This paper investigates static output feedback guaranteed cost control for a class of nonlinear
discrete-time systems where the delay in state vector is inconsistent with the delay in nonlinear
perturbations. Based on the output measurement, the controller is designed to ensure the robust exponentially
stability of the closed-loop system and guarantee the performance of system to achieve an
adequate level. By using the Lyapunov-Krasovskii functional method, some sufficient conditions for
the existence of robust output feedback guaranteed cost controller are established in terms of linear
matrix inequality. A numerical example is provided to show the effectiveness of the results obtained.
1. Introduction
In control theory and practice, one of the most important open problems is the static output feedback (SOF) problem. The main principle of the SOF control is to utilize the measured output to excite the plant. Since the controller can be easily implemented in practice, the SOF control has attracted a lot of attention over the past few decades and has been applied to many areas such as economic, communication, and biological systems [1, 2]. The goal of design SOF controller is to ensure asymptotically stable or exponential stable of the original system [3]. However, in many practical systems, controller designed is to not only ensure asymptotically or exponentially stable of the system but also guarantee the performance of system to achieve an adequate level. One method of dealing with this problem is the guaranteed cost control first introduced by Chang and Peng [4]. This method has the advantage of providing an upper bound on a given performance index and thus the system performance degradation is guaranteed to be no more than this bound. Based on this idea, a lot of significant results have been addressed for continuous-time systems in [5–7] and for discrete-time systems in [8].
It is well known that time-delays as well as parameter uncertainties frequently lead to instability of systems. Moreover, the existing of time-delays and uncertainties make the system more complex [9, 10].
In the past studies for guaranteed cost control, almost most of the articles considered linear systems [11, 12]. However, in majority dynamic systems, the nonlinear perturbations appear more and more frequently. Therefore, we not only deal with the time-varying delays and uncertainties, but also deal with the nonlinearities. Difficulties then arise when one attempts to derive exponential stabilization conditions. Hence in this case, the methods in linear systems [11, 12] can not be directly applied to nonlinear systems. This calls for a fresh look at the problem with an improved Lyapunov-Krasovskii functionals and a new set of LMI conditions. In this paper, we aim to design robust static output feedback guaranteed cost controller for a class of nonlinear discrete-time systems with time-varying delays. By constructing a set of improved Lyapunov-Krasovskii functionals, a new criterion for the existence of robust static output feedback guaranteed cost controller is established and described in terms of linear matrix inequality. A numerical example is provided to show the effectiveness of the results obtained.
Notations. In this paper, a matrix A is symmetric if A=AT. λmax(A)(λmin(A)) denotes the maximum (minimum) value of the real parts of eigenvalues of A. The symmetric terms in a matrix are denoted by *. X>0 (resp., X≥0), for X∈Rn×n, means that the matrix is real symmetric positive definite (resp., positive semidefinite). N+ denotes the set of all real nonnegative integers.
2. Preliminaries
Consider the following control system:
(1)x(k+1)=[A+ΔA(k)]x(k)+[A1+ΔA1(k)]x(k-d(k))+Bu(k)+Ff(x(k))+Gg(x(k-h(k))),y(k)=Cx(k)+C1x(k-d(k)),k∈N+,x(k)=ϕk,k=-τ,-τ+1,…,0,
where x(k)∈Rn is the state vector, y(k)∈Rr is the observation output, and u(k)∈Rm is the control intput. A, A1, B, F, G, C, and C1 are given constant matrices with appropriate dimensions. ΔA(k), ΔA1(k) are the time-varying parameter uncertainties that are assumed to satisfy the following admissible condition:
(2)[ΔA(k)ΔA1(k)]=MΔ(k)[N1N2],ΔT(k)Δ(k)⩽I,∀k∈N+,
where M, N1, and N2 are some given constant matrices with appropriate dimensions. The positive integers d(k) and h(k) are time-varying delays satisfying
(3)0≤d(k)≤d,0≤h(k)≤h,
where d, h are known positive integers. f(x(k))∈Rn and g(x(k-h(k))∈Rn are unknown nonlinear functions, assumed as
(4)fT(x(k))f(x(k))⩽β1xT(k)L1TL1x(k),gT(x(k-h(k)))g(x(k-h(k)))⩽β2xT(k-h(k))L2TL2x(k-h(k)),
where β1, β2 are known positive integers and L1, L2 are known real matrices. The initial condition ϕ=(ϕ-τ,ϕ-τ+1,…,ϕ0)∈R(τ+1)n with the norm
(5)∥ϕ∥=max{∥x(-τ)∥,…,∥x(0)∥},
where τ=max{h,d}.
The corresponding cost function is as follows:
(6)J=∑k=0∞{xT(k)Sx(k)+xT(k-d(k))S1x(k-d(k))+xT(k-h(k))S2x(k-h(k))+uT(k)Ru(k)},
where S, S1, S2, and R are given symmetric positive definite matrices with appropriate dimensions.
Substituting the output feedback controller u(k)=Ky(k) into system (1), we have
(7)x(k+1)=[A~(k)+BKC]x(k)+[A~1(k)+BKC1]x(k-d(k))+Ff(x(k))+Gg(x(k-h(k))),
where A~(k)=A+ΔA(k) and A~1(k)=A1+ΔA1(k).
The objective of this paper is to design an output feedback controller u(k)=Ky(k) for system (1) and cost function (6) such that the resulting closed-loop system is robust exponentially stable with an upper bound for cost function (6).
We first give the following definitions, which will be used in the next theorems and proofs.
Definition 1.
Given α>0, the closed-loop system (7) is said to be robust exponentially stable with a decay rate α, if there exists scalars σ>0 such that for every solution x(k,ϕ) of the system satisfies the condition:
(8)∥x(k,ϕ)∥≤σe-αk∥ϕ∥,∀k∈N+.
Definition 2.
For system (1) and cost function (6), if there exist a static output feedback control law u*(k) and a positive constant J* such that the closed-loop system (7) is robust exponentially stable with a decay rate α and the value (6) satisfies J≤J*, then J* is said to be a guaranteed cost index and u*(k) is said to be a robust output feedback guaranteed cost control law of the system.
The following lemmas are essential in establishing our main results.
Lemma 3 (see [11]).
For any x, y∈Rn, and positive symmetric definite matrix N∈Rn×n, we have
(9)±2yTx≤xTN-1x+yTNy.
Lemma 4 (Schur complement lemma [13]).
Given constant matrix X, Y, and Z with appropriate dimensions satisfying X=XT, Y=YT>0. Then X+ZTY-1Z<0 if and only if S=(XZTZ-Y)<0.
3. Main Results
In this section, by constructing a new set of Lyapunov-Krasovskii functionals, we give a sufficient condition for the existence of robustly output feedback guaranteed cost control for system (1).
Theorem 5.
For a given scalar α>0, the control u(k)=Ky(k) is a robustly static output feedback guaranteed cost controller for nonlinear system (1), if there exist symmetric positive definite matrices P, Q1, Q2, R1, R2, and K, arbitrary matrix N, and scalars ε1≥0, ε2≥0, such that the following LMI holds:
(10)(Θ10Θ13Θ14Θ15*Θ2000**Θ300***Θ40****Θ5)<0,
where
(11)Θ1=(Δ11Δ120*Δ22Δ23**Δ33),Θ14=(0000NBNMNFNG0000),Θ13=(CTKTCTKT0000),Θ15=(0000C1TKTC1TKT),Θ2=diag{Δ44,Δ55,Δ66},Θ3=diag{-0.5R-1,-I},Θ4=diag{Λ1,Λ2,-I,-I},Θ5=diag{-0.5e-2αdR-1,-e-2αI},Δ11=-P+d¯(Q1+R1)+eαN1TN1+h¯(Q2+R2)+β1(e2α+ε1)L1TL1+S,Δ12=eαNA,Δ23=eαd¯NA1,Δ33=-Q1-R1+eαd¯N2TN2+e2αdS1,Δ22=P-(N+NT),Δ44=-Q2-R2+(ε2+1)β2e2αhL2TL2+e2αhS2,Δ55=-ε1I,Δ66=-ε2I,Λ1=-e-2αd¯I,Λ2=-1eαd¯+eαI,d¯=d+1,h¯=h+1,
and the guaranteed cost value is given by J*=μ2∥ϕ∥2, where
(12)μ2=λmax(P)+d(d+1)[λmax(Q1)+λmax(R1)]+h(h+1)[λmax(Q2)+λmax(R2)].
Proof.
We first introduce the new variable z(k)=eαkx(k). The closed-loop system (7) is reduced to
(13)z(k+1)=[A~(k)+BKC]eαz(k)+[A~1(k)+BKC1]eα(d(k)+1)z(k-d(k))+Feα(k+1)f¯(z(k))+Geα(k+1)g¯(z(k-h(k))),
where f¯(z(k))=f(z(k)/eαk) and g¯(z(k-h(k)))=g(z(k-h(k))/eα(k-h(k))).
Associated with (2), the above equality is reduced to
(14)z(k+1)=[A¯(k)+B¯(k)KC¯(k)]z(k)+[A1¯(k)+B¯(k)KC1¯(k)]z(k-d(k))+F¯(k)f¯(z(k))+G¯(k)g¯(z(k-h(k))),
where
(15)A¯(k)=eα[A+MΔ(k)N1],C¯(k)=e-αkC,A1¯(k)=eα(d(k)+1)[A1+MΔ(k)N2],B¯(k)=eα(k+1)B,C1¯(k)=e-α(k-d(k))C1,F¯(k)=eα(k+1)F,G¯(k)=eα(k+1)G,(16)f¯T(z(k))f¯(z(k))≤β1e-2αkzT(k)L1TL1z(k),(17)g¯T(z(k-h(k)))g¯(z(k-h(k)))≤β2e-2α(k-h(k))zT(k-h(k))L2TL2z(k-h(k)).
Consider a Lyapunov-Krasovskii functional candidate for the closed-loop system (14) as
(18)V(k)=V1(k)+V2(k)+V3(k)+V4(k)+V5(k),
where
(19)V1(k)=zT(k)Pz(k),V2(k)=∑i=k-d(k)k-1zT(i)Q1z(i)+∑i=k-h(k)k-1zT(i)Q2z(i),V3(k)=∑j=-d+10∑i=k+jk-1zT(i)Q1z(i)+∑j=-h+10∑i=k+jk-1zT(i)Q2z(i),V4(k)=∑i=k-d(k)k-1zT(i)R1z(i)+∑i=k-h(k)k-1zT(i)R2z(i),V5(k)=∑j=-d+10∑i=k+jk-1zT(i)R1z(i)+∑j=-h+10∑i=k+jk-1zT(i)R2z(i).
Calculating the difference of V(k) we have
(20)ΔV1(k)=zT(k+1)Pz(k+1)-zT(k)Pz(k),(21)ΔV2(k)≤zT(k)(Q1+Q2)z(k)-zT(k-d(k))Q1z(k-d(k))-zT(k-h(k))Q2z(k-h(k))+∑i=k-d+1kzT(i)Q1z(i)+∑i=k-h+1kzT(i)Q2z(i),(22)ΔV3(k)=∑j=-d+10{∑i=k+j+1kzT(i)Q1z(i)-∑i=k+jk-1zT(i)Q1z(i)}+∑j=-h+10{∑i=k+j+1kzT(i)Q2z(i)-∑i=k+jk-1zT(i)Q2z(i)}=zT(k)(dQ1+hQ2)z(k)-∑j=k-d+1kzT(j)Q1z(j)-∑j=k-h+1kzT(j)Q2z(j).
Combine (21) and (22), we have
(23)ΔV2(k)+ΔV3(k)≤zT(k)(d¯Q1+h¯Q2)z(k)-zT(k-d(k))Q1z(k-d(k))-zT(k-h(k))Q2z(k-h(k)).
Similarly, we can get
(24)ΔV4(k)+ΔV5(k)≤zT(k)(d¯R1+h¯R2)z(k)-zT(k-d(k))R1z(k-d(k))-zT(k-h(k))R2z(k-h(k)).
Therefore, from (17)–(24), we have
(25)ΔV(k)≤zT(k+1)Pz(k+1)+zT(k)[d¯(Q1+R1)+h¯(Q2+R2)-P]z(k)-zT(k-d(k))(Q1+R1)z(k-d(k))-zT(k-h(k))(Q2+R2)z(k-h(k)).
Multiplying 2zT(k+1)N both sides of the identity (14),
(26)-2zT(k+1)Nz(k+1)+2zT(k+1)×N(A¯(k)+B¯(k)KC¯(k))z(k)+2zT(k+1)×N(A1¯(k)+B¯(k)KC1¯(k))z(k-d(k))+2zT(k+1)NF¯(k)f¯+2zT(k+1)NG¯(k)g¯=0.
Note that for any ε1≥0, ε2≥0, it follows from (16) to (17)
(27)ε1[β1e-2αkzT(k)L1TL1z(k)-f¯T(z(k))f¯(z(k))]≥0,ε2[(z(k-h(k)))β2e-2α(k-h(k))zT(k-h(k))L2TL2z(k-h(k))-g¯T(z(k-h(k)))g¯(z(k-h(k)))]≥0.
Substituting (26)-(27) into (25), we have
(28)ΔV(k)≤zT(k+1)Pz(k+1)+zT(k)[d¯(Q1+R1)+h¯(Q2+R2)-P]z(k)-zT(k-d(k))(Q1+R1)z(k-d(k))-zT(k-h(k))(Q2+R2)z(k-h(k))-2zT(k+1)Nz(k+1)+2zT(k+1)×N(A¯(k)+B¯(k)KC¯(k))z(k)+2zT(k+1)×N(A1¯(k)+B¯(k)KC1¯(k))z(k-d(k))+2zT(k+1)NF¯(k)f¯(z(k))+2zT(k+1)NG¯(k)g¯(z(k-h(k)))+ε1[β1e-2αkzT(k)L1TL1z(k)-f¯T(z(k))f¯(z(k))]+ε2[(z(k-h(k)))β2e-2α(k-h(k))zT(k-h(k))L2TL2z(k-h(k))-g¯T(z(k-h(k)))g¯(z(k-h(k)))].
Dealing with partial idem in (28) using Lemma 3, it follows
(29)2zT(k+1)N(A¯(k)+B¯(k)KC¯(k))z(k)=2zT(k+1)eα[NA+NMΔ(k)N1]z(k)+2zT(k+1)eαNBKCz(k)≤2zT(k+1)eαNAz(k)+eαzT(k+1)NMMTNTz(k+1)+eαzT(k)N1TN1z(k)+zT(k+1)e2αNBBTNTz(k+1)+zT(k)CTKTKCz(k),2zT(k+1)N(A1¯(k)+B¯(k)KC1¯(k))z(k-d(k))=zT(k+1)eα(d(k)+1)(NA1+A1TNT)z(k-d(k))+2eα(d(k)+1)zT(k+1)NMΔ(k)N2z(k-d(k))+2zT(k+1)eαd(k)NBeαKC1z(k-d(k))≤zT(k+1)eαd¯(NA1+A1TNT)z(k-d(k))+eαd¯zT(k+1)NMMTNTz(k+1)+eαd¯zT(k-d(k))N2TN2z(k-d(k))+zT(k+1)e2αdNBBTNTz(k+1)+e2αzT(k-d(k))C1TKTKC1z(k-d(k)),2zT(k+1)NF¯(k)f¯(z(k))=2zT(k+1)NFeα(k+1)f¯(z(k))≤zT(k+1)NFFTNTz(k+1)+e2α(k+1)f¯T(z(k))f¯(z(k))≤zT(k+1)NFFTNTz(k+1)+β1e2αzT(k)L1TL1z(k).
Similarly, we have
(30)2zT(k+1)NG¯(k)g¯(z(k-h(k)))≤zT(k+1)NGGTNTz(k+1)+β2e2αhzT(k-h(k))L2TL2z(k-h(k)).
Adding the following relation to inequality (28)
(31)[zT(k)S¯z(k)+zT(k-d(k))S1¯z(k-d(k))+zT(k-h(k))S2¯z(k-h(k))+uT(k)Ru(k)]-[zT(k)S¯z(k)+zT(k-d(k))S1¯z(k-d(k))+zT(k-h(k))S2¯z(k-h(k))+uT(k)Ru(k)]=0,
where S¯=e-2αkS, S1¯=e-2α(k-d(k))S1, and S2¯=e-2α(k-h(k))S2, and using
(32)uT(k)Ru(k)≤zT(k)CTKTRKCz(k)+2eαdzT(k)CTKTRKC1z(k-d(k))+e2αdzT(k-d(k))C1TKTRKC1z(k-d(k))≤2zT(k)CTKTRKCz(k)+2e2αdzT(k-d(k))C1TKTRKC1z(k-d(k)),
we can get
(33)ΔV(k)≤ξT(k)Ωξ(k)-[zT(k)S¯z(k)+zT(k-d(k))S1¯z(k-d(k))+zT(k-h(k))S2¯z(k-h(k))+uT(k)Ru(k)],
where ξ(k)=[zT(k)zT(k+1)zT(k-d(k))zT(k-h(k))f¯T(z(k))g¯T(z(k-h(k)))]T and
(34)Ω=(Ω11Δ120000*Ω22Δ23000**Ω33000***Δ4400****Δ550*****Δ66),Ω11=Δ11+CTKT(I+R+RT)KC,Ω22=Δ22+e2αd¯NBBTNT+eαNMMTNT+eαd¯NMMTNT+NFFTNT+NGGTNT,Ω33=Δ33+e2αC1TKTKC1+2e2αdC1TKTRKC1.
By Lemma 4, the condition Ω<0 is equivalent to LMI (10). Therefore, from (33) it follows that
(35)ΔV(k)<0,
which implies that V(k)≤V(0), ∀k∈N+.
We can easily get
(36)μ1∥z(k)∥2≤V(k)≤μ2∥zk∥2,
where ∥zk∥=max{∥z(k-τ)∥,…,∥z(k)∥}, μ1=λmin(P),
(37)μ2=λmax(P)+d(d+1)[λmax(Q1)+λmax(R1)]+h(h+1)[λmax(Q2)+λmax(R2)].
From (35) and (36), we can get
(38)μ1∥z(k)∥2≤μ2∥ϕ¯∥2,∥z(k)∥≤μ2μ1∥ϕ¯∥.
Using the relation z(k)=eαkx(k), we can get
(39)∥x(k)∥≤μ2μ1e-αk∥ϕ∥,∀k∈N+.
Therefore, the closed-loop system (7) is exponentially stable. Next we will find the guaranteed cost value, from (33), we can get
(40)ΔV(k)≤-[xT(k)Sx(k)+xT(k-d(k))S1x(k-d(k))+xT(k-h(k))S2x(k-h(k))+uT(k)Ru(k)].
Summing up both sides of (40) from 0 to n-1, we can get
(41)∑k=0n-1[xT(k)Sx(k)+xT(k-d(k))S1x(k-d(k))+xT(k-h(k))S2x(k-h(k))+uT(k)Ru(k)]≤V(0)-V(n).
Letting n→+∞, noting that V(n)→0, we can get
(42)J=∑k=0∞[xT(k)Sx(k)+xT(k-d(k))S1x(k-d(k))+xT(k-h(k))S2x(k-h(k))+uT(k)Ru(k)xT(k)]≤V(0),
associated with (36), and we have J≤μ2∥ϕ∥2=J*.
Remark 6.
When time-delay in state vector keeps consistent with the delay in nonlinear perturbations and uncertain items disappear, the system (1) induced to
(43)x(k+1)=Ax(k)+A1x(k-d(k))+Bu(k)+Ff(x(k))+Gg(x(k-d(k))),y(k)=Cx(k)+C1x(k-d(k)),k∈N+,x(k)=ϕk,k=-d,-d+1,…,0,
at the same time, the closed-loop system (7), and cost function (6) are reduced to
(44)x(k+1)=[A+BKC]x(k)+[A1+BKC1]x(k-d(k))+Ff(x(k))+Gg(x(k-d(k))),J=∑k=0∞{xT(k)Sx(k)+xT(k-d(k))S1+uT(k)Ru(k)}.
Then we give a sufficient condition for the existence of static output feedback control for system (43).
Theorem 7.
For a given scalar α>0, the control u(k)=Ky(k) is a static output feedback guaranteed cost controller for system (43), if there exist symmetric positive definite matrices P, Q1, R1, and K, arbitrary matrix N, and scalars ε1≥0, ε2≥0, such that the following LMI holds:
(45)(Γ10Θ13Γ14Γ15*Γ2000**Θ300***Γ40****Γ5)<0,
where
(46)Γ1=(Δ11′Δ120*Δ22Δ23**Δ33′),Γ14=(000NBNFNG000),Θ15=(0000C1TKTC1TKT),Γ2=diag{Δ44′,Δ55′},Γ4=diag{Λ1,Λ2,-I,-I},Γ5=diag{-0.5e-2αdR-1,-e-2αI},Δ11′=-P+d¯(Q1+R1)+β1(e2α+ε1)L1TL1+S,Δ33′=-Q1-R1+e2αdS1+β2(ε2+e2αd)L2TL2,Δ44′=-ε1I,Δ55′=-ε2I,
and the guaranteed cost value is given by J*=μ2∥ϕ∥2, where μ2′=λmax(P)+d(d+1)[λmax(Q1)+λmax(R1)].
Proof.
The proof of Theorem 7 is similar to that of Theorem 5, which are omitted.
Remark 8.
In this paper, we design the controller directly from the LMI without variable transformation [11] which reduces the amount of calculation. Moreover, based on Theorem 5, one can deduce the criteria for linear discrete-time systems with time-delay and nonlinear discrete-time systems with constant time-delay.
4. Numerical Example
Consider the nonlinear uncertain discrete-time system (1) with the following parameters:
(47)A=(-0.020.12-0.10.10.030.1-0.10.030.2),A1=(0.020.030.020.010.02-0.01-0.050.020.08),B=(-0.050.10.20.10.320.10.05-0.060.12),F=(0.16-0.30.02-0.20.140.20.130.140.1),G=(-0.030.060.020.40.10.10.02-0.030.02),C=(0.120.280-0.060.090.020.20.030.2),C1=(0.04-0.060.070.050.080.070.040.010.05),N1=(-0.50.20.10.20.30.1-0.80.20.1),N2=(-0.60.30.1-0.1-0.20.2-0.1-0.50.3),M=(0.030.020.04-0.050.120.10.03-0.020.02),S=(0.60.20.10.20.60.40.10.40.2),S1=(0.4-0.10.3-0.10.30.20.30.20.6),S2=(0.50.10.20.10.30.40.20.40.6),R=(1.120.3-0.10.31.50-0.101.6),L1=(-0.350.20.10.20.30.10.80.20.1),L2=(1.20.232-1.30.40.21.5-0.31.6),d(k)=1.5+1.5sinkπ2,h(k)=1+coskπ2,k=0,1,2,….
Given α=0.032, d=3, h=2, β1=0.23, β2=0.25, and the initial condition
(48)x(-3)=(-1.61.31.2),x(-2)=(-1.30.8-1),x(-1)=(0.4-0.50.2),x(0)=(1.20-1.2),
using the LMI Toolbox in MATLAB [14], the LMI (10) in Theorem 5 is satisfied with
(49)P=(4.11840.0455-0.38180.04551.6976-0.2973-0.3818-0.297310.3666),Q1=(0.0311-0.0020-0.0207-0.00200.0115-0.0158-0.0207-0.01580.1978),Q2=(0.0478-0.0032-0.0334-0.00320.0163-0.0254-0.0334-0.02540.3176),R1=(0.1001-0.0069-0.0710-0.00690.0329-0.0541-0.0710-0.05410.6715),R2=(0.1046-0.0077-0.0783-0.00770.0299-0.0596-0.0783-0.05960.7340),N=(5.22700.12110.00220.12202.2186-0.48080.0003-0.480013.0762),ε1=7.3374,ε2=6.7127,
and the controller parameter
(50)K=(0.6351-0.9829-0.4757-0.98292.79650.8749-0.47570.87490.9870).
Moveover, the solution of the closed-loop system satisfies
(51)∥x(k)∥≤4.0345e-0.032k∥ϕ∥,
and the guaranteed cost of the closed-loop system is as follows
(52)J*=27.4600∥ϕ∥2.
Simulation result is presented in Figure 1, which shows the convergence behavior of the proposed methods.
State trajectories of the closed-loop system.
5. Conclusion
In this paper, the problem of robust output feedback guaranteed cost control for nonlinear uncertain disctere system is researched. For all admissible uncertainties, an output feedback guaranteed cost controller has been designed such that the resulting closed-loop system is robust exponentially stable and guarantees an adequate level of system performance. A numerical example has been presented to illustrate the efficiency of the result.
Conflict of Interests
The authors declare that there is no conflict of interests regarding the publication of this paper.
YaeshI.ShakedU.H∞ optimization with pole constraints of static output-feedback controllers-a non-smooth optimization approach20122041066107210.1109/TCST.2011.21588242-s2.0-84861719730DongJ. X.YangG. H.Static output feedback control synthesis for linear systems with time-invariant parametric uncertainties200752101930193610.1109/TAC.2007.906227MR23558032-s2.0-36249014749HoD. W. C.LuG.Robust stabilization for a class of discrete-time non-linear systems via output feedback: the unified LMI approach200376210511510.1080/0020717031000067367MR19658472-s2.0-0037454724ChangS. S. L.PengT.Adaptive guaranteed cost control of systems with uncertain parameters1972174474483MR04498182-s2.0-001538427310.1109/TAC.1972.1100037DongJ. X.YangG. H.Robust static output feedback control synthesis for linear continuous systems with polytopic uncertainties20134961821182910.1016/j.automatica.2013.02.047MR30492342-s2.0-84877584286LienC. H.Non-fragile guaranteed cost control for uncertain neutral dynamic systems with time-varying delays in state and control input200731488989910.1016/j.chaos.2005.10.080MR22621842-s2.0-33747348121WeiJ.DongY.SuY.Guaranteed cost control of uncertain T-S fuzzy systems via output feedback approach20111093063172-s2.0-80055008107ChenW.-H.GuanZ.-H.LuX.Delay-dependent guaranteed cost control for uncertain discrete-time systems with delay2003150441241610.1049/ip-cta:200305722-s2.0-0141633842ChenL.Unfragile guaranteed-cost H∞ control of uncertain state-delay sampling system20122933593363QiuF.CuiB.JiY.Further results on robust stability of neutral system with mixed time-varying delays and nonlinear perturbations201011289590610.1016/j.nonrwa.2009.01.032MR25712622-s2.0-70449623984ThuanM. V.PhatV. N.TrinhH. M.Dynamic output feedback guaranteed cost control for linear systems with interval time-varying delays in states and outputs201221821106971070710.1016/j.amc.2012.04.039MR29270852-s2.0-84861799162ParkJ. H.On dynamic output feedback guaranteed cost control of uncertain discrete-delay systems: LMI optimization approach2004121114716210.1023/B:JOTA.0000026135.99294.32MR20629742-s2.0-4043097506BoydS.El GhaouiL.FeronE.BalakrishanV.199415Philadelphia, Pa, USASIAM10.1137/1.9781611970777MR1284712GahinetP.NemirovskiiA.LaubA. J.ChilaliM.1995Natick, Mass, USAMathWorks, Inc.