An instrumental matrix approach to design output feedback passive controller
for switched singular systems is proposed in this paper. The nonlinear
inequality condition including Lyapunov inverse matrix and controller gain
matrix is decoupled by introducing additional instrumental matrix variable.
Combined with multiple Lyapunov function method, the nonlinear inequality
is transformed into linear matrix inequality (LMI). An LMI condition is
presented for switched singular system to be stable and passive via static
output feedback under designed switching signal. Moreover, the conditions
proposed do not require the decomposition of Lyapunov matrix and its inverse
matrix or fixing to a special structure. The theoretical results are verified by
means of an example. The method introduced in the paper can be effectively
extended to a single singular system and normal switched system.
1. Introduction
The switched singular systems arise from, for instance, power systems, economic systems, and complex networks. As pointed out in [1], when the interrelationships among different industrial sectors are described and the capital and the demand are variable depending on seasons, the dynamic economic systems are modelled as periodically switched singular systems. In some complex hybrid networks, some algebraic constraints have to be considered. The special algebraic constraints that, for example, communication resources are always limited and required to be allocated to different levels of privileged users, are needed in the resource allocation process. Thus, constructing the network model with a set of constraints is reasonable and indispensable. The model can be denoted by a class of singular hybrid systems [2]. There has been increasing interest in analysis and synthesis for switched singular systems. The stability issues are discussed for continuous-time switched singular systems [3], discrete-time switched singular systems [4], linear switched singular systems [5, 6], nonlinear switched singular systems [7], and time-delay switched singular systems [8], respectively. In [9, 10], reachability conditions and admissibility criteria are presented, respectively. Reference [11] studies the initial instantaneous jumps at switching points and a sufficient stability condition is obtained for the switched singular system with both stable and unstable subsystems. At arbitrary switching instant, inconsistent state jump for switched singular systems can be compressed by hybrid impulsive controllers in [12]. Filters and observers are designed for switched singular systems in [13, 14] and [15], respectively.
It has been shown that passivity is a suitable design approach in power systems [16], neural networks [17], network control [18–20], signal processing [21, 22], Markovian jumping systems [23–25], switched systems [26, 27], and singular systems [28, 29]. In [16], the problem of passive control is considered for uncertain singular time-delay systems, and three types of controllers were designed, namely, state feedback controller, observer-based state feedback controller, and dynamic output feedback controller. Under these controllers, the closed-loop systems are quadratically stable and passive. Contingent failures are possible for a real system, which may cause performance of the system to be degraded and even hazard [30]. The work [17] applies passivity-based fault-tolerant synchronization control to chaotic neural networks against actuator faults by using the semi-Markov jump model approach. The work [27] investigates the problem of robust reliable passive control for a class of uncertain stochastic switched time-delay systems with actuator failures. Sufficient condition for the stochastic switched time-delay systems to be passive and exponentially stable under switching state feedback controller is derived. The work [29] designs a state feedback controller such that, for all possible actuator failures, the closed-loop singular system is exponentially stable and passive.
Up to now, little attention has been paid to passive control problem for switched singular systems. This motivates us to investigate this problem. Furthermore, considering the operational cost and the reliability of systems and the simplicity of implementation, output feedback is always adopted to stabilize a system. Thus, we study passive control of switched singular systems through output feedback.
In this paper, by introducing instrumental matrix variable, the nonlinear inequality including Lyapunov inverse matrix and controller gain matrix is decoupled, which makes the design of output feedback passive controllers for continuous-time switched singular systems easy. Based on multiple Lyapunov functions and variable substitution techniques, a new and simple sufficient condition is presented in terms of LMI, by solving which static output feedback passive controller can be designed. The novelty of the conditions proposed in this paper lies in the following aspect. Decomposition of Lyapunov matrix and its inverse matrix is not required. Moreover, the Lyapunov inverse matrix is not fixed to a special structure.
The rest of this paper is organized as follows. Problem statement and preliminaries are given in Section 2. Output feedback passive control is studied in Section 3. In Section 4, an example shows the efficiency of main results in the paper. Section 5 concludes this paper.
2. Problem Statement and Preliminaries
Consider the following switched singular system:(1)Ecix˙t=Acixt+Bciut+Gciωtzt=Hcixt+Dciωtyt=Ccixt,where i∈χ={1,2,…,m} is the switching signal. x(t)∈Rn is the state vector, u(t)∈Rm is the control input vector, ω(t)∈L2r[0,∞) is the external disturbance vector, z(t)∈Rr is the control input vector, and y(t)∈Rp is the measured controlled output vector. Eci,Aci,Bci,Cci,Dci,Gci,Hci are constant matrices with appropriate dimensions. Eci∈Rn×n and rank Eci=ri≤n. Without loss of generality, we assume that Cci is full row rank.
Let us consider the following static output feedback controller:(2)ut=Kiyt,where Ki is the controller gain matrix to be designed.
Then, the resulting closed-loop system can be described as(3)Ecix˙t=Aci+BciKiCcixt+Gciωtzt=Hcixt+Dciωtyt=Ccixt.
We are now considering the output feedback passive control problem for system (3). In this paper, we aim to design output feedback passive controller such that system (3) simultaneously satisfies the following two requirements.
System (3) with ω(t)=0 is stable.
For a give scalar η>0, the dissipation inequality(4)∫0TωTz-ηωTωdt≥0,∀T>0,
holds for all trajectories with zero initial condition. In this case, the closed-loop switched singular system (3) is said to be stable and passive with dissipation rate η.
To obtain the main results of this paper, the following transformation is introduced.
Since Cci is full row rank, there exists a nonsingular matrix Ti such that CciTi-1=[Ip0]. Using the nonsingular Ti as a similarity transformation for system (3), the closed-loop system (3) is equivalent to the following system:(5)Eix˙t=A¯ixt+Giωtzt=Hixt+Diωtyt=Cixt,where Ei=TiEciTi-1, Ai=TiAciTi-1, Bi=TiBci, Ci=CciTi-1=[Ip0], Gi=TiGci, Hi=HciTi-1, Di=Dci, and A¯i=Ai+BiKiCi.
3. Output Feedback Passive Control
The following theorem provides a sufficient condition under which system (5) is stable and passive with dissipation rate η.
Theorem 1.
If there exist simultaneously nonnegative real number βij,η and matrices Xi>0, Xj>0 and matrices Ki such that for any i,j∈χ, i≠j,(6)XiTEiT=EiXi≥0and inequality(7)XiTA¯iT+A¯iXi+∑j=1mβijEiTXi-EjTXjGi-XiTHiTGiT-HiXi2ηI-Di-DiT<0hold, system (5) is stable and passive with dissipation rate η under static output feedback u(t)=Kiy(t) via switching signal(8)i=argmaxxTtEiTXixt,i∈χ.
Proof.
When βij is simultaneously nonnegative, for x∈Rn/{x∣∏i=1mEix=0}, there must exist a i∈χ=1,…,m, such that for any j≠i, j∈χ, xT(EiTXi-EjTXj)x≥0 holds. Then ∑j=1mxT(EiTXi-EjTXj)x≥0 holds. Let(9)Ωi=x∈Rn∣xTEiTXi-EjTXjx≥0,∏i=1m∏i=1mEix≠0,j≠i,j∈χ.Clearly, ⋃i=1mΩi=Rn/{x∣∏i=1mEix=0}.
Construct Ω¯1=Ω1,…,Ω¯i=Ωi-⋃j=1i-1Ω¯j,…,Ω¯m=Ωm-⋃j=1m-1Ω¯j. Obviously, ⋃i=1mΩ¯i=Rn/{x∣∏i=1mEix=0} and Ω¯i⋂Ω¯j=ϕ, i≠j, hold.
Design switching signal as(10)i=argmaxxTtEiTXixt,i∈χ.
When x∈Ω¯i, choose Lyapunov function as (11)Vxt=xTtEiTXi-1xt.Premultiplying Xi-T and postmultiplying Xi-1 to XiTEiT=EiXi in (6), respectively, one gets EiTXi-1=Xi-TEi. Then, the derivation of Lyapunov function V(x(t)) along the closed-loop system (5) is(12)V˙xt=x˙TtEiTXi-1xt+xTtEiTXi-1x˙t=A¯ixt+GiωtTXi-1xt+xTtXi-TA¯ixt+Giωt.Therefore(13)V˙xt-2ωTtzt+2ηωTtωt=A¯ixt+GiωtTXi-1xt+xTtXi-TA¯ixt+Giωt-2ωTtHixt+Diωt+2ηωTtωt=xtωtTA¯iTXi-1+Xi-TA¯iXi-TGi-HiTGiTXi-1-Hi2ηI-Di-DiTxtωt=ζTtΞiζt,where ζ(t)=x(t)ω(t), Ξi=A¯iTXi-1+Xi-TA¯iXi-TGi-HiTGiTXi-1-Hi2ηI-Di-DiT. Pre- and postmultiplying Ξi by XiT00I and Xi00I, respectively, we obtain(14)XiT00IΞiXi00I=Φi,where Φi=XiTA¯iT+A¯iXiGi-XiTHiTGiT-HiXi2ηI-Di-DiT. Then, Ξi<0 is equivalent to Φi<0. Next, we prove that Φi<0 holds.
Suppose that (7) holds. Inequality (7) can be rewritten as(15)Φi+∑j=1mβijEiTXi-EjTXj000<0.According to switching signal (8), Φi<0 when (15) holds, and by (13), we get that the following inequality holds:(16)V˙xt-2ωTtzt+2ηωTtωt≤0.Taking the integral on the two sides of (16) from 0 to T, we obtain(17)∫0TV˙xt-2ωTtzt+2ηωTtωtdt≤0.To get the result, we introduce(18)ST=∫0T-2ωTtzt+2ηωTtωtdt,where T>0. Noting the zero initial condition, it can be shown that, for any T>0,(19)ST≤∫0TV˙xt-2ωTtzt+2ηωTtωtdt.It follows from (19) that S(T)≤0, and hence(20)∫0T-2ωTtzt+2ηωTtωtdt≤0is satisfied for any T>0. The system is passive with dissipation η.
Next, we prove the stability of system (5) with ω(t)=0. Following the similar procedures as used above, we have(21)V˙xt=xTtA¯iTXi-1+Xi-TA¯ixt.Note that (7) implies A¯iTXi-1+Xi-TA¯i<0. Therefore, we have V˙(x(t))<0 when (7) holds. According to Lyapunov theory, system (5) is stable. This completes the proof.
Condition (7) in Theorem 1 is not linear matrix inequality, which cannot be solved by MATLAB. The controller gains Ki are also difficult to be computed from these conditions. In order to solve output feedback passive controller, we induce the following important lemma.
Lemma 2.
For any i,j∈χ, inequality (7) holds if there exists an instrumental matrix Yi such that(22)ΨiXiT-Yi+A¯iYiTGi-XiTHiTXi-YiT+YiA¯iT-Yi-YiT0GiT-HiXi02ηI-Di-DiT<0,where Ψi=YiA¯iT+A¯iYiT+∑j=1mβij(EiTXi-EjTXj).
Proof.
Let Λi=IA¯i000I; then Λi is full row rank. By setting Ψi=YiA¯iT+A¯iYiT+∑j=1mβij(EiTXi-EjTXj), the result of Lemma 2 can be evidently derived from the fact that(23)Λi·ΨiXiT-Yi+A¯iYiTGi-XiTHiTXi-YiT+YiA¯iT-Yi-YiT0GiT-HiXi02ηI-Di-DiT·ΛiT=XiTA¯iT+A¯iXi+∑j=1mβijEiTXi-EjTXjGi-XiTHiTGiT-HiXi2ηI-Di-DiT.
Remark 3.
The instrumental matrix variable Yi is introduced to decouple Lyapunov inverse matrix Xi and controller gain matrix Ki, which makes the design of output feedback passive controllers feasible, and, at the same time, Xi and Xi-1 do not require decomposition or fixing to a certain structure.
Based on the above lemma, an LMI condition is presented, under which system (5) is stable and passive in the following theorem.
Theorem 4.
If there exist simultaneously nonnegative real number βij,η and matrices Zi>0, Zj>0, Yi,Wi,Ui,Uj such that for any i,j∈χ, i≠j,(24)Θi11Θi12Θi13Θi12T-Yi-YiT0Θi13T02ηI-Di-DiT<0,where Θi11=YiAiT+AiYiT+BiWi+WiTBiT+∑j=1mβij(EiTZiEiT+EiTLiUi-EjTZjEjT-EjTLjUj), Θi12=EiZiT+UiTLiT-Yi+AiYiT+BiWi, Θi13=Gi-EiZiTHiT-UiTLiTHiT, Yi=Yi11Yi120Yi22, and Wi=Wi10, then, system (5) is stable and passive with dissipation rate η under static output feedback(25)Ki=Wi1Yi11-Tvia switching signal(26)i=argmaxxTtEiTZiEiT+LiUixt,i∈χ.
Proof.
Since Ci=[Ip0] in A¯i=Ai+BiKiCi, then BiKiCiYiT=Bi[Ki0]YiT. Letting Yi=Yi11Yi120Yi22, one gets YiT=Yi11T0Yi12TYi22T. Thus, Bi[Ki0]YiT=Bi[KiYi11T0]. Let Wi=[Wi10], Wi1=KiYi11T. A static output feedback gain is denoted by Ki=Wi1Yi11-T.
By making use of the existent methods (e.g., [31, 32]), we introduce the matrix Li∈Rn×(n-ri) with rank Li=n-ri satisfying EiLi=0. By setting Xi=ZiEiT+LiUi, one obtains that Xi-TEiT=EiXi≥0 holds when Zi>0. Then, according to Lemma 2, inequality (6) and (7) can be rewritten as inequality (24). Inequality (24) is a linear matrix inequality. This completes the proof.
Remark 5.
If βij is nonpositive simultaneously, Theorem 4 also holds through choosing the switching signal as i=argmin{xT(t)EiT(ZiEiT+LiUi)x(t),i∈χ}.
Remark 6.
When switched singular system reduces to a single singular system (i.e., no switching), Theorem 4 also holds by setting Ai=A, Bi=B, Gi=G, and so forth. This will be illustrated in Corollary 7. When switched singular system reduces to a normal switched system (i.e., Eci=I), the method introduced in this paper is also available, and, at the same time, Theorem 4 is rewritten as Corollary 8.
Corollary 7.
If there exist a real number η>0 and matrices Z>0,Y,W,U such that(27)Θ11Θ12Θ13Θ12T-Y-YT0Θ13T02ηI-D-DT<0,where Θ11=YAT+AYT+BW+WTBT, Θ12=EZT+UTLT-Y+AYT+BW, Θ13=G-EZTHT-UTLTHT, Y=Y11Y120Y22, and W=W10, then, the corresponding singular system is stable and passive with dissipation rate η under static output feedback(28)K=W1Y11-T.
Corollary 8.
If there exist simultaneously nonnegative real number βij,η and matrices Xi>0, Xj>0, Yi,Wi such that for any i,j∈χ, i≠j,(29)Θi11Θi12Θi13Θi12T-Yi-YiT0Θi13T02ηI-Di-DiT<0,where Θi11=YiAiT+AiYiT+BiWi+WiTBiT+∑j=1mβij(Xi-Xj), Θi12=XiT-Yi+AiYiT+BiWi, Θi13=Gi-XiTHiT, Yi=Yi11Yi120Yi22, and Wi=Wi10, then, the corresponding switched system is stable and passive with dissipation rate η under static output feedback(30)Ki=Wi1Yi11-Tvia switching signal(31)i=argmaxxTtXixt,i∈χ.
4. Example
Consider the switched singular system composed of two subsystems(32)Ecix˙t=Acixt+Bciut+Gciωtzt=Hcixt+Dciωtyt=Ccixti=1,2,where(33)Ec1=1000,Ec2=1000,Ac1=5-210-15,Ac2=-6-1-11.5,Bc1=-32.5,Bc2=-42,Cc1=21,Cc2=-12,Gc1=01,Gc2=01,Dc1=2,Dc2=1,Hc1=11,Hc2=12.Choose similarity transformation matrices as(34)T1=21-1-1,T2=-121-1.Then, we get an equivalent state-space description of the above system:(35)Eix˙t=Aixt+Biut+Giωtzt=Hixt+Diωtyt=Cixt,i=1,2,where(36)E1=T1Ec1T1-1=22-1-1,E2=T2Ec2T2-1=-1-212,A1=T1Ac1T1-1=3958-32-49,A2=T2Ac2T2-1=812-7.5-12.5,B1=T1Bc1=-3.50.5,B2=T2Bc2=8-6,C1=Cc1T1-1=10T,C2=Cc2T2-1=10T,D1=Dc1=2,D2=Dc2=1,G1=T1Gc1=1-1,G2=T2Gc2=2-1,H1=Hc1T1-1=0-1,H2=Hc2T2-1=34.Choose β12=β21=0.5, L1=1-1, L2=-21. By solving linear matrix inequality (24) in Theorem 4, we can obtain(37)Z1=6.11071.65281.65280.9175,Z2=2.0404-1.1790-1.17900.9889,Y1=0.6280-0.079100.0364,Y2=2.7362-1.390200.5073,W1=10.04250,W2=-2.24750,X1=19.9846-6.81140.6831-3.5225,X2=7.5201-3.3638-4.40002.3218,X1-1=0.0536-0.10360.0104-0.3040,X2-1=0.87291.26471.65432.8273.By linear matrix inequality (24), output feedback passive control gains can be obtained:(38)K1=15.9905,K2=-0.8214.The dissipation rate can be obtained:(39)η=0.5.
The corresponding switching signal is chosen as(40)i=1,xt∈Ω12,xt∈Ω2Ω1,Ω1=x∈Rn∣xTE1TX1-E2TX2x≥0,xTE1TX1-E2TX2E1x≠0,E2x≠0Ω2=x∈Rn∣xTE2TX2-E1TX1x≥0,E1TvdE1TvdE1x≠0,E2x≠0.Under the switching signal, the response curve of the above system is exhibited in Figure 1, where ωt=sinte-0.1t. From Figure 1, it is obvious that the resulting closed-loop system is stable.
State response of the corresponding closed-loop system.
5. Conclusion
In this paper, the output feedback passive control problem for a class of switched singular systems is investigated. A novel method is proposed to solve static output feedback passive controllers. Sufficient linear matrix inequality condition is presented by means of introducing instrumental matrix variable Yi and multiple Lyapunov function technique under designed switching law. An example is given to verify the LMI condition proposed for the resulting closed-loop system to be stable and passive with a lower dissipation rate η. Passivity is a suitable design approach in network control. How to extend the results of this paper to network-based control is an interesting problem. This problem deserves a further study and it remains as our future work.
Conflict of Interests
The authors declare that there is no conflict of interests regarding the publication of this paper.
Acknowledgments
This work is supported by the National Natural Science Foundation of China under Grants nos. 61203001, 61104066, and 61473140, Liaoning Educational Committee Foundation under Contract L2014525, and the Natural Science Foundation of Liaoning Province under Grant no. 2014020106.
CantóB.CollC.SánchezE.Positive N-periodic descriptor control systems200453540741410.1016/j.sysconle.2004.05.0172-s2.0-7444257432XiongW.HoD. W. C.CaoJ.Synchronization analysis of singular hybrid coupled networks2008372446633663710.1016/j.physleta.2008.09.0302-s2.0-53149096505XiaY. Q.BoukasE.-K.ShiP.ZhangJ. H.Stability and stabilization of continuous-time singular hybrid systems200945615041509MR28794562-s2.0-6734924470910.1016/j.automatica.2009.02.008ZhaiG. S.XuX. P.HoD. W. C.Stability of switched linear discrete-time descriptor systems: a new commutation condition2012851117791788MR29802522-s2.0-8486702630210.1080/00207179.2012.703788LiberzonD.TrennS.On stability of linear switched differential algebraic equationsProceedings of the 48th IEEE Conference on Decision and Control (CDC '09)December 2009215621612-s2.0-7795084625710.1109/cdc.2009.5400076ClotetJ.FerrerJ.MagretM. D.Switched singular linear systemsProceedings of the 17th Mediterranean Conference on Control and Automation (MED '09)June 2009Thessaloniki, Greece1343134710.1109/med.2009.5164733LongF.XiaoY.ChenX.ZhangZ. H.Disturbance attenuation for nonlinear switched descriptor systems based on neural network2013237-82211221910.1007/s00521-012-1171-32-s2.0-84887425844ZamaniI.ShafieeM.On the stability issues of switched singular time-delay systems with slow switching based on average dwell-time201424459562410.1002/rnc.29032-s2.0-84892535303MR3160375MengB.ZhangJ.-F.Reachability conditions for switched linear singular systems200651348248810.1109/tac.2005.864196MR22056872-s2.0-33645009873DingX. Y.LiuX.ZhongS. M.Delay-independent criteria for exponential admissibility of switched descriptor delayed systems20142284324452-s2.0-84891044853MR315193010.1016/j.amc.2013.11.107ZhouL.HoD. W.ZhaiG.Stability analysis of switched linear singular systems20134951481148710.1016/j.automatica.2013.02.002MR30440312-s2.0-84876689594ShiS. H.ZhangQ. L.YuanZ. H.LiuW. Q.Hybrid impulsive control for switched singular systems20115110311110.1049/iet-cta.2009.04442-s2.0-79551715077LinJ. X.FeiS. M.WuQ.Reliable H∞ filtering for discrete-time switched singular systems with time-varying delay20123131191121410.1007/s00034-011-9361-2MR29129252-s2.0-84860637719ZhangD.YuL.WangQ.-G.OngC.-J.WuZ.-G.Exponential H∞ filtering for discrete-time switched singular systems with time-varying delays201234972323234210.1016/j.jfranklin.2012.04.006MR29606102-s2.0-84864666417KoenigD.MarxB.JacquetD.Unknown input observers for switched nonlinear discrete time descriptor systems2008531373379MR239159310.1109/tac.2007.9142262-s2.0-65949091343LiQ.ZhangQ. L.YiN.YuanY. H.Robust passive control for uncertain time-delay singular systems200956365366310.1109/tcsi.2008.2002663MR25912432-s2.0-63449091969WangJ.ShenH.Passivity-based fault-tolerant synchronization control of chaotic neural networks against actuator faults using the semi-Markov jump model approach2014143515610.1016/j.neucom.2014.06.022GaoH. J.ChenT. W.ChaiT. Y.Passivity and passification for networked control systems2007464129913222-s2.0-4824912344610.1137/060655110MR2346383LiuK.FridmanE.Wirtinger's inequality and Lyapunov-based sampled-data stabilization201248110210810.1016/j.automatica.2011.09.029MR28794162-s2.0-84355161690LiuK.FridmanE.HetelL.Stability and L2-gain analysis of networked control systems under Round-Robin scheduling: a time-delay approach201261666675FengZ.LamJ.Robust reliable dissipative filtering for discrete delay singular systems20129212301030252-s2.0-8486428134210.1016/j.sigpro.2012.06.003WuZ.-G.ParkJ. H.SuH.SongB.ChuJ.Mixed H∞ and passive filtering for singular systems with time delays20139371705171110.1016/j.sigpro.2013.01.0032-s2.0-84874585747XiaZ. L.LiJ. M.LiJ. R.Passivity-based resilient adaptive control for fuzzy stochastic delay systems with Markovian switching201435173818383610.1016/j.jfranklin.2013.03.016MR32149452-s2.0-84877112314HeS.LiuF.Optimal finite-time passive controller design for uncertain nonlinear Markovian jumping systems201435173782379610.1016/j.jfranklin.2013.03.006MR32149432-s2.0-84877120957YaoX. M.GuoL.Composite disturbance-observer-based output feedback control and passive control for Markovian jump systems with multiple disturbances201481087388110.1049/iet-cta.2013.0659MR3235565GeromelJ. C.ColaneriP.BolzernP.Passivity of switched linear systems: analysis and control design201261454955410.1016/j.sysconle.2012.02.0082-s2.0-84859547584JiaH. M.XiangZ. R.KarimiH. R.Robust reliable passive control of uncertain stochastic switched time-delay systems201423125426710.1016/j.amc.2014.01.020MR31740282-s2.0-84893370252WangG. L.ZhangQ. L.YangC. Y.Dissipative control for singular Markovian jump systems with time delay201233441543210.1002/oca.1004MR29578302-s2.0-84864468956WuZ.-G.ParkJ. H.SuH.ChuJ.Reliable passive control for singular systems with time-varying delays20132381217122810.1016/j.jprocont.2013.07.0092-s2.0-84882768944LiH.GaoY.WuL.LamH. K.Fault detection for T-S fuzzy time-delay systems: delta operator and input-output methods201545222924110.1109/tcyb.2014.23239942-s2.0-84901866003ZhangL. Q.HuangB.LamJ.LMI synthesis of H2 and mixed H2/H∞ controllers for singular systems200350961562610.1109/tcsii.2003.8169042-s2.0-0141954040UezatoE.IkedaM.Strict LMI conditions for stability, robust stabilization, and H∞ control of descriptor systems4Proceedings of the 38th IEEE Conference on Decision and Control (CDC '99)December 1999Phoenix, Ariz, USAIEEE409240972-s2.0-003331227310.1109/CDC.1999.828001