Processing math: 32%
Research article

Robust stabilization for uncertain saturated systems with multiple time delays

  • Received: 20 May 2022 Revised: 11 August 2022 Accepted: 15 August 2022 Published: 30 August 2022
  • MSC : 93D09

  • This paper is concerned with the robust stabilization problem for uncertain saturated linear systems with multiple discrete delays. First of all, a new distributed-delay-dependent polytopic approach is proposed, and a new type of Lyapunov-Krasovskii functional is constructed. Then, by further incorporating some integral inequalities, both stabilization and robust stabilization conditions are proposed in terms of linear matrix inequalities under which the closed-loop systems are asymptotically stable for admissible initial conditions. Finally, a simulation example is given to illustrate the feasibility and advantages of the obtained results.

    Citation: Yuzhen Chen, Haoxin Liu, Rui Dong. Robust stabilization for uncertain saturated systems with multiple time delays[J]. AIMS Mathematics, 2022, 7(10): 19180-19201. doi: 10.3934/math.20221053

    Related Papers:

    [1] Danhua He, Baizeng Bao, Liguang Xu . Robust stability and boundedness of uncertain conformable fractional-order delay systems under input saturation. AIMS Mathematics, 2023, 8(9): 21123-21137. doi: 10.3934/math.20231076
    [2] Guijun Xing, Huatao Chen, Zahra S. Aghayan, Jingfei Jiang, Juan L. G. Guirao . Tracking control for a class of fractional order uncertain systems with time-delay based on composite nonlinear feedback control. AIMS Mathematics, 2024, 9(5): 13058-13076. doi: 10.3934/math.2024637
    [3] Xinxin Zhang, Yunpeng Ma, Shan Gao, Jiancai Song, Lei Chen . Robust synchronization analysis of delayed fractional order neural networks with uncertain parameters. AIMS Mathematics, 2022, 7(10): 18883-18896. doi: 10.3934/math.20221040
    [4] Zhifu Jia, Xinsheng Liu . New stability theorems of uncertain differential equations with time-dependent delay. AIMS Mathematics, 2021, 6(1): 623-642. doi: 10.3934/math.2021038
    [5] Aziz Belmiloudi . Time-varying delays in electrophysiological wave propagation along cardiac tissue and minimax control problems associated with uncertain bidomain type models. AIMS Mathematics, 2019, 4(3): 928-983. doi: 10.3934/math.2019.3.928
    [6] R. Sriraman, R. Samidurai, V. C. Amritha, G. Rachakit, Prasanalakshmi Balaji . System decomposition-based stability criteria for Takagi-Sugeno fuzzy uncertain stochastic delayed neural networks in quaternion field. AIMS Mathematics, 2023, 8(5): 11589-11616. doi: 10.3934/math.2023587
    [7] Liping Luo, Yonggang Chen, Jishen Jia, Kaixin Zhao, Jinze Jia . Event-triggered anti-windup strategy for time-delay systems subject to saturating actuators. AIMS Mathematics, 2024, 9(10): 27721-27738. doi: 10.3934/math.20241346
    [8] Ahmed M. Elaiw, Ghadeer S. Alsaadi, Aatef D. Hobiny . Global co-dynamics of viral infections with saturated incidence. AIMS Mathematics, 2024, 9(6): 13770-13818. doi: 10.3934/math.2024671
    [9] Biwen Li, Qiaoping Huang . Synchronization issue of uncertain time-delay systems based on flexible impulsive control. AIMS Mathematics, 2024, 9(10): 26538-26556. doi: 10.3934/math.20241291
    [10] Li Wan, Qinghua Zhou, Hongbo Fu, Qunjiao Zhang . Exponential stability of Hopfield neural networks of neutral type with multiple time-varying delays. AIMS Mathematics, 2021, 6(8): 8030-8043. doi: 10.3934/math.2021466
  • This paper is concerned with the robust stabilization problem for uncertain saturated linear systems with multiple discrete delays. First of all, a new distributed-delay-dependent polytopic approach is proposed, and a new type of Lyapunov-Krasovskii functional is constructed. Then, by further incorporating some integral inequalities, both stabilization and robust stabilization conditions are proposed in terms of linear matrix inequalities under which the closed-loop systems are asymptotically stable for admissible initial conditions. Finally, a simulation example is given to illustrate the feasibility and advantages of the obtained results.



    Over the past decades, time-delay systems have received significant research attention due to the ubiquity of time delays in many dynamical systems, see, e.g., [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29] and the references therein. In particular, in [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19], some researchers have considered the systems with multiple time delays. The motivation for studying multiple time-delay systems is that multiple time delays are often encountered in many practical systems such as power systems [5,6,7,8], complex networks [11,12], neural networks [4,13], multi-agent systems [18,19]. For example, in [6,7], multi-area power systems with time delays have been modeled as linear systems with multiple time delays. The main techniques for studying multiple time-delay systems are the characteristic roots approach [1,2], the state trajectory approach [3] and the Lyapunov-Krasovskii (L-K) approach [30,31,32]. Using the L-K approach, the obtained results can be expressed by linear matrix inequalities (LMIs), which can be easily solved by existing software. In particular, in [31,32], to derive some less conservative stability criteria in the framework of LMIs, some cross-terms related to multiple different time delays have been introduced into the constructed L-K functionals. However, it is worth mentioning that the results proposed in [31,32] are based on the free-weighting matrices technique and the Jensen inequality, which are still conservative to some extent.

    On the other hand, it is well known that almost all practical feedback control systems are limited by the amplitude or rate of physical actuators for safety reasons or the limitations of the physical components. For example, a motor cannot generate infinite torque, and the amplifier output voltage cannot be unlimited. When an aircraft's vertical tail controls the horizontal steering, its steering angle has a maximum limit, and its rotation rate cannot be too fast; otherwise, the aircraft will roll over and cause an accident. Actuator saturation can cause instability and performance degradation of the overall system. During the past two decades, much effort has been focused on linear systems with actuator saturation [33,34,35,36,37,38,39,40,41,42,43,44]. One of the important research questions is related to global and semi-global stabilization [33,34,35,36]. Global stabilization is not possible if the open-loop system is exponentially unstable [33]. Furthermore, global stabilization cannot, in general, be achieved by linear feedback even for open-loop systems that are not exponentially unstable [35]. The low-gain design technique and the dynamic scheduling approach have been utilized in [36], and the parametric Lyapunov equation-based low-gain design was proposed in [37]. The local stabilization problem has been well discussed for open-loop systems that are unstable [33,34,38]. In particular, two dominant approaches to deal with saturation nonlinearities are the polytopic models [33,38,45] and the generalized sector condition [34,46].

    By utilizing the polytopic approach in [33], time-delay systems with actuator saturations have been transformed into the convex polytope of linear systems [39,40]. Then, by incorporating the L-K approach, LMI-based sufficient conditions have been established in [39,40], under which the local stability of closed-loop systems can be guaranteed. Based on the generalized sector condition, the asynchronous H control problem has been addressed in [41] for time-delayed switched systems with actuator saturation via the anti-windup scheme. By introducing auxiliary time-delay feedbacks, the delay-dependent polytopic approach was developed in [42,43,44,47] to reduce the potential conservatism. However, the results in [39,40,41,42,43,44,48,49] are mainly concerned with systems with a single time delay. To the best of our knowledge, the local stabilization problem for multiple time-delay systems with actuator saturation has not been considered, likely due to the mathematical complexity.

    Motivated by the above discussions, in this paper, the problems of local stabilization and the corresponding estimate of the domain of attraction are considered for multiple time-delay systems with actuator saturation. First, a distributed-delay-dependent polytopic approach is proposed, and the saturation nonlinearly is represented by the convex combination of state feedback and auxiliary distributed-delay feedback. Then, based on the L-K approach, both the stabilization and robust stabilization conditions are established in terms of LMIs. Subsequently, the optimization problems regarding the maximization of the domain of attraction are discussed. The main contributions of this work are summarized below.

    1) The results proposed in this paper are quite general since many factors are considered, such as norm-bounded uncertainties, multiple time delays, and actuator saturation. Therefore, the results obtained in this paper generalize to existing results in the literature.

    2) Different from the L-K functionals used in [30,31,32,34], a novel L-K functional is constructed in this paper. Specifically, the proposed functional contains augmented state vectors and some interconnected terms about multiple delays, which will lead to less conservative results.

    3) A new polytopic model is proposed to represent the saturation nonlinearity. In particular, some interconnected terms concerning multiple delays are utilized, which is useful in reducing the possible conservatism.

    Notation: The superscript "T" denotes matrix transposition; I denotes the identity matrix of appropriate dimensions; λM(P) denotes the maximum eigenvalue of matrix P; Sym(E) is a shorthand notation for matrix E+ET; and Cn,d=C([d,0],Rn) denotes the Banach space of continuous vector functions mapping interval [d,0] into Rn using the topology of uniform convergence. φ(t)cmaxt[d,0]φ(t)2 stands for the norm of a function φ(t)Cn,d; Dm is the set of m×m diagonal matrices, where the diagonal elements are either 1 or 0; \(e_{m, i} \in \mathbb{R}^{1 \times m}\) denotes a row vector whose i-th element is 1 and the others are 0; I[1,N] denotes the set {1,2,,N}; and co{h1,h2,,hm} denotes the convex hull of the vectors.

    Consider the following uncertain saturated linear system with multiple discrete delays:

    ˙x(t)=A0(t)x(t)+Nj=1Aj(t)x(tdj)+B(t)sat(u(t)),t>0, (1)
    x(t)=φ(t),t[d,0],d=max1jN{dj}, (2)

    where x(t)Rn is the system state; u(t)Rm is the control input; the time delays dj,jI[1,N] are known scalars; φ(t)Cn,d denotes the initial function; sat(u)=[sat(u1)sat(u2)sat(um)]T is the standard saturation function with sat(ui)=sgn(ui)min{1,|ui|}; and A0(t),A1(t),,AN(t) and B(t) are time-varying matrices that satisfy A0(t)=A0+ΔA0(t), Aj(t)=Aj+ΔAj(t)(jI[1,N]) and B(t)=B+ΔB(t), where Aj and B are known constant matrices with appropriate dimensions.

    Remark 1. Different from most existing literature concerning control systems with actuator saturation [32,33,34,35,36,37,38,39,40,41], the systems (1) and (2) considered in this paper contains multiple time delays. Note that multiple time delays are often encountered in many practical systems such as power systems [5,6,7,8], complex networks [11,12], neural networks [4,13,29], and multi-agent systems [18,19]. Therefore, the results obtained in this paper can be seen as an indispensable supplement to the existing literature.

    Assumption 1. The uncertain matrices ΔA0(t), ΔAj(t)(jI[1,N]), ΔAN(t) and ΔB(t) satisfy

    [ΔA0(t)ΔA1(t)ΔAN(t)ΔB(t)]Δ=MF(t)[E0E1ENEB], (3)

    where M,E0,E1,,EN and EB are known real constant matrices, and F(t) is an unknown time-varying matrix satisfying FT(t)F(t)I.

    This paper employs the following state feedback controller:

    u(t)=Kx(t), (4)

    where KRm×n is the gain matrix to be designed.

    Lemma 1. [38] Denote   mΔ=m2m1. Let wR  m be such that ||w||1, DkDm, kI[1,2m]. The function fm is defined as fm(0)=0 and

    fm(k)={fm(k1)+1,Dk+DkIm,kI[1,k],fm(k),Dk+Dk=Im,kI[1,k].

    Letting denote the Kronecker product, for any uRm, the following holds:

    sat(u)co{Dku+Dkw:kI[1,2m]},

    where DkRm×  m is defined as Dk=e2m1,fm(k)Dk with Dk=IDk.

    Lemma 2. [50] For a given symmetric positive definite R and any differentiable function x in [d,0]Rn, the following inequality holds:

    0d˙xT(θ)R˙x(θ)dθ1d[Θ0Θ1]T[R003R][Θ0Θ1],

    where

    Θ0=x(0)x(d),Θ1=x(0)+x(d)2d0dx(θ)dθ.

    In addtion, to represent the saturation nonlinearity in a delay-depdent framwork, we introduce the following lemma directly derived from Lemma 1.

    Lemma 3. Let us define the following functional:

    w(t)=Ux(t)+Nj=1Vjttdjx(s)ds+N1ˆi=1Nˆj=ˆi+1Vˆiˆjtdˆitdˆjx(s)ds, (5)

    where U, VjR  m×n, jI[1,N], VˆiˆjR  m×n, ˆiI[1,N1], and ˆjI[ˆi+1,N]. If the constraint

    ||w(t)||1 (6)

    is satisfied, the saturation nonlinearity sat(u(t)) can be written as follows:

    sat(u(t))=2mk=1λtk[(DkK+DkU)x(t)+DkNj=1Vjttdjx(s)ds+DkN1ˆi=1Nˆj=ˆi+1Vˆiˆjtdˆitdˆjx(s)ds], (7)

    where λt10, λt20, , λt2m0 and 2mk=1λtk=1.

    Proof. From (5) and Lemma 1, the (7) can be easily obtained.

    Using (1) and (7), one can obtain the following closed-loop system:

    ˙x(t)=2mk=1λtk{[A0(t)+B(t)(DkK+DkU)]x(t)+Nj=1Aj(t)x(tdj)+Nj=1B(t)DkVjttdjx(s)ds+N1ˆi=1Nˆj=ˆi+1B(t)DkVˆiˆjtdˆitdˆjx(s)ds}Δ=η(t). (8)

    Remark 2. Compared with the traditional polytope model [38], it can be seen that the polytope model (7) contains the auxiliary time-delay feedback Nj=1Vjttdjx(s)ds and N1ˆi=1Nˆj=ˆi+1Vˆiˆjtdˆitdˆjx(s)ds related to different time delays. Since the time delay information is sufficiently used in representing the saturation nonlinearity, it is expected that less conservative stabilization conditions can be achieved.

    In this paper, we assume that the admissible initial conditions φ(t) belong to the following set:

    XρΔ={φ(t)Cn,d:||φ(t)||cρ1,||˙φ(t)||cρ2}, (9)

    where ρ1 and ρ2 are some positive scalars to be optimized.

    In this paper, we are interested in designing the state feedback controller (4) such that the closed-loop system (8) is locally stable with an estimate of the region of attraction that is as large as possible.

    In this paper, we choose the following L-K functional:

    V(t)=V1(t)+V2(t)+V3(t), (10)

    where V1(t)=ξT(t)Pξ(t) and

    V2(t)=Nj=1[ttdjxT(s)Sjx(s)ds+dj0djtt+θ˙xT(s)Rj˙x(s)dsdθ],V3(t)=N1ˆi=1Nˆj=ˆi+1dˆjˆi[tdˆitdˆjxT(s)Qˆiˆjx(s)ds+dˆidˆjtt+θ˙xT(s)Zˆiˆj˙x(s)dsdθ],

    with P>0, Rj>0, Sj>0, Qˆiˆj>0, Zˆiˆj>0, dˆjˆiΔ=dˆjdˆi (jI[1,N], ˆiI[1,N1], ˆjI[ˆi+1,N]).

    For convenience in subsequent presentation, we denote

    ξ(t)Δ=[xT(t)ttd1xT(s)dsttdNxT(s)dstd1td2xT(s)dstd1tdNxT(s)dstd2td3xT(s)dstd2tdNxT(s)dstd3td4xT(s)dstd3tdNxT(s)dstdN1tdNxT(s)ds]T,ζ(t)Δ=[xT(t)xT(td1)xT(tdN)ttd1xT(s)dsttdNxT(s)dstd1td2xT(s)dstd1tdNxT(s)dstdN1tdNxT(s)ds˙xT(t)]T,ˆAΔ=[A1A2AN]T,N1Δ=(N2+N+2)n/2,N2Δ=N(N1)n/2,N3Δ=N(N+1)n/2,N4Δ=(N2+3N+2)n/2,˜dˆiˆjΔ=12/(dˆjdˆi)2,ˉdˆiˆjΔ=2/(dˆjdˆi),˜djΔ=12/d2j,ˉdjΔ=1/dj,ˆΦ2Δ=[IT1IT2IT1IT3IT1ITN+1IT2IT3IT2ITN+1ITNITN+1]T,IjΔ=[0n×(j1)nIn0n×(Nj)n],ItΔ=[0n×(t1)nIn0n×(N+1t)n]T,

    where jI[1,N], tI[1,N+1].

    First, we consider the stabilization problem for the system (1) without uncertainties, i.e., the system matrices in (1) satisfy F(t)=0.

    Theorem 1. Let the constants δ0 and dj, jI[1,N] be given. System (1) without uncertainties can be asymptotically stabilized by controller (4) with K=YXT if for any initial function φ(t)Cn,d satisfying V(0)1, there exist N1×N1 matrix ˉP>0, n×n matrices ˉRj>0, ˉSj>0, ˉQˆiˆj>0, ˉZˆiˆj>0, n×n invertible matrix X, m×n matrix Y, and   m×n matrices G, Hj, jI[1,N], Hˆiˆj, ˆiI[1,N1], ˆjI[ˆi+1,N], such that the following LMIs hold:

    (ˉΩkrs)5×5+Sym(ΦT1ˉPΦ2)ˉΩk<0,kI[1,2m], (11)
    [1ˉFlˉP]+[001×n01×N3ˉΣ11ˉϑˉΞ]0,lI[1,m], (12)

    where ˉFl is the lth row of ˉF=[GH1H2HNH12H1NH23H2NH(N1)N], lI[1,  m] and

    ˉΩk11=Nj=1(ˉSj4ˉRj)+Sym[A0XT+B(DkY+DkG)],ˉΩk12=[2ˉR1+A1XT2ˉR2+A2XT2ˉRN+ANXT],ˉΩk13=[ˉβk1ˉβk2ˉβkN],ˉΩk14=[ˉγk12ˉγk13ˉγk1Nˉγk23ˉγk2Nˉγk(N1)N],ˉΩk15=XT+δXAT0+δ(DkY+DkG)TBT,ˉΩk22=[ˉα22ˉZ122ˉZ1(N1)2ˉZ1Nˉα32ˉZ2(N1)2ˉZ2NˉαN2ˉZ(N1)NˉαN+1],ˉΩk23=6diag{ˉd1ˉR1ˉd2ˉR2ˉdNˉRN},ˉΩk24=3[ˉξ12ˉξ13ˉξ1Nˉξ23ˉξ2Nˉξ(N1)N],ˉΩk25=δXˆA,ˉΩk33=diag{˜d1ˉR1˜d2ˉR2˜dNˉRN},ˉΩk34=0Nn×N2,ˉΩk35=[ˉσk1ˉσk2ˉσkN]T,ˉΩk44=diag{˜d12ˉZ12˜d1NˉZ1N˜d23ˉZ23˜d2NˉZ2N˜d(N1)NˉZ(N1)N},ˉΩk45=[ˉμk12ˉμk1Nˉμk23ˉμk2Nˉμk(N1)N]T,ˉΩk55=δSym(X)+Nj=1d2jˉRj+N1ˆi=1Nˆj=ˆi+1d2ˆiˆjˉZˆiˆj,Φ1=[I0n×Nn0000n×n00n×NnI000n×n00n×Nn0I00n×n00n×Nn00I0n×n],Φ2=[0n×N4IˆΦ20n×N3],ˉΣ11=2Nj=1djˉRj+2N1ˆi=1Nˆj=ˆi+1dˆjˆiˉZˆiˆj,ˉϑ=[ˉΣ12ˉΣ13ˉΣ1(N+1)ˉΣ121ˉΣ131ˉΣ(N1)N1],ˉΞ=diag{ˉΣ22ˉΣ(N+1)(N+1)ˉΣ1212ˉΣ1313ˉΣ(N1)N(N1)N},

    with

    ˉα2=4ˉR1ˉS1+Nˆj=2dˆj1ˉQ1ˆj4Nˆj=2ˉZ1ˆj,ˉαt+1=4ˉRtˉSt+Nˆj=t+1(dˆjtˉQtˆj4ˉZtˆj)t1ˆi=1(dtˆiˉQˆit+4ˉZˆit),tI[2,N1],ˉαN+1=4ˉRNˉSNN1ˆi=1dNˆiˉQˆiN4N1ˆi=1ˉZˆiN,ˉβkj=6ˉdjˉRj+BDkHj,ˉΣ1(j+1)=2ˉRj,ˉγkˆiˆj=BDkHˆiˆj,ˉξˆiˆj=ˉdˆiˆjˉZˆiˆj(Iˆi+Iˆj),ˉσkj=δBDkHj,ˉμkˆiˆj=δBDkHˆiˆj,ˉΣˆiˆj1=2ˉZˆiˆj,ˉΣˆiˆjˆiˆj=ˉQˆiˆj+ˉdˆiˆjˉZˆiˆj,ˉΣ(j+1)(j+1)=ˉdj(2ˉRj+ˉSj),jI[1,N],ˆiI[1,N1],ˆjI[ˆi+1,N].

    Proof. Differentiating V(t) in (10) along the closed-loop system (8) yields

    \begin{align*} \dot V(t) = &2{\xi ^T}(t)P\dot \xi (t)+\sum\limits_{j = 1}^N \bigg[x^T(t){S_j}x(t)+ d_j^2\dot x^T(t)R_j\dot x(t) -x^T(t-d_j){S_j}x(t-d_j) \\ &-d_j\int\nolimits_{t-d_j}^t\dot x^T(s)R_j\dot x(s)\mathrm{d}s \bigg]+\sum\limits_{\hat i = 1}^{N-1} \sum\limits_{\hat j = \hat i+1}^N d_{\hat j\hat i}\bigg\{ \bigg[ {x^T}(t-d_{\hat i})Q_{\hat i\hat j} {x(t-d_{\hat i})} \\ &- {x^T}(t - d_{\hat j})Q_{\hat i \hat j}x(t -d_{\hat j}) \bigg]+ d_{\hat j \hat i}{{\dot x}^T}(t)Z_{\hat i\hat j}\dot x(t) - \int\nolimits_{t - d_{\hat j}}^{t - d_{\hat i}} {{{\dot x}^T}(s)Z_{\hat i\hat j}\dot x(s)\mathrm{d}s} \bigg\}. \end{align*} (13)

    Using the Wirtinger-based inequality (Lemma 2) for \int\nolimits_{t - {d_j}}^t {{{\dot x}^T}(s){R_j}\dot x(s)\mathrm{d}s} and \int_{t - {d_{\hat j}}}^{t - {d_{\hat i}}} {{{\dot x}^T}(s){Z_{\hat i \hat j}}\dot x(s)\mathrm{d}s} , we have

    \begin{align*} -{d_j}\int\nolimits_{t-{d_j}}^t {{{\dot x}^T}(s){R_j}\dot x(s)\mathrm{d}s} \leqslant -{[x(t)-x(t-{d_j})]^T}{R_j} [x(t)-x(t-{d_j})]-3\Psi _j^T{R_j}{\Psi _j}, \end{align*} (14)
    -{d_{\hat j\hat i}}\int\nolimits_{t-{d_{\hat j}}}^{t - {d_{\hat i}}} {{{\dot x}^T}(s){Z_{\hat i \hat j}}\dot x(s)\mathrm{d}s} \leqslant \\ -{[x(t-{d_{\hat i}})-x(t-{d_{\hat j}})]^T}{Z_{\hat i \hat j}}[x(t-{d_{\hat i}})-x(t-{d_{\hat j}})]- 3\Gamma _{\hat i \hat j}^T{Z_{\hat i\hat j}}{\Gamma _{\hat i\hat j}}, (15)

    where

    \begin{align*} {\Psi _j} = x(t) + x(t - {d_j}) - 2{\bar d_j}\int_{t - {d_j}}^t {x(s)\mathrm{d}s},\\ {\Gamma _{\hat i\hat j}} = x(t - {d_{\hat i}}) + x(t - {d_{\hat j}}) - {\bar d_{\hat i\hat j}}\int_{t - {d_{\hat j}}}^{t - {d_{\hat i}}} {x(s)\mathrm{d}s}. \end{align*}

    For any matrices {T_1} , {T_2} \in {\mathbb{R}^{n \times n}} , it follows from the closed-loop system (8) that

    \begin{align*} 2[{x^T}(t){T_1} + {\dot x^T}(t){T_2}][\eta (t) - \dot x(t)] = 0. \end{align*} (16)

    Adding the left side of (16) to \dot V(t) in (13) and substituting (14) and (15) into (13), one can obtain the following inequality:

    \begin{align*} \dot V(t) \leqslant \sum\limits_{k = 1}^{{2^m}} {\lambda _k^t{\zeta ^T}} (t){\Omega ^k}\zeta (t), \end{align*} (17)

    where

    \begin{align*} {\Omega}^k \buildrel \Delta \over = ({\Omega}^k_{rs})_{5\times5}+{\rm {Sym}}({\Phi^T_1}{ P}{\Phi_2}), \end{align*}

    with

    \begin{array}{*{20}{c}} {\Omega}^k_{11} = \sum\limits_{j = 1}^N {({{ S}_j} - 4{{ R}_j}} )+{T_1}\big[ {{A_0} + B({D_k}K + \mathcal{D}_k^ - U)} \big] + {\big[ {{A_0} + B({D_k}K + \mathcal{D}_k^ - U)} \big]^T}T_1^T,\\ {\Omega}^k_{12} = \big[ {-2{{R}_1}+T_1{A_1}}\; {-2{{R}_2}+T_1{A_2}}\; \cdots\; - 2{{R}_N}+T_1{A_N} \big],\; {\Omega}^k_{13} = \big[ {\beta _1^k}\; {\beta _2^k}\; \cdots\; {\beta _N^k} \big],\\ {\Omega}^k_{14} = \big[ {\gamma _{12}^k}\; {\gamma _{13}^k} \; \cdots\; {\gamma _{1N}^k}\; {\gamma _{23}^k} \; \cdots\; {\gamma _{2N}^k} \; \cdots\; {\gamma _{(N - 1)N}^k} \big],\; {\Omega}^k_{15} = -{T_1}+{\big[{A_0}+B({D_k}K+\mathcal{D}_k^- U)\big]^T}T_2^T,\\ {\Omega}^k_{22} = \left[ {\begin{array}{*{20}{c}} {{{\alpha }_2}}&{- 2{{Z}_{12}}}& \cdots &{-2{{Z}_{1(N - 1)}}}&{-2{{Z}_{1N}}} \\ * &{{{\alpha }_3}}& \cdots &{ - 2{{Z}_{2(N - 1)}}}&{ - 2{{Z}_{2N}}} \\ \vdots & \vdots & \ddots & \vdots & \vdots \\ * & * & * &{{{\alpha }_N}}&{ - 2{{Z}_{(N - 1)N}}} \\ * & * & * & * &{{{\alpha }_{N + 1}}} \end{array}} \right] ,\\ {\Omega}^k_{23} = 6{{\rm{diag}}}\left\{ {{{\bar d}_1}{{R}_1}\; {{\bar d}_2}{{ R}_2}\; \cdots\; {{\bar d}_N}{{R}_N}} \right\},\\ {\Omega}^k_{24} = 3\big[ {\xi_{12}}\; {\xi _{13}} \; \cdots\; {\xi_{1N}}\; {\xi _{23}} \; \cdots\; {\xi_{2N}} \; \cdots\; {\xi _{(N-1)N}} \big],\\ {\Omega}^k_{25} = \hat A T_2^T,\; {\Omega}^k_{33} = {\rm{diag}}\left\{\tilde d_1{R_1}\; \tilde d_2 R_2\; \cdots\; \tilde d_N R_N\right\},\\ {\Omega}^k_{34} = {{0_{Nn \times {N_2}}}},\; {\Omega}^k_{35} = {\big[ {\sigma _1^k}\; {\sigma _2^k}\; \cdots \; {\sigma _N^k} \big]^T},\\ {\Omega}^k_{44} = {{\rm{diag}}}\Big\{ {{\tilde d}_{12}}{{ Z}_{12}}\; \cdots\; {{\tilde d}_{1N}}{{Z}_{1N}}\; {{\tilde d}_{23}}{{Z}_{23}}\; \cdots\; {{\tilde d}_{2N}}{{Z}_{2N}}\; \cdots\; {{\tilde d}_{(N - 1)N}}{{Z}_{(N - 1)N}} \Big\},\\ {\Omega}^k_{45} = {\big[ {\mu _{12}^k}\; \cdots \; {\mu _{1N}^k}\; {\mu _{23}^k}\; \cdots\; {\mu _{2N}^k}\; \cdots\; {\mu _{(N - 1)N}^k} \big]^T},\\ {\Omega}^k_{55} = -{\rm {Sym}}(T_2) + \sum\limits_{j = 1}^N {d_j^2{{R}_j} + } \sum\limits_{\hat i = 1}^{N - 1} {\sum\limits_{\hat j = \hat i + 1}^N {d_{\hat i\hat j}^2{{Z}_{\hat i\hat j}}} },\\ {\Phi_1} = \left[ {\begin{array}{*{20}{c}} I&0_{n\times Nn}&0&0&\cdots&0&0_{n\times n}\\ 0&0_{n\times Nn}&I&0&\cdots&0&0_{n\times n}\\ 0&0_{n\times Nn}&0&I&\cdots&0&0_{n\times n}\\ \vdots& \vdots& \vdots& \vdots& \ddots& \vdots& \vdots&\\ 0&0_{n\times Nn}&0&0&\cdots&I&0_{n\times n} \end{array}} \right],\; {\Phi_2} = \left[ {\begin{array}{*{20}{c}} 0_{n\times N_4}&I\\ {\hat \Phi }_2&0_{n\times N_3} \end{array}} \right],\\ {\alpha _{t + 1}} = -4{R_t} - {S_t} + \sum\limits_{\hat j = t + 1}^N {\left( {{d_{\hat jt}}{{Q}_{t\hat j}} - 4{{Z}_{t\hat j}}} \right)} -\sum\limits_{\hat i = 1}^{t - 1} {\left( {{d_{t\hat i}}{{Q}_{\hat it}} + 4{{Z}_{\hat it}}} \right)},\; t \in {{\bf{I}}}[2,N - 1],\\ {\alpha_{N+1}} = -4{R_N}-{S_N}-\sum\limits_{\hat i = 1}^{N-1} {{d_{N\hat i}}{{Q}_{\hat iN}}} - 4\sum\limits_{\hat i = 1}^{N - 1} {{{Z}_{\hat iN}}},\; {\alpha _2} = - 4{R_1} - {S_1} + \sum\limits_{\hat j = 2}^N {{d_{\hat j1}}{{Q}_{1\hat j}} - 4\sum\limits_{\hat j = 2}^N {{{Z}_{1\hat j}}} },\\ \gamma _{\hat i\hat j}^k = T_1 B\mathcal{D}_k^- {V_{\hat i\hat j}},\; \xi _{\hat i\hat j} = d_{\hat i\hat j}Z_{\hat i\hat j}({I'}_{\hat i} + {I'}_{\hat j}),\; \mu _{\hat i\hat j}^k = T_2 B\mathcal{D}_k^ - {V_{\hat i\hat j}},\; \beta _j^k = 6{\bar d_j}{R_j} + T_1 B\mathcal{D}_k^ - {V_j},\\ \sigma _j^k = T_2 B\mathcal{D}_k^ - {V_j},\; \hat i \in {{\bf{I}}}[1,N - 1],\; \hat j \in {{\bf{I}}}[\hat i + 1,N],\; j \in{{\bf{I}}}[1,N]. \end{array}

    Suppose that the following matrix inequality holds:

    \begin{align*} {\Omega ^k} < 0. \end{align*}

    Then, it is seen from (17) that \dot V(t) < 0 can be ensured. Furthermore, \dot V(t) < 0 means that

    \begin{align*} V(t) \leqslant V(0),\; t \geqslant 0. \end{align*} (18)

    Next, we will show that the condition described by (6) can be ensured. For the functional V(t) defined in (10), using Jensen inequalities for the terms {d_{\hat j\hat i}}\int_{t - {d_{\hat j}}}^{t - {d_{\hat i}}} {{x^T}(s){Q_{\hat i \hat j}}x(s)\mathrm{d}s} and {d_{\hat j \hat i}}\int_{ - {d_{\hat j}}}^{ - {d_{\hat i}}} {\int_{t + \theta }^t {{{\dot x}^T}(s){Z_{\hat i \hat j}}\dot x(s)\mathrm{d}s\mathrm{d}\theta } } , \hat i \in {{\bf{I}}}[1, N - 1] , \hat j \in {{\bf{I}}}[\hat i + 1, N] and {d_j}\int_{ - {d_j}}^0 {\int_{t + \theta }^t {{{\dot x}^T}(s){R_j}\dot x(s)\mathrm{d}s\mathrm{d}\theta } } , j \in {{\bf{I}}}[1, N] , it follows that

    \begin{align*} V(t)& \geqslant \sum\limits_{j = 1}^N \bigg\{2{x^T}(t){R_j}\bigg[ {d_j}x(t)-2\int\nolimits_{t-{d_j}}^t {x(s)\mathrm{d}s} \bigg] +\bar d_j\int\nolimits_{t- d_j}^t {x^T}(s)\mathrm{d}s \left(2R_j + S_j\right)\int\nolimits_{t-d_j}^t x(s)\mathrm{d}s \bigg\}\\ &+\sum\limits_{\hat i = 1}^{N-1} \sum\limits_{\hat j = \hat i+1}^N \bigg\{ 2{d_{\hat j\hat i}}{x^T}(t){Z_{\hat i\hat j}}x(t)- \bigg[4x^T(t)+\bar d_{\hat i\hat j}\int\nolimits_{t-d_{\hat j}}^{t-d_{\hat i}} x^T(s)\mathrm{d}s\bigg]Z_{\hat i\hat j}\int\nolimits_{t-d_{\hat j}}^{t-d_{\hat i}} {x(s)\mathrm{d}s}\bigg\}\\ &+ \sum\limits_{\hat i = 1}^{N - 1} {\sum\limits_{\hat j = \hat i + 1}^N {\int\nolimits_{t - {d_{\hat j}}}^{t - {d_{\hat i}}} {{x^T}(s)\mathrm{d}s} {Q_{\hat i \hat j}}\int\nolimits_{t - {d_{\hat j}}}^{t - {d_{\hat i}}} {x(s)\mathrm{d}s} } } + {\xi ^T}(t)P\xi (t) = {\xi ^T}(t)\Big( {\Big[ {\begin{array}{*{20}{c}} {{\Sigma _{11}}}&\vartheta \\ * &\Xi \end{array}} \Big] + P} \Big)\xi (t), \end{align*} (19)

    where

    \begin{array}{*{20}{c}} \Sigma _{11} = 2\sum\limits_{j = 1}^N {{d_j}{R_j}}+ 2\sum\limits_{\hat i = 1}^{N - 1} {\sum\limits_{\hat j = \hat i+1}^N {{d_{\hat j\hat i}}} {Z_{\hat i\hat j}}},\\ \vartheta = \Big[ {{\Sigma _{12}}}\; {{\Sigma _{13}}}\; \cdots \; {{\Sigma _{1(N + 1)}}}\; {\Sigma _1^{12}}\; {\Sigma _1^{13}} \; \cdots\; {\Sigma _1^{(N - 1)N}} \Big], \\ \Xi = {\rm{diag}}\Big\{ {{\Sigma _{22}}\; \cdots\; {\Sigma _{(N+1)(N+1)}}\; \Sigma _{12}^{12}\; \Sigma _{13}^{13}\; \cdots\; \Sigma _{(N-1)N}^{(N-1)N}} \Big\}, \end{array}

    with

    \begin{align*} {\Sigma _{1(j + 1)}} = -2{R_j},\; \Sigma _1^{\hat i \hat j} = -2{Z_{\hat i\hat j}},\; {\Sigma _{(j+1)(j+1)}} = {\bar d_j}\left( {2{R_j}+{S_j}}\right),\\ \Sigma _{\hat i\hat j}^{\hat i\hat j} = {Q_{\hat i\hat j}} + {\bar d_{\hat i\hat j}}{Z_{\hat i\hat j}},\; j \in {{\bf{I}}}[1,N],\; \hat i \in {{\bf{I}}}[1,N - 1],\; \hat j \in {{\bf{I}}}[\hat i + 1,N]. \end{align*}

    Assume that the following matrix inequalities hold:

    \begin{align*} F_l^T{F_l} \leqslant \left[ {\begin{array}{*{20}{c}} {{\Sigma _{11}}}&\vartheta \\ * &\Xi \end{array}} \right] + P,l \in {{\bf{I}}}[1,\mathop m\limits^ \leftrightarrow], \end{align*} (20)

    where F = \big[ U\; {V_1}\; \cdots \; {V_N}\; {V_{12}}\; \cdots \; {V_{1N}}\; {V_{23}}\; \cdots \; {V_{2N}}\; \cdots\; V_{(N - 1)N} \big] , and {F_l} is the l{\rm{ - th}} row of matrix F , l \in {{\bf{I}}}[1, \overset{\lower0.5em\hbox{ $\smash{ \leftrightarrow}$ }} {m} ] . From (18)-(20), it can be seen that

    \begin{align*} \left| {{w_l}(t)} \right|^2 \leqslant V(t) \leqslant V(0),l \in {{\bf{I}}}[1, \mathop m\limits^ \leftrightarrow ], \end{align*} (21)

    where {V_{(\hat i \hat j)l}} is the l{\rm{-th}} row of matrix {V_{\hat i \hat j}} . For any \varphi (t) \in {\mathbb{C}_{n, d}} satisfying V(0) \leqslant 1 , it can be seen from (5) and (21) that \left| {{w_l}(t)} \right| \leqslant 1 holds for l \in {{\bf{I}}}[1, \overset{\lower0.5em\hbox{ $\smash{ \leftrightarrow}$ }} {m} ] , which implies that the assumption (6) can be guaranteed. Then, it can be concluded that the closed-loop system (8) without uncertainties is locally asymptotically stable for any initial function \varphi (t) \in {\mathbb{C}_{n, d}} satisfying V(0) \leqslant 1 .

    To obtain LMI-based conditions, we set {T_2} \triangleq \delta {T_1} , \delta \ne 0 in (16) and introduce the following new matrix variables:

    \left\{ \begin{array}{l} T_1^{ - 1} \triangleq X,\; K{X^T} \triangleq Y,\; {\rm{ }}U{X^T} \triangleq G,\; {V_j}{X^T} \triangleq {H_j}, {\rm{ }}\tilde XP{{\tilde X}^T} \triangleq \bar P,\; X{R_j}{X^T} \triangleq {{\bar R}_j},\; {\rm{ }}X{S_j}{X^T} \triangleq {{\bar S}_j},\hfill \\ X{Q_{\hat i\hat j}}{X^T} \triangleq {{\bar Q}_{\hat i \hat j}},\; {\rm{ }}X{Z_{\hat i\hat j}}{X^T} \triangleq {{\bar Z}_{\hat i\hat j}},\; {V_{\hat i\hat j}}{X^T} \triangleq {H_{\hat i\hat j}},\; {\rm{ }}\tilde X = {\rm{diag}}\big\{ {X, X, \cdots , X} \big\}. \hfill \\ \end{array} \right. (22)

    Performing some congruence transformations as in [42] and noting (22), one can ascertain that \Omega^k < 0 is equivalent to linear matrix inequality (11), and nonlinear matrix inequality (20) is equivalent to linear matrix inequality (12). This completes the proof.

    Remark 3. In this paper, we are mainly concerned with the systems with multiple delays. In the L-K functional (10), N is required to be greater than 1. For the case that N = 1 , the system (1) becomes the case with a single delay, which has been investigated in some existing literature.

    Remark 4: For the case that d_1 < d_2 < d_3 < \cdots < d_N , we can choose the following L-K functional:

    V(t) = \varpi^T(t)P\varpi(t)+ \\ \sum\limits_{j = 1}^{N} \Big[\int_{t-d_{j}}^{t-d_{j-1}}x^T(s) S_{j} x(s)\mathrm{d}s +(d_{j}-d_{j-1})\int_{-d_{j}}^{-d_{j-1}} \int_{t + \theta }^t {{\dot x}^T}(s) R_{j}\dot x(s)\mathrm{d}s\mathrm{d}\theta \Big], (23)

    where

    \begin{align*} \varpi(t) \buildrel \Delta \over = \bigg[ x^{T}(t)\; \int_{t-{d_1}}^t {x^T}(s) \mathrm{d}s\; \int_{t-{d_2}}^{t-{d_1}} {x^T}(s) \mathrm{d}s\; \int_{t-{d_3}}^{t-{d_2}} {x^T}(s) \mathrm{d}s \; \cdots\; \int_{t-{d_{N}}}^{t-{d_{N-1}}} {x^T}(s) \mathrm{d}s \bigg]^T, \end{align*}

    and P > 0 , S_j > 0 , R_j > 0 , d_0 = 0 ( j \in {{\bf{I}}}[1, N] ).

    Correspondingly, under the constraint

    \begin{align*} ||Ux(t) + \sum\limits_{j = 1}^N V_j \int\nolimits_{t - {d_j}}^{t-d_{j-1}} {x(s)\mathrm{d}s} ||_\infty \leqslant 1, \end{align*} (24)

    we have the following closed-loop system:

    \dot x(t) = \sum\limits_{k = 1}^{2^m} {\lambda _k^t}\bigg\{ \left[{A_0}(t) + B(t)({D_k}K + \mathcal{D}_k^ - U)\right]x(t) + \sum\limits_{j = 1}^N {{A_j}(t)x(t - {d_j})} \\ +\sum\limits_{j = 1}^{N} B(t) {\cal D}_k^ - V_{j} \int_{t-d_{j}}^{t-d_{j-1}} {x(s)\mathrm{d}s} \bigg\} , (25)

    where \lambda _1^t \ge 0 , \lambda _2^t \ge 0 , \cdots , \lambda _{{2^m}}^t \ge 0 and \sum_{k = 1}^{{2^m}} {\lambda _k^t} = 1 . Follow the derivation steps in Theorem 1, the corresponding results can be readily obtained.

    Remark 5. First, different from the L-K functionals in [10,15], some integral terms are introduced in V_1(t) . Second, by considering the relationships between multiple delays {d_j}\; (j \in {{\bf{I}}}[1, N]) , this paper introduces the term V_3(t) in the L-K functional (10). In addition, the Theorem 1 is slacker since the variables H_j\; (j \in {{\bf{I}}}[1, N]) and H_{\hat i \hat j}\; (\hat i \in {{\bf{I}}}[1, N - 1], \hat j \in {{\bf{I}}}[\hat i + 1, N]) are additionally introduced in the conditions (11) and (12). The matrices H_j and H_{\hat i \hat j} can be seen as slack variables in this paper. These new techniques can result in less conservative stabilization criteria.

    Remark 6. In proving Theorem 1, the stability criterion of the open-loop system can be represented as \tilde \Omega ^k < 0 , where \tilde \Omega ^k is obtained by setting U = 0 , K = 0 , V_j = 0 , and V_{\hat i \hat j} = 0 in \Omega ^k as defined below in (15). For convenience of comparison, the stability criterion is referred to as Corollary 1.

    Now, we consider system (1) with F(t) \ne 0 .

    Theorem 2. Let the constants \delta \ne 0 and {d_j} , j \in {{\bf{I}}}[1, N] be given. The system (1) can be robustly stabilized by the controller (4) with K = Y{X^{ - T}} if for any initial function \varphi (t) \in {\mathbb{C}_{n, d}} satisfying V(0) \leqslant 1, there exist {N_1} \times {N_1} matrix \bar P > 0 , n \times n matrices {\bar R_j} > 0 , {\bar S_j} > 0 , {\bar Q_{\hat i \hat j}} > 0 , {\bar Z_{\hat i\hat j}} > 0 , X , m \times n matrix Y , \overset{\lower0.5em\hbox{ $\smash{ \leftrightarrow}$ }} {m} \times n matrices G , {H_j} , j \in {{\bf{I}}}[1, N] , {H_{\hat i \hat j}} , \hat i \in {{\bf{I}}}[1, N - 1] , \hat j \in {{\bf{I}}}[\hat i + 1, N] , and a scalar \varepsilon > 0 such that for \forall k \in {{\bf{I}}}[1, {2^m}] and \forall l \in {{\bf{I}}}[1, \overset{\lower0.5em\hbox{ $\smash{ \leftrightarrow}$ }} {m} ] , the LMI (12) and the following LMI hold:

    \begin{align*} \left( {\begin{array}{*{20}{c}} {{\bar \Omega }^k}&{\varepsilon \bar M}&{{({{\bar E}^k})}^T} \\ * &{ - \varepsilon I}&0 \\ * & * &{ - \varepsilon I} \end{array}} \right) < 0, \end{align*} (26)

    where {\bar \Omega ^k} are defined in Theorem 1 and

    \begin{array}{*{20}{c}} \bar M = \big[ M^T\; 0_{n\times \tfrac{N(N +3)}{2}n}\; \delta M^T \big]^T,\\ \bar E^k = \big[ \bar \nu _0^k\; {E_1}{X^T}\; \cdots \; {E_N}{X^T}\; \bar \nu _1^k\; \cdots\; \bar \nu _N^k\; \bar \nu _12^k\; \cdots\; \bar \nu _1N^k\; \bar \nu _23^k\; \cdots\; \bar \nu _2N^k\; \cdots \; \bar \nu _{(N - 1)N}^k\; 0 \big], \end{array}

    with

    \begin{array}{*{20}{c}} \bar \nu _0^k = {E_0}{X^T}+E_{N + 1}({D_k}Y+\mathcal{D}_k^-G),\; \bar\nu _j^k = E_{N+1}\mathcal{D}_k^-{H_j},\; j\in {\bf{I}}[1,N], \\ \bar\nu_{\hat i\hat j}^k = E_{N + 1}\mathcal{D}_k^-{H_{\hat i\hat j}},\; \hat i\in {\bf{I}}[1,N - 1],\; \hat j \in {\bf{I}}[\hat i+1,N]. \end{array}

    Proof. Denote

    \begin{align*} \Delta {\Omega ^k} \triangleq \left[ {\begin{array}{*{20}{c}} {\Delta \Omega ^k_{11}}&{\Delta \Omega ^k_{12}}&{\Delta \Omega ^k_{13}}&{\Delta \Omega ^k_{14}}&{\Delta \Omega ^k_{15}} \\ * &0&0&0&{\Delta \Omega ^k_{25}} \\ * & * &0&0&{\Delta \Omega ^k_{35}} \\ * & * & * &0&{\Delta \Omega ^k_{45}} \\ * & * & * & * &0 \end{array}} \right], \end{align*} (27)

    where

    \begin{array}{*{20}{c}} \Delta \Omega ^k_{11} = {\rm {Sym}}\big[ {{T_1}\Delta {A_0} + {T_1}\Delta B({D_k}K + \mathcal{D}_k^ - U)} \big],\\ \Delta \Omega ^k_{12} = \left[ {{T_1}\Delta {A_1}}\; {{T_2}\Delta {A_2}}\; \cdots \; {{T_2}\Delta {A_N}} \right],\; \Delta \Omega ^k_{13} = \left[ {\Delta \beta _1^k}\; {\Delta \beta _2^k}\; \cdots \; {\Delta \beta _N^k} \right],\\ \Delta \Omega ^k_{14} = \big[ {\Delta \gamma _{12}^k}\; {\Delta \gamma _{13}^k}\; \cdots \; {\Delta \gamma _{1N}^k}\; {\Delta \gamma _{23}^k} \; \cdots\; {\Delta \gamma _{2N}^k}\; \cdots \; {\Delta \gamma _{(N - 1)N}^k} \big],\\ \Delta \Omega ^k_{15} = \Delta A_0^TT_2^T + {\left[\Delta B({D_k}K + \mathcal{D}_k^ - U)\right]^T}T_2^T,\\ \Delta \Omega ^k_{25} = {\big[ {\Delta {A_1}T_2^T}\; {\Delta {A_2}T_2^T}\; \cdots \; {\Delta {A_N}T_2^T} \big]^T},\; \Delta \Omega ^k_{35} = {\left[ {\Delta \sigma _1^k}\; {\Delta \sigma _2^k}\; \cdots \; {\Delta \sigma _N^k} \right]^T},\\ \Delta \Omega ^k_{45} = \big[ {\Delta \mu_{12}^k}\; {\Delta \mu _{13}^k}\; \cdots \; {\Delta \mu _{1N}^k}\; {\Delta \mu _{23}^k} \; \cdots \; {\Delta \mu _{2N}^k}\; \cdots \; {\Delta \mu _{(N - 1)N}^k} \big]^T, \end{array}

    with

    \begin{align*} \Delta \beta _j^k = & {T_1}\Delta B\mathcal{D}_k^ - {V_j},\; \Delta \sigma _j^k = T_2^{}\Delta B\mathcal{D}_k^ - {V_j}, \Delta \gamma _{\hat i\hat j}^k = {T_1}\Delta B\mathcal{D}_k^ - {V_{\hat i\hat j}},\\\Delta \mu _{\hat i\hat j}^k = & {T_2}\Delta B\mathcal{D}_k^ - {V_{\hat i \hat j}},\; j \in {{\bf{I}}}[1,N],\; \hat i \in {{\bf{I}}}[1,N - 1],\; \hat j \in {{\bf{I}}}[\hat i + 1,N]. \end{align*}

    Here, we choose the same L-K functional (10); then, the LMI (12) and the following LMI hold:

    \begin{align*} {\Omega ^k}{\rm{ + }}\Delta {\Omega ^k} < 0. \end{align*} (28)

    Using the assumptions (3) and {T_2} \triangleq \delta {T_1} , \delta \ne 0 , it is easy to obtain that

    \begin{align*} \Delta {\Omega ^k}{\rm{ = }}{T_1}\bar MF(t){E^k} + {({E^k})^T}{F^T}(t){\bar M^T}T_1^T, \end{align*} (29)

    where

    \begin{align*} {E^k} = & \big[ {\nu _0^k}\; {{E_1}}\; \cdots \; {{E_N}}\; {\nu _1^k}\; \cdots \; {\nu _N^k}\; {\nu _{12}^k}\; \cdots\; {\nu _{1N}^k}\; {\nu _{23}^k}\; \cdots \; {\nu _{2N}^k}\; \cdots \; {\nu _{(N - 1)N}^k}\; 0 \big], \end{align*}

    with

    \nu _{0}^k = {E_{0}}+{E_B}({D}_kK+\mathcal{D}_k^ - U),\; \nu _{j}^k = {E_B}\mathcal{D}_k^ - V_j,\; \nu _{\hat i \hat j}^k \\ = {E_{N + 1}}\mathcal{D}_k^ - {V_{\hat i\hat j}},\; j \in {{\bf{I}}}[1,N],\; \hat i \in {{\bf{I}}}[1,N - 1], \hat j \in {{\bf{I}}}[\hat i + 1,N].

    From (28) and (29), one can obtain that

    \begin{align*} {\Omega ^k}+{T_1}\bar MF(t){E^k} + {({E^k})^T}{F^T}(t){\bar M^T}T_1^T < 0. \end{align*} (30)

    Performing congruence transformation, one can ascertain that the nonlinear matrix inequality (30) is equivalent to the following matrix inequality:

    \begin{align*} {\bar \Omega ^k}+\bar MF(t){\bar E^k} + {({\bar E^k})^T}{F^T}(t){\bar M^T} < 0. \end{align*} (31)

    Using the well-known Schur complement, it is clear that (31) is equivalent to the LMI (26). The proof is completed.

    Remark 7. Note that the Wirtinger integral inequality is a special case of the Bessel-Legendre inequality [51]. Therefore, the obtained conditions in this paper remain conservative to a certain extent. By modifying the sector condition and using the Bessel-Legendre inequality, some more effective conditions are expected to be obtained.

    To obtain a larger estimate of the domain of attraction {X_\rho } when designing a controller, we discuss the estimate and maximization of the domain of attraction. Noting the L-K functional (10) and (22) and using the Jensen integral inequality to estimate the first term {\xi ^T}(0)P\xi (0) of V(0) , it can be seen that the domain of attraction {X_\rho } can be bounded by the following inequality:

    \begin{align*} V(0) \leqslant&\varPsi_1\rho_1^2+\varPsi_2\rho_2^2, \end{align*} (32)

    where

    \begin{array}{*{20}{c}} \varPsi_1\buildrel \Delta \over = \lambda_M(\tilde \Lambda_0) + \sum\limits_{j = 1}^N{d_j} \Big[ \lambda_M(d_j{\tilde \Lambda }_j+X^{-1}{\bar S}_j{X^{-T}}) \Big] +\sum\limits_{\hat i = 1}^{N- 1} \sum\limits_{\hat j = \hat i + 1}^N d_{\hat j \hat i}^2 \Big[\lambda _M({\tilde \Lambda }_{\hat i\hat j})+\lambda_M(X^{-1}{\bar Q}_{\hat i\hat j }{X^{-T}})\Big], \\ \varPsi_2\buildrel \Delta \over = (1/2)\Big[ \sum\limits_{j = 1}^N d_j^3\lambda_M(X^{-1}{\bar R_j}X^{-T}) +\sum\limits_{\hat i = 1}^{N- 1} \sum\limits_{\hat j = \hat i + 1}^N d_{\hat j \hat i}^2 {({d_{\hat j}} + {d_{\hat i}}){\lambda _M}({X^{ - 1}}{{\bar Z}_{\hat i\hat j }}{X^{ - T}})}\Big], \end{array}

    with {\tilde \Lambda _{\tilde j}} = {X^{ - 1}}{\Lambda _{\tilde j}}{X^{ - T}} , \tilde j \in {{\bf{I}}}[0, N] , {\tilde \Lambda _{\hat i \hat j}} = {X^{ - 1}}{\Lambda _{\hat i \hat j}}{X^{ - T}} , \hat i \in {{\bf{I}}}[1, N - 1] , \hat j \in {{\bf{I}}}[\hat i + 1, N] , and the LMI constraint \bar P \leqslant {\rm{diag}}\{ {\Lambda _0}\; {\Lambda _1}\; \cdots \; {\Lambda _N}\; {\Lambda _{12}}\; \cdots\; {\Lambda _{1N}}\; {\Lambda _{23}}\cdots {\Lambda _{2N}} \; \cdots \; {\Lambda _{(N - 1)N}}\} \triangleq \Lambda.

    As in [42], we introduce the matrix inequality {X^{ - 1}}{X^{ - T}} \leqslant rI , which can be guaranteed by the LMI

    \begin{align*} \left( {\begin{array}{*{20}{c}} {rI}&I \\ I&{X + {X^T} - I} \end{array}} \right) \geqslant 0. \end{align*} (33)

    Meanwhile, we define the following LMIs:

    \left\{ \begin{array}{l} {\Lambda _{\tilde j}} \leqslant {p_{\tilde j}}I,\; \tilde j \in {{\bf{I}}}[0,N],\; \bar P \leqslant \Lambda,\; {{\bar R}_j} \leqslant {r_j}I,\; {{\bar S}_j} \leqslant {s_j}I,\; j \in {{\bf{I}}}[1,N],\hfill \\ {\Lambda _{\hat i\hat j}} \leqslant {p_{\hat i\hat j}}I,\; {{\bar Q}_{\hat i\hat j}} \leqslant {q_{\hat i\hat j}}I,\; {{\bar Z}_{\hat i\hat j}} \leqslant {z_{\hat i\hat j}}I, \hat i \in {{\bf{I}}}[1,N - 1],\; \hat j \in {{\bf{I}}}[\hat i + 1,N]. \hfill \\ \end{array} \right. (34)

    Then, it can be seen from (32) that the maximization of the estimate of the domain of attraction {X_\rho } for system (1) without uncertainties in Theorem 1 can be formulated as the following optimization problem:

    Prob. 1.

    \begin{align*} \begin{gathered} \mathop {\min }\limits_{\bar P,\; {{\bar R}_j},\; {{\bar S}_j},\; {{\bar Q}_{\hat i\hat j}},\; {{\bar Z}_{\hat i\hat j}},\; {\Lambda _{\tilde j}},\; {\Lambda _{\hat i\hat j}},\; X,\; Y,\; G,\; {H_j},\; {H_{\hat i\hat j}},\; r,\; {p_{\tilde j}},\; {p_{\hat i\hat j}},\; {r_j},\; {s_j},\; {q_{\hat i\hat j}},\; {z_{\hat i\hat j}}}\lambda, s.t. \; {\rm{LMIs}}\; (11),(12), (33)\; {\rm{and}}\; (34) \; {\rm{hold}}. \hfill \end{gathered} \end{align*}

    Additionally, the maximization of the estimate of the domain of attraction {X_\rho } for system (1) in Theorem 2 can be formulated as the following optimization problem:

    Prob. 2.

    \begin{align*} \begin{gathered} \mathop {\min }\limits_{\bar P,\; {{\bar R}_j},\; {{\bar S}_j},\; {{\bar Q}_{\hat i\hat j}},\; {{\bar Z}_{\hat i\hat j}},\; {\Lambda _{\tilde j}},\; {\Lambda _{\hat i\hat j}},\; X,\; Y,\; G,\; {H_j},\; {H_{\hat i\hat j}},\; \varepsilon,\; r,\; {p_{\tilde j}},\; {p_{\hat i\hat j}},\; {r_j},\; {s_j},\; {q_{\hat i\hat j}},\; {z_{\hat i\hat j}}}\lambda , s.t. \; {\rm{LMIs}}\; (12),(26), (33) \; {\rm{and}}\; (34) \; {\rm{hold}}, \hfill \end{gathered} \end{align*}

    where \lambda = e*r+{p_0}+\sum_{j = 1}^N {{d_j}\big({s_j + {d_j}{p_j}+0.5d_j^2{r_j}}\big)} +\sum_{\hat i = 1}^{N-1} {\sum_{\hat j = \hat i+1}^N {d_{\hat i \hat j}^2\big[{{p_{\hat i\hat j}}+{q_{\hat i\hat j}}+0.5\big({d_{\hat j}}+{d_{\hat i}}\big){z_{\hat i\hat j}}}\big]}} and e is a weighting parameter.

    By solving the optimization problems Prob. 1 and Prob. 2, the values of \varPsi_1 and \varPsi_2 can be easily obtained. Then, we can determine the admissible bounds \rho_1 and \rho_2 of the set X_\rho by the equation \varPsi_1\rho_1^2+\varPsi_2\rho_2^2 = 1 . In particular, when {\rho _1} = {\rho _2} \triangleq \rho , the maximum admissible scalar \rho can be obtained by {\rho _{\max}}\triangleq1/\sqrt {\varPsi_1+\varPsi_2} . For any initial conditions \varphi (t) in the set X_\rho satisfying the equation \varPsi_1\rho_1^2+\varPsi_2\rho_2^2 = 1 , it can be seen that the constraint V(0)\leqslant1 can be guaranteed.

    Remark 8. In this section, by using the novel L-K functional (12) and the Wirtinger-based integral inequality, two sufficient conditions are established by LMIs under which the closed-loop system (6) is asymptotically stable for the case without uncertainties and robustly asymptotically stable for the case with uncertainties. In addition, the corresponding optimization problems are proposed to maximize the estimate of the domain of attraction.

    Example 1. [31,32] Consider the time-delay system (1), where N = 2 and

    \begin{align*} {A_0} = \left[ {\begin{array}{*{20}{c}} { - 2}&0 \\ 0&{ - 0.9} \end{array}} \right],\; {A_1} = \left[ {\begin{array}{*{20}{c}} { - 1}&{0.6} \\ { - 0.4}&{ - 1} \end{array}} \right],\; {A_2} = \left[ {\begin{array}{*{20}{c}} 0&{ - 0.6} \\ { - 0.6}&0 \end{array}} \right],\; B = \left[ {\begin{array}{*{20}{c}} {10} \\ {10} \end{array}} \right]. \end{align*} (35)

    First, we show the effectiveness of the proposed method. For the case {d_1} = 5 , {d_2} = 6 , it is clear from Figure 1 that the open-loop system (1) is not stable. By solving Prob.1 in this paper ( \delta = 1 , e = 5 \times {10^9} ), one can obtain the scalar {\rho _{\max }} = 32.2055 associated with the set {X_\rho } . Meanwhile, we have U = [-0.0032\; -0.0092] , {V_1} = [-0.0001845 \; -0.0002640] , {V_2} = [-0.000000009548 \; -0.000000009128] , {V_{12}} = [0.0040\; 0.0067] and the controller gain matrix K = [0.0078\; -0.0228] . In Figure 2, we plot the state responses of the closed-loop system and the signal w(t) , where \varphi (t) = [24\; 20]^T\in{X_\rho} . In Figure 3, we plot the state trajectories of the closed-loop systems. Figure 2 shows that the closed-loop system (1) is stable and that the functional w(t) defined in (5) satisfies the constraints {\left\| {w(t)} \right\|_\infty} < 1 . From Figure 3, it can be seen that the trajectories starting on the periphery of the ball never leave this ball and converge to the origin.

    Figure 1.  State responses of the open-loop system ( d_1 = 5 , d_2 = 6 ).
    Figure 2.  State responses of the closed-loop system and w(t) defined in (5).
    Figure 3.  System trajectories.

    By solving Prob.1 with \delta = 1 , e = 5*10^9 , we obtained the scalars \rho_{\max} related to the set {X_\rho } for different d_1 and d_2 , which are listed in Table 1. In Table 1, we also list the scalars \rho_{\max} obtained by Case 1 and Case 2, where Case 1 is obtained by setting {V_j} = 0 , {S_j} = 0 in LMIs (11) and (12), and Case 2 is obtained by setting {V_{\hat i\hat j}} = 0 , {Q_{\hat i\hat j}} = 0 and {Z_{\hat i\hat j}} = 0 in LMIs (11) and (12). From Table 1, it is clear that Prob. 1 can provide a larger estimate of the domain of attraction. Noting that Case 1 and Case 2 are based on the traditional techniques that address saturations and time delays, it can be concluded that our proposed result is effective in reducing the possible conservatism.

    Table 1.  The scalars {\rho _{\max }} are associated with the set {X_\rho } for the given {d_1} and {d_2} .
    ({d_1}, {d_2}) (1, 0.1) (1, 0.5) (1, 1.5) (1, 2) (1, 4)
    {\rho _{\max }} (Prob. 1) 161.9218 145.9009 116.4877 97.2339 75.8305
    {\rho _{\max }} (Case 1) 161.8783 141.6443 100.1767 83.0704 47.7115
    {\rho _{\max }} (Case 2) 159.4388 128.9824 99.8500 93.4002 75.7668
    ({d_1}, {d_2}) (2, 0.1) (2, 0.5) (2, 1.4) (2, 2.1) (2, 3.5)
    {\rho _{\max }} (Prob. 1) 89.1132 87.3734 81.7620 86.3634 47.9163
    {\rho _{\max }} (Case 1) 88.9580 86.6829 52.8286 54.5928
    {\rho _{\max }} (Case 2) 87.9240 77.7870 57.6496 48.0497 35.4435

     | Show Table
    DownLoad: CSV

    In Table 2, we list the admissible ranges of the time delay d_2 for different d_1 obtained by Corollary 1 in this paper and the results in [31,32]. It is clear from Table 2 that Corollary 1 can provide the larger range of the time delay d_2 , which shows that our proposed stability analysis approach is less conservative than that in [31,32].

    Table 2.  Admissible range of the time delay {d_2} to ensure stability of the open-loop system for a given {d_1} .
    {d_1} 2.3 2.4 2.5 3.0 3.5
    {d_2} ([31]) [0.08, 3.57] [0.22, 3.61] [0.35, 3.65] [1.04, 3.77] [1.88, 3.90]
    {d_2} ([32]) [0.07, 3.57] [0.17, 3.61] [0.28, 3.65] [0.80, 3.77] [1.60, 3.90]
    {d_2} (Corollary 1) [0, 4.54] [0, 4.54] [0, 4.55] [0, 4.69] [0, 4.77]

     | Show Table
    DownLoad: CSV

    Example 2. [52] Consider the uncertain system (1), where N = 2 and

    \begin{align*} {A_0} = & \left[ {\begin{array}{*{20}{c}} {-1}&2 \\ 0&{1} \end{array}} \right],\; {A_1} = \left[ {\begin{array}{*{20}{c}} {0.6}&{-0.4} \\ {0}&{0} \end{array}} \right],\; {A_2} = \left[ {\begin{array}{*{20}{c}} 0&{0} \\ {0}&-0.5 \end{array}} \right],\; B = \left[ {\begin{array}{*{20}{c}} {1} \\ {1} \end{array}} \right],\; M = I,\\ {E_0} = & \left[ {\begin{array}{*{20}{c}} {0.16}&0 \\ 0&{0.16} \end{array}} \right],\; {E_1} = \left[ {\begin{array}{*{20}{c}} {0}&{0} \\ {0}&{0.04} \end{array}} \right],\; {E_2} = \left[ {\begin{array}{*{20}{c}} {0.04}&0 \\ 0&{0} \end{array}} \right],\; {E_B} = \left[ {\begin{array}{*{20}{c}} 0 \\ 0 \end{array}} \right]. \end{align*} (36)

    This system can be seen as a linearized model of combustion in a liquid monopropellant rocket motor chambers[53]. For this example, by solving Prob. 2 in this paper with N = 2 , \delta = 0.7 , and e = 10 , we can obtain the scalars \rho _{\max } , which are listed in Table 3. Table 3 shows that the scalars \rho _{\max } decrease with the larger d_2 . By using the proposed controller ( K = [-0.9926 -2.9132] ), the state responses of the closed-loop system are plotted in Figure 4, where \varphi (t) = [0.4\; 0.1]^T . Figure 4 shows that the closed-loop system (1) is stable, which also shows the effectiveness of the proposed conditions.

    Table 3.  The scalars {\rho _{\max }} are associated with the set {X_\rho } for the given {d_1} and {d_2} .
    ({d_1}, {d_2}) (1, 0.1) (1, 0.5) (1, 1.5) (1, 2) (1, 4)
    {\rho _{\max }} (Prob. 2) 1.1163 1.0439 0.8584 0.7185 0.3630
    ({d_1}, {d_2}) (2, 0.1) (2, 0.5) (2, 1.4) (2, 2.1) (2, 3.5)
    {\rho _{\max }} (Prob. 2) 0.9910 0.9285 0.7103 0.8890 0.5009

     | Show Table
    DownLoad: CSV
    Figure 4.  State responses of the closed-loop system and w(t) defined in (5).

    Remark 9. From Examples 1 and 2, it is clear that our proposed results can provide a larger estimate of the initial condition set than some existing ones. However, it is worth mentioning that more free-weighting matrices are involved in our conditions, and the computational complexity is increased when solving the optimization problems.

    In this paper, some new delay-dependent local stabilization criteria have been obtained for multiple time-delay systems with actuator saturation. The saturation nonlinearity is represented as the convex combination of the state feedback, auxiliary distributed-delay feedback, and some cross-terms related to different time delays. Then, by combining the saturation with an augmented L-K functional and some integral inequalities (the Wirtinger integral inequality and the Jensen integral inequality), the stabilization and robust stabilization criteria have been proposed in terms of LMIs. Moreover, the estimation of the domain of attraction has been discussed. A numerical example is provided to show the values of the proposed results. Our proposed results can be readily extended to more general systems such as switched systems and T-S fuzzy systems [54].

    In addition, is should be pointed out that our proposed results in this paper are mainly concerned the case with multiple constant delays. For the case with multiple time-varying delays, the corresponding results can also be established, which is our further work.

    This work was supported in part by the National Natural Science Foundation of China under Grant 61773156 and Grant 72031009 and in part by the Science and Technology Plan Project of Henan Province 192400410046.

    The authors declare that there are no conflicts of interest.



    [1] Q. Gao, N. Olgac, Stability analysis for LTI systems with multiple time delays using the bounds of its imaginary spectra, Syst. Control Lett., 102 (2017), 112–118. https://doi.org/10.1016/j.sysconle.2017.02.003 doi: 10.1016/j.sysconle.2017.02.003
    [2] H. Wu, Eigenstructure assignment-based robust stability conditions for uncertain systems with multiple time-varying delays, Automatica, 33 (1997), 97–102. https://doi.org/10.1016/S0005-1098(96)00134-3 doi: 10.1016/S0005-1098(96)00134-3
    [3] F. Mazenc, M. Malisoff, S. I. Niculescu, Stability analysis for systems with time-varying delay: Trajectory based approach, 2015 54th IEEE Conference on Decision and Control (CDC), 2015, 1811–1816. https://doi.org/10.1109/CDC.2015.7402473
    [4] C. Wang, Q. Yang, T. Jiang, N. Li, Synchronization analysis of a class of neural networks with multiple time delays, J. Math., 2021 (2021), 5573619. https://doi.org/10.1155/2021/5573619 doi: 10.1155/2021/5573619
    [5] F. Milano, Small-signal stability analysis of large power systems with inclusion of multiple delays, IEEE Trans. Power Syst., 31 (2016), 3257–3266. https://doi.org/10.1109/TPWRS.2015.2472977 doi: 10.1109/TPWRS.2015.2472977
    [6] Y. Sun, Y. Wang, Z. Wei, G. Sun, X. Wu, Robust H_\infty load frequency control of multi-area power system with time delay: A sliding mode control approach, IEEE/CAA J. Autom. Sinica, 5 (2018), 610–617. https://doi.org/10.1109/JAS.2017.7510649 doi: 10.1109/JAS.2017.7510649
    [7] Y. Sun, N. Li, X. Zhao, Z. Wei, G. Sun, C. Huang, Robust H_\infty load frequency control of delayed multi-area power system with stochastic disturbances, Neurocomputing, 193 (2016), 58–67. https://doi.org/10.1016/j.neucom.2016.01.066 doi: 10.1016/j.neucom.2016.01.066
    [8] J. Li, Z. Chen, D. Cai, W. Zhen, Q. Huang, Delay-dependent stability control for power system with multiple time-delays, IEEE Trans. Power Syst., 31 (2016), 2316–2326. https://doi.org/10.1109/TPWRS.2015.2456037 doi: 10.1109/TPWRS.2015.2456037
    [9] D. Ding, Z. Wang, B. Bo, H. Shu, {H}_{\infty } state estimation for discrete-time complex networks with randomly occurring sensor saturations and randomly varying sensor delays, IEEE T. Neur. Net. Lear., 23 (2012), 725–736. https://doi.org/10.1109/TNNLS.2012.2187926 doi: 10.1109/TNNLS.2012.2187926
    [10] R. Zhang, D. Zeng, S. Zhong, Y. Yu, J. Cheng, Sampled-data synchronisation for memristive neural networks with multiple time-varying delays via extended convex combination method, IET Control Theory Appl., 12 (2018), 922–932. https://doi.org/10.1049/iet-cta.2017.1172 doi: 10.1049/iet-cta.2017.1172
    [11] Z. Wang, Y. Wang, Y. Liu, Global synchronization for discrete-time stochastic complex networks with randomly occurred nonlinearities and mixed time delays, IEEE T. Neural Networ., 21 (2010), 11–25. https://doi.org/10.1109/TNN.2009.2033599 doi: 10.1109/TNN.2009.2033599
    [12] Y. Dong, J. Xian, D. Han, New conditions for synchronization in complex networks with multiple time-varying delays, Commun. Nonlinear Sci. Numer. Simul., 18 (2013), 2581–2588. https://doi.org/10.1016/j.cnsns.2013.01.006 doi: 10.1016/j.cnsns.2013.01.006
    [13] Y. Wang, J. Cao, H. Wang, State estimation for markovian coupled neural networks with multiple time delays via event-triggered mechanism, Neural Process. Lett., 53 (2021), 893–906. https://doi.org/10.1007/s11063-020-10396-4 doi: 10.1007/s11063-020-10396-4
    [14] F. Zheng, Q. Wang, T. Lee, Adaptive robust control of uncertain time delay systems, Automatica, 41 (2005), 1375–1383. https://doi.org/10.1016/j.automatica.2005.03.014 doi: 10.1016/j.automatica.2005.03.014
    [15] C. Hua, G. Feng, X. Guang, Robust controller design of a class of nonlinear time delay systems via backstepping method, Automatica, 44 (2008), 567–573. https://doi.org/10.1016/j.automatica.2007.06.008 doi: 10.1016/j.automatica.2007.06.008
    [16] R. Dong, Y. Chen, W. Qian, An improved approach to robust {H}_{\infty } filtering for uncertain discrete-time systems with multiple delays, Circuits Syst. Signal Process., 39 (2020), 65–82. https://doi.org/10.1007/s00034-019-01162-6 doi: 10.1007/s00034-019-01162-6
    [17] F. Treviso, R. Trinchero, F. G. Canavero, Multiple delay identification in long interconnects via LS-SVM regressio, IEEE Access, 9 (2021), 39028–39042. https://doi.org/10.1109/ACCESS.2021.3063713 doi: 10.1109/ACCESS.2021.3063713
    [18] Y. Li, Y. Lu, Y. Wu, S. He, Robust cooperative control for micro/nano scale systems subject to time-varying delay and structured uncertainties, Int. J. Adv. Manuf. Technol., 105 (2019), 4863–4873. https://doi.org/10.1007/s00170-019-03832-w doi: 10.1007/s00170-019-03832-w
    [19] Y. Yan, J. Huang, Cooperative robust output regulation problem for discrete-time linear time-delay multi-agent systems via the distributed internal model, 2017 IEEE 56th Annual Conference on Decision and Control (CDC), 2017, 4680–4685. https://doi.org/10.1109/CDC.2017.8264350 doi: 10.1109/CDC.2017.8264350
    [20] Z. Zhao, W. Qian, X. Xu, Stability analysis for delayed neural networks based on a generalized free-weighting matrix integral inequality, Syst. Sci. Control Eng., 9 (2021), 6–13. https://doi.org/10.1080/21642583.2020.1858363 doi: 10.1080/21642583.2020.1858363
    [21] L. Zou, Z. Wang, H. Gao, X. Liu, State estimation for discrete-time dynamical networks with time-varying delays and stochastic disturbances under the Round-Robin protocol, IEEE T. Neur. Net. Lear., 28 (2017), 1139–1151. https://doi.org/10.1109/TNNLS.2016.2524621 doi: 10.1109/TNNLS.2016.2524621
    [22] J. Hu, H. Zhang, H. Liu, X. Yu, A survey on sliding mode control for networked control systems, Int. J. Syst. Sci., 52 (2021), 1129–1147. https://doi.org/10.1080/00207721.2021.1885082 doi: 10.1080/00207721.2021.1885082
    [23] L. Zou, Z. Wang, J. Hu, Y. Liu, X. Liu, Communication-protocol-based analysis and synthesis of networked systems: Progress, prospects and challenges, Int. J. Syst. Sci., 52 (2021), 3013–3034. https://doi.org/10.1080/00207721.2021.1917721 doi: 10.1080/00207721.2021.1917721
    [24] H. Liu, W. Qian, W. Xing, Z. Zhao, Further results on delay-dependent robust H_{\infty} control for uncertain systems with interval time-varying delays, Syst. Sci. Control Eng., 9 (2021), 30–40. https://doi.org/10.1080/21642583.2020.1833785 doi: 10.1080/21642583.2020.1833785
    [25] Y. Chen, K. Ma, R. Dong, Dynamic anti-windup design for linear systems with time-varying state delay and input saturations, Int. J. Syst. Sci., 53 (2022), 2165–2179. https://doi.org/10.1080/00207721.2022.2043483 doi: 10.1080/00207721.2022.2043483
    [26] L. Ma, Z. Wang, Y. Liu, F. E. Alsaadi, Distributed filtering for nonlinear time-delay systems over sensor networks subject to multiplicative link noises and switching topology, Int. J. Robust Nonlinear Control, 29 (2019), 2941–2959. https://doi.org/10.1002/rnc.4535 doi: 10.1002/rnc.4535
    [27] E. Xu, K. Ma, Y. Chen, H_{\infty} control for a hyperchaotic finance system with external disturbance based on the quadratic system theory, Syst. Sci. Control Eng., 9 (2021), 41–49. https://doi.org/10.1080/21642583.2020.1848658 doi: 10.1080/21642583.2020.1848658
    [28] H. Geng, H. Liu, L. Ma, X. Yi, Multi-sensor filtering fusion meets censored measurements under a constrained network environment: Advances challenges and prospects, Int. J. Syst. Sci., 52 (2021), 3410–3436. https://doi.org/10.1080/00207721.2021.2005178 doi: 10.1080/00207721.2021.2005178
    [29] W. Qian, W. Xing, S. Fei, H_\infty state estimation for neural networks with general activation function and mixed time-varying delays, IEEE T. Neur. Net. Lear., 32 (2021), 3909–3918. https://doi.org/10.1109/TNNLS.2020.3016120 doi: 10.1109/TNNLS.2020.3016120
    [30] E. Fridman, U. Shaked, Delay-dependent stability and {H}_{\infty } control: constant and time-varying delays, Int. J. Control, 76 (2003), 48–60. https://doi.org/10.1080/0020717021000049151 doi: 10.1080/0020717021000049151
    [31] Y. He, M. Wu, J. H. She, Delay-dependent stability criteria for linear systems with multiple time delays, IEE Proc., Control Theory Appl., 153 (2006), 447–452. http://dx.doi.org/10.1049/ip-cta:20045279 doi: 10.1049/ip-cta:20045279
    [32] J. Wang, L. Kong, Y. Chen, Further results on robust stability of uncertain linear systems with multiple time-varying delays, ICIC Express Lett., 9 (2015), 2879–2885.
    [33] T. Hu, Z. Lin, Control systems with actuator saturation: Analysis and design, Springer Science & Business Media, 2001. https://doi.org/10.1007/978-1-4612-0205-9
    [34] S. Tarbouriech, G. Garcia, J. M. G. da Silva Jr, I. Queinnec, Stability and stabilization of linear systems with saturating actuators, Springer London, 2011. https://doi.org/10.1007/978-0-85729-941-3
    [35] A. T. Fuller, In-the-large stability of relay and saturating control systems with linear controllers, Int. J. Control, 10 (1969), 457–480. https://doi.org/10.1080/00207176908905846 doi: 10.1080/00207176908905846
    [36] B. Zhou, Z. Lin, G. Duan, Robust global stabilization of linear systems with input saturation via gain scheduling, Int. J. Robust Nonlinear Control, 20 (2010), 424–447. https://doi.org/10.1002/rnc.1436 doi: 10.1002/rnc.1436
    [37] B. Zhou, G. Duan, Z. Lin, A parametric lyapunov equation approach to the design of low gain feedback, IEEE T. Automat Contr., 53 (2008), 1548–1554. https://doi.org/10.1109/TAC.2008.921036 doi: 10.1109/TAC.2008.921036
    [38] B. Zhou, Analysis and design of discrete-time linear systems with nested actuator saturations, Syst. Control Lett., 62 (2013), 871–879. https://doi.org/10.1016/j.sysconle.2013.06.012 doi: 10.1016/j.sysconle.2013.06.012
    [39] E. Fridman, A. Pila, U. Shaked, Regional stabilization and H_\infty control of time-delay systems with saturating actuators, Int. J. Robust Nonlinear Control, 13 (2003), 885–907. https://doi.org/10.1002/rnc.852 doi: 10.1002/rnc.852
    [40] L. Zhang, E. K. Boukas, A. Haidar, Delay-range-dependent control synthesis for time-delay systems with actuator saturation, Automatica, 44 (2008), 2691–2695. https://doi.org/10.1016/j.automatica.2008.03.009 doi: 10.1016/j.automatica.2008.03.009
    [41] H. He, X. Gao, W. Qi, Asynchronous H_\infty control of time-delayed switched systems with actuator saturation via anti-windup design, Optim. Control. Appl. Methods, 39 (2018), 1–18. https://doi.org/10.1002/oca.2330 doi: 10.1002/oca.2330
    [42] Y. Chen, S. Fei, Y. Li, Robust stabilization for uncertain saturated time-delay systems: A distributed-delay-dependent polytopic approach, IEEE T. Automat. Contr., 62 (2017), 3455–3460. https://doi.org/10.1109/TAC.2016.2611559 doi: 10.1109/TAC.2016.2611559
    [43] Y. Chen, Z. Wang, S. Fei, Q. L. Han, Regional stabilization for discrete time-delay systems with actuator saturations via a delay-dependent polytopic approach, IEEE T. Automat. Contr., 64 (2019), 1257–1264. https://doi.org/10.1109/TAC.2018.2847903 doi: 10.1109/TAC.2018.2847903
    [44] Y. Chen, Z. Wang, B. Shen, H. Dong, Exponential synchronization for delayed dynamical networks via intermittent control: Dealing with actuator saturations, IEEE T. Neur. Net. Lear., 30 (2019), 1000–1012. https://doi.org/10.1109/tnnls.2018.2854841 doi: 10.1109/tnnls.2018.2854841
    [45] Y. Chen, S. Fei, K. Zhang, Stabilization of impulsive switched linear systems with saturated control input, Nonlinear Dyn., 69 (2012), 793–804. https://doi.org/10.1007/s11071-011-0305-y doi: 10.1007/s11071-011-0305-y
    [46] L. Ma, Z. Wang, Y. Chen, X. Yi, Probability-guaranteed distributed filtering for nonlinear systems with innovation constraints over sensor networks, IEEE T. Control. Netw., 8 (2021), 951–963. https://doi.org/10.1109/TCNS.2021.3049361 doi: 10.1109/TCNS.2021.3049361
    [47] Y. Chen, S. Fei, Y. Li, Stabilization of neutral time-delay systems with actuator saturation via auxiliary time-delay feedback, Automatica, 52 (2015), 242–247. https://doi.org/10.1016/j.automatica.2014.11.015 doi: 10.1016/j.automatica.2014.11.015
    [48] M. Basin, J. Rodriguez-Gonzalez, L. Fridman, Optimal and robust control for linear state-delay systems, J. Franklin Inst., 344 (2007), 830–845. https://doi.org/10.1016/j.jfranklin.2006.10.002 doi: 10.1016/j.jfranklin.2006.10.002
    [49] N. M. Dmitruk, Optimal robust control of constrained linear time-delay systems, IFAC Proc. Vol., 40 (2007), 168–173. https://doi.org/10.1016/S1474-6670(17)69282-6 doi: 10.1016/S1474-6670(17)69282-6
    [50] A. Seuret, F. Gouaisbaut, Wirtinger-based integral inequality: Application to time-delay systems, Automatica, 49 (2013), 2860–2866. https://doi.org/10.1016/j.automatica.2013.05.030 doi: 10.1016/j.automatica.2013.05.030
    [51] A. Seuret, F. Gouaisbaut, Hierarchy of LMI conditions for the stability analysis of time-delay systems, Syst. Control Lett., 81 (2015), 1–7. https://doi.org/10.1016/j.sysconle.2015.03.007 doi: 10.1016/j.sysconle.2015.03.007
    [52] M. Wu, Y. He, J. H. She, Delay-dependent stabilization for systems with multiple unknown time-varying delays, Int. J. Control Autom. Syst., 4 (2006), 682–688.
    [53] L. Xie, E. Fridman, U. Shaked, Robust H_\infty control of distributed delay systems with application to combustion control, IEEE T. Automat Contr., 46 (2001), 1930–1935. https://doi.org/10.1109/9.975483 doi: 10.1109/9.975483
    [54] Z. Gu, P. Shi, D. Yue, Z. Ding, Decentralized adaptive event-triggered H_\infty filtering for a class of networked nonlinear interconnected systems, IEEE T. Cybernetics, 49 (2019), 1570–1579. https://doi.org/10.1109/TCYB.2018.2802044 doi: 10.1109/TCYB.2018.2802044
  • This article has been cited by:

    1. Xiaoyu Sun, Huabo Liu, Bin Wang, Robust stability analysis of large-scale interconnected systems with time-varying delays, 2025, 10, 2473-6988, 9674, 10.3934/math.2025445
  • Reader Comments
  • © 2022 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(2019) PDF downloads(62) Cited by(1)

Figures and Tables

Figures(4)  /  Tables(3)

Other Articles By Authors

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog