooperative stabilizing approach. Nonlinear Anal. Hybrid Syst. 2019, 33, 380–392. [Google Scholar] [CrossRef]
  26. Fu, J.; Li, T.F.; Chai, T.; Su, C.Y. Sampled-data-based stabilization of switched linear neutral systems. Automatica 2016, 72, 92–99. [Google Scholar] [CrossRef]
  27. Huang, J.; Ma, X.; Zhao, X.; Che, H.; Chen, L. Interval observer design method for asynchronous switched systems. IET Control Theory Appl. 2020, 14, 1082–1090. [Google Scholar] [CrossRef]
  28. Ho, D.W.C.; Lu, G. Robust stabilization for a class of discrete-time nonlinear systems via output feedback: The unified LMI approach. Int. J. Control 2003, 76, 105–115. [Google Scholar] [CrossRef]
  29. Hespanha, P.; Morse, A.S. Stability of switched systems with average dwell-time. In Proceedings of the 38th IEEE Conference on Decision and Control (Cat. No.99CH36304), Phoenix, AZ, USA, 7–10 December 1999; Volume 3, pp. 2655–2660. [Google Scholar]
  30. Mahmoud, M. Robust Control and Filtering for Time-Delay Systems; Marcel Dekker: New York, NY, USA, 2000. [Google Scholar]
  31. Taghieh, A.; Shafiei, M.H. Static output feedback control of switched nonlinear systems with time-varying delay and parametric uncertainties under asynchronous switching. Trans. Inst. Meas. Control 2021, 43, 1156–1167. [Google Scholar] [CrossRef]
  32. Mancilla-Aguilar, J.L.; Haimovich, H.; Feketa, P. Uniform stability of nonlinear time-varying impulsive systems with eventually uniformly bounded impulse frequency. Nonlinear Anal. Hybrid Syst. 2020, 38, 100933. [Google Scholar] [CrossRef]
  33. Feketa, P.; Bajcinca, N. On robustness of impulsive stabilization. Automatica 2019, 104, 48–56. [Google Scholar] [CrossRef]
  34. Feketa, P.; Klinshov, V.; Lücken, L. A survey on the modeling of hybrid behaviors: How to account for impulsive jumps properly. Commun. Nonlinear Sci. Numer. Simul. 2021, 103, 105955. [Google Scholar] [CrossRef]
  35. Li, L.; Song, L.; Li, T.; Fu, J. Event-Triggered Output Regulation for Networked Flight Control System Based on an Asynchronous Switched System Approach. IEEE Trans. Syst. Man Cybern. Syst. 2020, 51, 7675–7684. [Google Scholar] [CrossRef]
Figure 1. Example 1: Switching signal.
Figure 1. Example 1: Switching signal.
Mathematics 09 02968 g001
Figure 2. Example 1: First state variable and its corresponding estimation.
Figure 2. Example 1: First state variable and its corresponding estimation.
Mathematics 09 02968 g002
Figure 3. Example 1: Second state variable and its corresponding estimation.
Figure 3. Example 1: Second state variable and its corresponding estimation.
Mathematics 09 02968 g003
Figure 4. Example 1: Estimation error.
Figure 4. Example 1: Estimation error.
Mathematics 09 02968 g004
Figure 5. Example 1: Output signal.
Figure 5. Example 1: Output signal.
Mathematics 09 02968 g005
Figure 6. Example 1: Control signal.
Figure 6. Example 1: Control signal.
Mathematics 09 02968 g006
Figure 7. Example 2: Switching signal.
Figure 7. Example 2: Switching signal.
Mathematics 09 02968 g007
Figure 8. Example 2: First state variable and its estimation.
Figure 8. Example 2: First state variable and its estimation.
Mathematics 09 02968 g008
Figure 9. Example 2: Second state variable and its estimation.
Figure 9. Example 2: Second state variable and its estimation.
Mathematics 09 02968 g009
Figure 10. Example 2: Time response of the estimation error.
Figure 10. Example 2: Time response of the estimation error.
Mathematics 09 02968 g010
Figure 11. Example 2: Time response of output.
Figure 11. Example 2: Time response of output.
Mathematics 09 02968 g011
Figure 12. Example 2: Control signal.
Figure 12. Example 2: Control signal.
Mathematics 09 02968 g012
Figure 13. Example 3: Switching signal.
Figure 13. Example 3: Switching signal.
Mathematics 09 02968 g013
Figure 14. Example 3: First state and its estimation.
Figure 14. Example 3: First state and its estimation.
Mathematics 09 02968 g014
Figure 15. Example 3: Second state and its estimation.
Figure 15. Example 3: Second state and its estimation.
Mathematics 09 02968 g015
Figure 16. Example 3: Estimation error.
Figure 16. Example 3: Estimation error.
Mathematics 09 02968 g016
Figure 17. Example 3: Time response of the F 18 aircraft system’s output.
Figure 17. Example 3: Time response of the F 18 aircraft system’s output.
Mathematics 09 02968 g017
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Share and Cite

MDPI and ACS Style

Taghieh, A.; Mohammadzadeh, A.; Tavoosi, J.; Mobayen, S.; Rojsiraphisal, T.; Asad, J.H.; Zhilenkov, A. Observer-Based Control for Nonlinear Time-Delayed Asynchronously Switching Systems: A New LMI Approach. Mathematics 2021, 9, 2968. https://0-doi-org.brum.beds.ac.uk/10.3390/math9222968

AMA Style

Taghieh A, Mohammadzadeh A, Tavoosi J, Mobayen S, Rojsiraphisal T, Asad JH, Zhilenkov A. Observer-Based Control for Nonlinear Time-Delayed Asynchronously Switching Systems: A New LMI Approach. Mathematics. 2021; 9(22):2968. https://0-doi-org.brum.beds.ac.uk/10.3390/math9222968

Chicago/Turabian Style

Taghieh, Amin, Ardashir Mohammadzadeh, Jafar Tavoosi, Saleh Mobayen, Thaned Rojsiraphisal, Jihad H. Asad, and Anton Zhilenkov. 2021. "Observer-Based Control for Nonlinear Time-Delayed Asynchronously Switching Systems: A New LMI Approach" Mathematics 9, no. 22: 2968. https://0-doi-org.brum.beds.ac.uk/10.3390/math9222968

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop