For a class of linear time-invariant neutral systems with neutral and discrete constant delays, several existing asymptotic stability criteria in the form of linear matrix inequalities (LMIs) are simplified by using matrix analysis techniques. Compared with the original stability criteria, the simplified ones include fewer LMI variables, which can obviously reduce computational complexity. Simultaneously, it is theoretically shown that the simplified stability criteria and original ones are equivalent; that is, they have the same conservativeness. Finally, a numerical example is employed to verify the theoretic results investigated in this paper.
1. Introduction
Some practical systems, such as population ecology, neural networks, heat exchangers, and robots in contract with rigid environments, have been modeled by functional differential equations of neutral type (e.g., [1–6]). Since stability analysis is the primary task of analyzing and synthesizing a system, many researchers have paid more and more attention to establish stability criteria for delayed neutral systems (see [7, 8] and the references therein).
For a class of linear time-invariant neutral systems with neutral and discrete constant delays, both neutral- and discrete-delay-dependent stability criteria have been investigated in [9–27]. Many approaches have been provided for obtaining these stability criteria. Here we mention only some critical and representative approaches. Model transformation approach first transforms the neutral system with discrete delay into the one with distributed delays, and then delay-dependent stability criteria can be obtained by constructing a Lyapunov-Krasovskii functional [9]. However, model transformation approach introduces some additional dynamics, which leads still to conservative results [28]. To overcome this shortage, the so-called descriptor model transformation approach is introduced in [15, 24], which transforms the original system to an equivalent descriptor form. The descriptor model transformation approach will not introduce additional dynamics in the sense defined in [28]. Although the results obtained from descriptor model transformation approach may be less conservative than some existing ones, they can be improved by employing some other approaches (e.g., free weighting matrix approach [21, 22] and augmented Lyapunov-Krasovskii functional approach [14, 23, 27]) to get a larger bound for discrete delay.
By employing free weighting matrices and the Leibniz-Newton formula, He et al. [21] presented a neutral- and discrete-delay-dependent stability criterion for neutral systems with neutral and discrete constant delays, which reduces the conservativeness of methods involving a fixed model transformation. For the same system models, Qian et al. [27] established neutral- and discrete-delay-dependent stability criteria by constructing a Lyapunov-Krasovskii functional with additional functional parameters and employing free weighting matrices. To the best of our knowledge, the large-scale employment of functional parameters and/or free weighting matrices can cause complex stability criteria, and thereby the computational complexity increases obviously. Therefore, in order to reduce computational complexity, it is necessary to reduce the number of functional parameters and free weighting matrices. This motivates the present study.
The aim of this paper is to simplify several stability criteria proposed by He et al. [21] and Qian et al. [27]. The main contributions of the paper are as follows. (i) In the premise of not increasing conservativeness, a pair of stability criteria proposed in [27, Theorem 1 and Corollary 1] are simplified by using matrix analysis techniques; (ii) one of the pair of simplified stability criteria is theoretically presented to be less conservative than [26, Corollary 1], and another is theoretically proven to be equivalent to [21, Theorem 1]; and (iii) by one numerical example, our theoretic results are verified to be more effective than many existing results.
2. Problem Statement and Preliminary Results
Consider a class of delayed neutral systems described as(1a)x˙(t)-Cx˙(t-η)=Ax(t)+A1x(t-τ),t≥0,(1b)x(t)=ϕ(t),t∈[-ω,0],
where x(t) is the n-dimensional state vector, A, A1, and C are known real constant matrices with appropriate dimensions, ω=max{η,τ}, ϕ(s) is an ℝn-valued continuous initial function defined on [-ω,0], and positive scalars η and τ represent the neutral and discrete delays, respectively.
Define C([-η,0],ℝn) as the set of continuous ℝn valued function on the interval [-η,0], and let xt∈C([-η,0],ℝn) be a segment of system trajectory defined as
(2)xt(θ)=x(t+θ),θ∈[-η,0].
Define an operator 𝒟:C([-η,0],ℝn)→ℝn by 𝒟xt=x(t)-Cx(t-η). The definitions on stability of operator 𝒟 and systems (1a) and (1b) can be seen in [29].
As mentioned in Introduction, neutral- and discrete-delay-dependent stability criteria for systems (1a) and (1b) have been investigated by He et al. [21] and Qian et al. [27]. In this paper, we will simplify the stability criteria [21, Theorem 1] and [27, Theorem 1 and Corollary 1] in the premise of not increasing conservativeness of stability criteria.
In order to present conveniently our main results, the following lemmas are required.
Lemma 1 (Schur complement lemma [30]).
Given constant matrices Ω1, Ω2, and Ω3 of appropriate dimensions, where Ω1T=Ω1 and Ω2T=Ω2, then
(3)Ω1+Ω3TΩ2-1Ω3<0
if and only if
(4)[Ω1Ω3TΩ3-Ω2]<0or[-Ω2Ω3Ω3TΩ1]<0.
Lemma 2 (see [31]).
Given a real symmetric matrix Ψ and a pair of real matrices P and Q, the following LMI problem
(5)Ψ+PTXTQ+QTXP<0
is solvable with respect to decision variable X if and only if
(6)𝒩PTΨ𝒩P<0,𝒩QTΨ𝒩Q<0,
where 𝒩P and 𝒩Q are matrices whose columns form a basis of the right null spaces of P and Q, respectively.
3. Several Stability Criteria in the Literature
In this section we introduce several existing stability criteria for systems (1a) and (1b), which will be useful to present conveniently the main results of this paper in the next section.
Proposition 3 (see [21, Theorem 1]).
For given scalars τ>0 and η>0, systems (1a) and (1b) are asymptotically stable, if the operator 𝒟 is stable and there exist real n×n matrices PT=P>0, QiT=Qi>0 (i=1,2), RT=R>0, XˇiiT=Xˇii, Yˇii=Yˇii (i=1,…,5), Xˇij, and Yˇij (1≤i<j≤5) such that the following LMIs are feasible:
(7)[XˇXˇ5TXˇ5Xˇ55]≥0,[YˇYˇ5TYˇ5Yˇ55]≥0,(8)[Φˇ+τXˇ+ηYˇ+ΦˇX+ΦˇYAˇTSSAˇ-S]<0,
where
(9)Xˇ=[Xˇ11Xˇ12Xˇ13Xˇ14Xˇ12TXˇ22Xˇ23Xˇ24Xˇ13TXˇ23TXˇ33Xˇ34Xˇ14TXˇ24TXˇ34TXˇ44],Yˇ=[Yˇ11Yˇ12Yˇ13Yˇ14Yˇ12TYˇ22Yˇ23Yˇ24Yˇ13TYˇ23TYˇ33Yˇ34Yˇ14TYˇ24TYˇ34TYˇ44],Xˇ5=[Xˇ15TXˇ25TXˇ35TXˇ45T],Yˇ5=[Yˇ15TYˇ25TYˇ35TYˇ45T],ΦˇX=Xˇ5TΠ1+Π1TXˇ5,ΦˇY=Yˇ5TΠ2+Π2TYˇ5,Π1=[In-In00],Π2=[In0-In0],S=R+τXˇ55+ηYˇ55,Aˇ=[AA10C],Φˇ=[Φˇ11PA1-ATPC0A1TP-Q1-A1TPC0-CTPA-CTPA1-Q20000-R],Φˇ11=PA+ATP+Q1+Q2.
Proposition 4 (see [27, Theorem 1]).
For given scalars τ>0 and η>0, systems (1a) and (1b) are asymptotically stable, if the operator 𝒟 is stable and there exist real n×n matrices U, PT=P, WT=W, QkT=Qk, NkT=Nk, RkT=Rk, TkT=Tk, Mk, Zk(k=1,2), Ei, Fi, Hi, Yii (i=1,…,5), and Yij (1≤i<j≤5) such that the following LMIs are feasible:
(10)[R2Z2Z2TT2]≥0,(11)Qk≥0,k=1,2,(12)W>0,[R1Z1Z1TT1]>0,(13)[PM1M2M1Tτ-1Q1+N1UTM2TUη-1Q2+N2]>0,(14)Ξ:=[Φ+ηYτG1-τFτG1T-τR1-τZ1-τFT-τZ1T-τT1]<0,(15)Ω:=[Y-G2H-G2TR2Z2HTZ2TT2]≥0,
where
(16)Φ=[Φ11⋯Φ15⋯⋯⋯Φ15T⋯Φ55],Y=[Y11⋯Y15⋯⋯⋯Y15T⋯Y55],F=[F1⋮F5],H=[H1⋮H5],Φ11=PA+ATP+Q1+Q2+τR1+ηR2Φ11=+M1+M1T+M2+M2T+F1+F1TΦ11=+H1+H1T-E1A-ATE1T,Φ12=PA1-M1+H2T-ATE2T+F2T-F1-E1A1,Φ13=-ATPC-M1TC-M2TC-M2Φ13=-H1+H3T+F3T-ATE3T,Φ14=-ATE4T+E1+F4T+H4T+τZ1+ηZ2,Φ15=-ATE5T+F5T+H5T-E1C,Φ22=-Q1-E2A1-A1TE2T-F2-F2T,Φ23=-A1TPC+M1TC-H2-F3T-A1TE3T,Φ24=E2-A1TE4T-F4T,Φ25=-E2C-A1TE5T-F5T,Φ33=-Q2+CTM2+M2TC-H3-H3T,Φ34=E3-H4T,Φ35=-E3C-H5T,Φ44=E4+E4T+W+τT1+ηT2,Φ45=-E4C+E5T,Φ55=-W-E5C-CTE5T,G1=[N1+UT-N1-UTM1T-M1TC]T,G2=[N2+U-U-N2M2T-M2TC]T.
Proposition 5 (see [27, Corollary 1]).
When η=τ, for a given scalar τ>0, systems (1a) and (1b) are asymptotically stable, if the operator 𝒟 is stable and there exist real n×n matrices WT=W>0, Q1T=Q1>0, PT=P, N1T=N1, R1T=R1, T1T=T1, M1, Z1, Ei, and Fi (i=1,2,4,5) such that the following LMIs are feasible:
(17)[R1Z1Z1TT1]>0,[PM1M1Tτ-1Q1+N1]>0,(18)[Π11Π12Π13Π14τN1-τF1Π12TΠ22Π23Π24-τN1-τF2Π13TΠ23TΠ33Π34τM1-τF4Π14TΠ24TΠ34TΠ44-τCTM1-τF5τN1-τN1τM1T-M1TC-τR1-τZ1-τF1T-τF2T-τF4T-τF5T-τZ1T-τT1]<0,
where
(19)Π11=PA+ATP+M1+M1T+Q1+τR1Π11=+F1+F1T-E1A-ATE1T,Π12=PA1-ATPC-M1-M1TCΠ12=-ATE2T-E1A1+F2T-F1,Π13=-ATE4T+E1+F4T+τZ1,Π14=-ATE5T+F5T-E1C,Π22=-Q1-A1TPC-CTPA1+CTM1+M1TCΠ22=-E2A1-A1TE2T-F2-F2T,Π23=E2-A1TE4T-F4T,Π24=-E2C-A1TE5T-F5T,Π33=E4+E4T+τT1+W,Π34=-E4C+E5T,Π44=-E5C-CTE5T-W.
Remark 6.
Proposition 3 simplifies the notations in [21, Theorem 1].
Remark 7.
Propositions 4 and 5 correct some slips of the pen in [27, Theorem 1 and Corollary 1]. After [27, Theorem 1] is amended as in Remark 7, [27, Theorem 2] will be correct.
4. Simplified Stability Criteria
In this section we will simplify the stability criteria introduced in the previous section. Firstly, the stability criterion presented in Proposition 4 can be simplified by the following theorem.
Theorem 8.
For given scalars τ>0 and η>0, systems (1a) and (1b) are asymptotically stable, if the operator 𝒟 is stable and one of the following cases, (i)–(iv), holds.
There exist real n×n matrices U, PT=P, WT=W, QkT=Qk, NkT=Nk, RkT=Rk, TkT=Tk, Mk, Zk(k=1,2), Ei, Fi, Hi, Yii (i=1,…,5), and Yij (1≤i<j≤5) such that the LMIs (10)–(15) hold.
There exist real n×n matrices U, PT=P, WT=W, QkT=Qk, NkT=Nk, RkT=Rk, TkT=Tk, Mk, Zk(k=1,2), Ei, Fi, Hi, Yii (i=1,…,5), and Yij (1≤i<j≤5) such that the LMIs (11)–(15) and the following LMI (20) hold:
(20)[R2Z2Z2TT2]>0.
There exist real n×n matrices U, PT=P, WT=W, QkT=Qk, NkT=Nk, RkT=Rk, TkT=Tk, Mk, and Zk (k=1,2) such that the LMIs (11), (12), and (13) and the following LMI (21) hold:
(21)[ΦτG1-τF-ηG2ηHτG1T-τR1-τZ100-τFT-τZ1T-τT100-ηG2T00-ηR2-ηZ2ηHT00-ηZ2T-ηT2]<0.
There exist real n×n matrices U, PT=P, WT=W, QkT=Qk, NkT=Nk, RkT=Rk, TkT=Tk, Mk, and Zk (k=1,2) such that the LMIs (11), (12), (13), and (20) and the following LMI (22) hold:
(22)Σ~:=[Φ~G~1G~2G~1T-τR10G~2T0-ηR2]<0,
where
(23)Φ~=[Φ~11Φ~12Φ~13Φ~14Φ~12TΦ~22Φ~23A1TSCΦ~13TΦ~23TΦ~330Φ~14TCTSA10CTSC-W],Φ~11=PA+ATP+Q1+Q2+τR1+ηR2+M1+M1TΦ~11=+M2+M2T+ATSA+τATZ1T+τZ1AΦ~11=+ηATZ2T+ηZ2A-τ-1T1-η-1T2,Φ~12=PA1-M1+τZ1A1+ηZ2A1+ATSA1+τ-1T1,Φ~13=-ATPC-M1TC-M2TC-M2+η-1T2,Φ~14=τZ1C+ηZ2C+ATSC,Φ~22=-Q1-τ-1T1+A1TSA1,Φ~23=-A1TPC+M1TC,Φ~33=-Q2+CTM2+M2TC-η-1T2,S=W+τT1+ηT2,G~1=[τN1T+τU-Z1T+τATM1-τN1T+Z1T+τA1TM1-τU0],G~2=[-ηN2T-ηUT+Z2T-ηATM2ηUT-ηA1TM2ηN2T-Z2T0].
Proof.
Due to Proposition 4, it suffices to show that the conditions (i)–(iv) are equivalent.
(i)⇒(ii) Since the eigenvalues of a matrix are continuous functions of its elements, it follows from (i) that there exists a sufficiently small positive scalar ϵ such that
(24)Ξ+diag(ηϵI,0,0,ηϵI,0,0,0)<0,Ω+diag(0,ϵI,ϵI)≥0,[R2+ϵIZ2Z2TT2+ϵI]>0.
Therefore, the LMIs (11)–(15) and (20) are feasible.
(ii)⇒(i) The proof is very easy, and hence it is omitted.
(ii)⇒(iii) By Lemma 1, it follows from (20) and (15) that
(25)Y≥[-G2H][R2Z2Z2TT2]-1[-G2H]T.
This, together with (14) and Lemma 1, implies that (21) holds.
(iii)⇒(ii) It follows from (21) and Lemma 1 that (20) holds and
(26)[Φ+ηYτG1-τFτG1T-τR1-τZ1-τFT-τZ1T-τT1]<0,
where
(27)Y=[-G2H][R2Z2Z2TT2]-1[-G2H]T,
and hence (ii) holds.
(iii)⇔(iv) Let
(28)F~=[FHE],E=[E1TE2TE3TE4TE15]T,K=[e1Te2Te3Te4Te5T]T,L=[(e1-e2-τe7)T(e1-e3+ηe9)TA~T]T,A~=[-A-A10I-C0n×4n],ei=[0n×(i-1)nI0n×(9-i)n],i=1,2,…,9.
Then LMI (21) can be written as
(29)Φ1+KTF~L+LTF~TK<0,
where the matrix Φ1 is obtained from the matrix on the left of (21) by deleting all parts containing one of F, H, and E (e.g., the (5,5)th subblock of Φ1 equals -W, which is obtained from Φ55 by deleting -E5C-CTE5T). Choose
(30)𝒩K=[e6Te7Te8Te9T],𝒩L=[e1+ATe4+τ-1e7-η-1e9e2+A1Te4-τ-1e7e3+η-1e9CTe4+e5e6e8]T.
Noting that 𝒩KTΦ1𝒩K<0, one can derive from (12) and Lemma 2 that the LMI (29) (i.e., (21)) is feasible if and only if LMIs (22) and (20) are feasible.
The proof is completed.
Clearly, the stability condition (iv) in Theorem 8 is more simpler than (i) in Theorem 8 (i.e., Proposition 4). By a process similar to investigating Theorem 8, one can easily obtain the following theorem which simplifies the stability criterion presented in Proposition 5.
Theorem 9.
When η=τ, for a given scalar τ>0, systems (1a) and (1b) are asymptotically stable, if the operator 𝒟 is stable, and one of the following cases, (i)-(ii), holds.
There exist real n×n matrices WT=W>0, Q1T=Q1>0, PT=P, N1T=N1, R1T=R1, T1T=T1, M1, Z1, Ei, and Fi (i=1,2,4,5) such that the LMIs (17) and (18) hold.
There exist real n×n matrices WT=W>0, Q1T=Q1>0, PT=P, N1T=N1, R1T=R1, T1T=T1, M1, and Z1 such that (17) and the following LMI (31) hold:
(31)[Π~11Π~12Π~13Π~14Π~12TΠ~22Π~23Π~24Π~13TΠ~23TΠ~330Π~14TΠ~24T0-τR1]<0,
where
(32)Π~11=PA+ATP+M1+M1T+Q1+τR1Π~11=+τZ1A+τATZ1T+ATS0A-τ-1T1,Π~12=PA1-ATPC-M1-M1TCΠ~12=+τZ1A1+ATS0A1+τ-1T1,Π~13=τZ1C+ATS0C,Π~14=τN1+τATM1-Z1T,Π~22=-Q1-A1TPC-CTPA1+CTM1Π~22=+M1TC+A1TS0A1-τ-1T1,Π~23=A1TS0C,Π~24=-τN1+τA1TM1+Z1T,Π~33=CTS0C-W,S0=W+τT1.
Remark 10.
Theorems 8 and 9 indicate that, without increasing conservativeness of stability criteria, the free weighting matrices required in Propositions 4 and 5 (i.e., [27, Theorem 1 and Corollary 1]) can be eliminated. This will obviously reduce the computational complexity, and hence the stability conditions (iv) of Theorem 8 and (ii) of Theorem 9 are more effective.
Remark 11.
In the special case that M1=0, N1=0, Z1=0, and R1=ϵIn (ϵ is a sufficiently small positive number), Theorem 9 reduces to [26, Corollary 1]. So, by choosing suitable M1, N1, Z1, and R1, Theorem 9 could reduce the conservativeness of [26, Corollary 1], by which the improvement over [26, Corollary 1] is theoretically demonstrated.
If we choose U=0, Mi=0, Ni=0, Zi=0, and Ri=ϵIn (ϵ is a sufficiently small positive number) (i=1,2) in Theorem 8, the following corollary can be easily derived.
Corollary 12.
For given scalars τ>0 and η>0, systems (1a) and (1b) are asymptotically stable, if the operator 𝒟 is stable and there exist real n×n matrices PT=P>0, WT=W>0, QkT=Qk>0, and TkT=Tk>0 (k=1,2) such that LMI (33) holds:
(33)Φ^:=[Φ^11Φ^12Φ^13ATSCΦ^12TΦ^22Φ^23A1TSCΦ^13TΦ^23TΦ^330CTSACTSA10CTSC-W]<0,
where
(34)Φ^11=PA+ATP+Q1+Q2+ATSA-τ-1T1-η-1T2,Φ^12=PA1+ATSA1+τ-1T1,Φ^13=-ATPC+η-1T2,Φ^22=-Q1-τ-1T1+A1TSA1,Φ^23=-A1TPC,Φ^33=-Q2-η-1T2,S=W+τT1+ηT2.
Next we will show that the pair of stability criteria presented in Proposition 3 (i.e., [21, Theorem 1]) and Corollary 12 have the same conservativeness. However, the stability criterion presented in Corollary 12 is more simpler than the one in Proposition 3 (i.e., [21, Theorem 1]), which reduces greatly the computational complexity.
Theorem 13.
For given scalars τ>0 and η>0, LMIs (7) and (8) are feasible if and only if the LMI (33) is feasible.
Proof.
The “Only If” Part. It follows from (7) and (8) that
(35)Xˇ≥Xˇ5TXˇ55-1Xˇ5,Yˇ≥Yˇ5TYˇ55-1Yˇ5,Φˇ+τXˇ+ηYˇ+ΦˇX+ΦˇY+AˇTSAˇ<0,
and hence
(36)Φˇ+τXˇ5TXˇ55-1Xˇ5+ηYˇ5TYˇ55-1Yˇ5+ΦˇX+ΦˇY+AˇTSAˇ<0.
Noting that
(37)τXˇ5TXˇ55-1Xˇ5+ΦˇX+τ-1Π1TXˇ55Π1≥0,ηYˇ5TYˇ55-1Yˇ5+ΦˇY+η-1Π2TYˇ55Π2≥0,
one can conclude from (36) that
(38)Φˇ-τ-1Π1TXˇ55Π1-η-1Π2TYˇ55Π2+AˇTSAˇ<0,
which becomes LMI (33) by letting W=R, T1=Xˇ55, and T2=Tˇ55. That is, LMI (33) is feasible.
The “If” Part. Since LMI (33) is feasible, there exist real n×n matrices PT=P>0, WT=W>0, QkT=Qk>0, and TkT=Tk>0 (k=1,2) such that LMI (33) is satisfied. Set R=W, Xˇ55=T1, and Yˇ55=T2. Then it follows from (33) and Lemma 1 that
(39)[Φˇ-τ-1Π1TXˇ55Π1-η-1Π2TYˇ55Π2AˇTSSAˇ-S]<0.
Let
(40)Xˇ5=-τ-1Xˇ55Π1,Xˇ=Xˇ5TXˇ55-1Xˇ5,Yˇ5=-η-1Yˇ55Π2,Yˇ=Yˇ5TYˇ55-1Yˇ5.
Then LMI (7) holds, and
(41)τXˇ5TXˇ55-1Xˇ5+ΦˇX+τ-1Π1TXˇ55Π1=0,ηYˇ5TYˇ55-1Yˇ5+ΦˇY+η-1Π2TYˇ55Π2=0.
Furthermore, LMI (8) is derived from (39).
The proof is completed.
5. Numerical Comparisons of Stability Criteria
In this section we will present one example to demonstrate effectiveness of the theoretic results described above.
Example 1 .
Consider the stability of systems (1a) and (1b) with
(42)A=[-0.90.20.1-0.9],A1=[-1.1-0.2-0.1-1.1],C=[-0.200.2-0.1].
Case 1 (τ=η).
The upper bounds on the delay τ under which robust stability of this system can be guaranteed using the methods in [9–27] and Theorem 9 of this paper are listed in Table 1. It is clear that our results are significantly better than those in [9–26] because our allowable maximum time delay τ is larger.
The maximum allowable bounds of τ=η.
Method
maxτ
[9]
0.3
[10]
0.5658
[11]
0.5937
[12]
0.6054
[13]
0.6612
[14]
0.7039
[15]
0.74
[16]
0.8844
[17]
1.0402
[18]
1.3718
[19]
1.6014
[20]
1.61
[21]
1.6527
[22]
1.7191
[23]
1.7220
[24]
1.78
[25]
1.7844
[26, Corollary 1]
1.7856
[27] and Theorem 9
1.7890
Case 2 (τ≠η).
For different values of η, Table 2 lists the maximum allowable bounds on τ that guarantee the stability of the system. It can be seen that the maximum allowable bound on τ decreases as η increases when η is small but that τ remains almost unchanged when η increases to certain extent.
The maximum allowable bounds of τ for different η.
Method
η
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
1.1
1.2
1.3
1.4
1.5
10000
100000
[21] and Corollary 12
τ
1.7100
1.6987
1.6883
1.6792
1.6718
1.6664
1.6624
1.6591
1.6563
1.6543
1.6531
1.6527
1.6527
1.6527
1.6527
1.6527
1.6527
[25]
τ
1.7844
1.7757
1.7669
1.7581
1.7495
1.7413
1.7338
1.7273
1.7226
1.7201
1.7193
1.7191
1.7191
1.7191
1.7191
1.7191
1.7191
[27] and Theorem 8
τ
1.8305
1.8174
1.8038
1.7897
1.7755
1.7616
1.7484
1.7366
1.7272
1.7213
1.7202
1.7202
1.7202
1.7202
1.7202
1.7193
1.7193
6. Conclusions
By using matrix analysis techniques, several stability criteria (i.e., [21, Theorem 1] and [27, Theorem 1 and Corollary 1]) for delayed neutral systems have been simplified. The numbers of LMI variables in [21, Theorem 1] and [27, Theorem 1 and Corollary 1] are reduced into the simplified ones, which can obviously reduce computational complexity. Furthermore, it is theoretically proven that the simplified stability criteria have the same conservativeness as the original ones. A numerical example is given to illustrate the theoretic results investigated in this paper.
Extending the idea of this paper to other system models, including singular delayed systems [32–35], stochastic systems [36], Markovian jump systems [37, 38], and genetic regulatory networks [39–41], is under consideration.
Conflict of Interests
The authors declare that there is no commercial or associative interest that represents a conflict of interest in connection with the work submitted.
Acknowledgments
This work is supported by the National Natural Science Foundation of China under Grant no. 11371006, the fund of Heilongjiang Province Innovation Team Support Plan under Grant no. 2012TD007, the fund of Heilongjiang University Innovation Team Support Plan under Grant no. Hdtd2010-03, the Fund of Key Laboratory of Electronics Engineering, College of Heilongjiang Province, (Heilongjiang University), China, and the fund of Heilongjiang Education Committee. The authors thank the anonymous referees for their helpful comments and suggestions which improve greatly this paper.
KolmanovskiiV. B.MyshkisA.1992Boston, Mass, USAKluwer Academic Publishers10.1007/978-94-015-8084-7KuangY.1993191Boston, Mass, USAAcademic Pressxii+398Mathematics in Science and EngineeringMR1218880NiculescuS.-I.2001269Berlin, GermanySpringerxvi+383Lecture Notes in Control and Information SciencesMR1880658LinX.ZhangX.WangY.Robust passive filtering for neutral-type neural networks with time-varying discrete and unbounded distributed delays2013350596698910.1016/j.jfranklin.2013.01.021MR3046050ZhangX.LinX.WangY.Robust fault detection filter design for a class of neutral-type neural networks with time-varying discrete and unbounded distributed delays201334559060710.1002/oca.2041MR3105675WangY.ZhangX.HuY.Robust H∞ control for a class of uncertain neutral stochastic systems with mixed delays: A CCL approach201332263164610.1007/s00034-012-9485-zMR3036050WangY.ZhangX.HeY.Improved delay-dependent robust stability criteria for a class of uncertain mixed neutral and Lur'e dynamical systems with interval time-varying delays and sector-bounded nonlinearity20121352188219410.1016/j.nonrwa.2012.01.014MR2911908ZBL1254.93129ZhangG. D.LinX.ZhangX.Exponential stabilization of neutral-type neural networks with mixed interval time-varying delays by intermittent control: a CCL approach20143337139110.1007/s00034-013-9651-yLienC.-H.YuK.-W.HsiehJ.-G.Stability conditions for a class of neutral systems with multiple time delays20002451202710.1006/jmaa.2000.6716MR1756574ZBL0973.34066ChenJ. D.LienC. H.FanK. K.ChouJ. H.Criteria for asymptotic stability of a class of neutral systems via a LMI approach200114864424472-s2.0-003550287810.1049/ip-cta:20010772HeY.WangQ.-G.LinC.WuM.Augmented Lyapunov functional and delay-dependent stability criteria for neutral systems2005151892393310.1002/rnc.1039MR2187877ZBL1124.34049WuM.HeY.SheJ.-H.New delay-dependent stability criteria and stabilizing method for neutral systems200449122266227110.1109/TAC.2004.838484MR2106758SunJ.LiuG. P.ChenJ.Delay-dependent stability and stabilization of neutral time-delay systems200919121364137510.1002/rnc.1384MR2537819ZBL1169.93399QianW.LiuJ.FeiS.Augmented Lyapunov functional approach for stability of neutral systems with mixed delays201214257257910.1002/asjc.294MR2908819FridmanE.New Lyapunov-Krasovskii functionals for stability of linear retarded and neutral type systems200143430931910.1016/S0167-6911(01)00114-1MR2008812ZBL0974.93028LienC.-H.ChenJ.-D.Discrete-delay-independent and discrete-delay-dependent criteria for a class of neutral systems2003125133412-s2.0-024262123910.1115/1.1540995ChenH.ZhangY.ZhaoY.Stability analysis for uncertain neutral systems with discrete and distributed delays201221823113511136110.1016/j.amc.2012.04.043MR2943980ZBL1277.93060ParkJ. H.KwonO.On new stability criterion for delay-differential systems of neutral type2005162262763710.1016/j.amc.2004.01.001MR2111860ZBL1077.34075HanQ.-L.On robust stability of neutral systems with time-varying discrete delay and norm-bounded uncertainty2004406108710922-s2.0-204253599710.1016/j.automatica.2004.01.007ZBL1073.93043HanQ.-L.A descriptor system approach to robust stability of uncertain neutral systems with discrete and distributed delays20044010179117962-s2.0-384311440110.1016/j.automatica.2004.05.002ZBL1075.93032HeY.WuM.SheJ.-H.LiuG.-P.Delay-dependent robust stability criteria for uncertain neutral systems with mixed delays2004511576510.1016/S0167-6911(03)00207-XMR2026262ZBL1157.93467KwonO. M.ParkJ. H.LeeS. M.On stability criteria for uncertain delay-differential systems of neutral type with time-varying delays2008197286487310.1016/j.amc.2007.08.048MR2400710ZBL1144.34052XuS.LamJ.ZouY.Further results on delay-dependent robust stability conditions of uncertain neutral systems200515523324610.1002/rnc.983MR2124633ZBL1078.93055FridmanE.ShakedU.A descriptor system approach to H∞ control of linear time-delay systems200247225327010.1109/9.983353MR1881892LiuX.-G.WuM.MartinR.TangM.-L.Stability analysis for neutral systems with mixed delays2007202247849710.1016/j.cam.2006.03.003MR2319971ZBL1120.34057ZhaoZ.WangW.YangB.Delay and its time-derivative dependent robust stability of neutral control system200718721326133210.1016/j.amc.2006.09.042MR2321337ZBL1114.93076QianW.LiuJ.SunY.FeiS.A less conservative robust stability criteria for uncertain neutral systems with mixed delays20108051007101710.1016/j.matcom.2009.12.007MR2610059ZBL1186.93057GuK.NiculescuS.-I.Further remarks on additional dynamics in various model transformations of linear delay systems200146349750010.1109/9.911431MR1819827ZBL1056.93511HaleJ. K.Verduyn LunelS. M.199399New York, NY, USASpringerx+447Applied Mathematical SciencesMR1243878BoydS.El GhaouiL.FeronE.BalakrishnanV.199415Philadelphia, Pa, USASociety for Industrial and Applied Mathematicsxii+193SIAM Studies in Applied Mathematics10.1137/1.9781611970777MR1284712GahinetP.ApkarianP.A linear matrix inequality approach to H∞ control19944442144810.1002/rnc.4590040403MR1286148ZBL0808.93024ZhuH.ZhangX.CuiS.Further results on H∞ control for discrete-time uncertain singular systems with interval time-varying delays in state and input201334332834710.1002/oca.2023MR3047730ZBL1270.93047LiF.ZhangX.A delay-dependent bounded real lemma for singular LPV systems with time-variant delay201222555957410.1002/rnc.1714MR2899488ZBL1273.93088LiF. B.ShiP.WuL. G.ZhangX.Fuzzy-model-based D-stability and non-fragile control for discrete-time descriptor systems with multiple delays201310.1109/TFUZZ.2013.2272647ZhangX.ZhuH.Robust stability and stabilization criteria for discrete singular time-delay LPV systems20121441084109410.1002/asjc.418MR2955449WuL. G.ZhengW. X.GaoH. J.Dissipativity-based sliding mode control of switched stochastic systems201358378579110.1109/TAC.2012.2211456WuL.SuX.ShiP.Sliding mode control with bounded L2 gain performance of Markovian jump singular time-delay systems20124881929193310.1016/j.automatica.2012.05.064MR2950452ZBL1268.93037LiF. B.WuL. G.ShiP.Stochastic stability of semi-Markovian jump systems with mode-dependent delays201310.1002/rnc.3057WangY. T.YuA. H.ZhangX.Robust stability of stochastic genetic regulatory networks with time-varying delays: a delay fractioning approach20132351217122710.1007/s00521-012-1034-yChenL. L.ZhouY.ZhangX.Guaranteed cost control for uncertain genetic regulatory networks with interval time-varying delays201413110511210.1016/j.neucom.2013.10.035ZhangX.YuA. H.ZhangG. D.M-matrix-based delay-range-dependent global asymptotical stability criterion for genetic regulatory networks with time-varying delays201311381510.1016/j.neucom.2012.12.046