In this note, an expansion for the determinant of the sum of two matrices, det(A +B) is given. A robustness analysis approach for state space systems with uncertain parameters is investigated by using Kharitonov's theorem.
In this note, an expansion for the determinant of the sum of two matrices, det(A +B) is given. A robustness analysis approach for state space systems with uncertain parameters is investigated by using Kharitonov's theorem.
In this note, an expansion for the determinant of the sum of two matrices, det(A +B) is given. A robustness analysis approach for state space systems with uncertain parameters is investigated by using Kharitonov's theorem.
1671 M. Hou, Th. Schmidt, R. Schupphaus, and P. C. Muller, Normal form and Luenberger observer for linear mechanical descriptor systems, 1. Dyn. Syst., Measurement, Contr., to be published. F. L. Lewis, A tutorial on the geometric analysis of linear time- invariant implicit system, Automatica, vol. 28, pp. 119-137, 1992. M. Hou and P. C. Muller, A singular matrix pencil theory for linear descriptor systems, Symposiumon Implicit and Nonlinear Systems, Ft. Worth, Texas, Dec. 14-15, 1992. P. Van Dooren, The computation of Kroneckers canonical form of a singular pencil, LinearAlgebra its Appl., vol. 27, pp. 103-140, 1979. Expansion of det(A +B ) and Robustness Analysis of Uncertain State Space Systems S. J . Xu, M. Darouach, and J. Schaefers Abstract-In this note, an expansion for the determinant of the sum of two matrices, det ( A +B) , is given. A robustness analysis approach for state space systems with uncertain parameters is investigated by using Kharitonovs theorem. Some useful results are obtained. Several illustrative examples are given. I. INTRODUCTION The stability problem of a family of polynomials has attracted much attention in recent literature (see [1]-[27]), in which the most notable result is Kharitonovs theorem introduced in con- trol literature by Barmish [2], and Bose and Zeheb [3]. However, there exist three restrictions in the application of Kharitonovs theorem; 1) it is only applicable to the interval polynomials, 2) it only considers the strict Hunvitz property of polynomials, and 3) it assumes that all of the coefficients of polynomials vary inde- pendently. More recently, many authors have become interested in the activity of generalization of Kharitonovs theorem. The strict Hunvitz property and the more general D-stability (i.e., the zeros of polynomials exist inside the special region D on the complex plane, where, D can be the left-half plane, the unit disk, or other special region) for the general polytopes of poly- nomials have been discussed in [4]-[14]. The edge Theorem is given [4]. Some corresponding frequency domain conditions are obtained by means of the Hermite-Biehler theorem in [lS], [16]. The studies indicate that the minimum number of polynomials needing to be checked is four for 6-order or more than 6-order systems. For the systems with orders 3, 4, and 5, [17] indicates that the polynomials needing to be checked are 1, 2, and 3, respectively. For the same problem, [ 181gives the corresponding frequency domain results. In practice, the majority of uncertain systems have the characteristic polynomials with dependent co- efficients. I n this case, Kharitonovs theorem degenerates as a sufficient condition. Using this theorem, one will obtain a con- servative robustness estimation. For the stability test of the polynomials with linearly and multilinearly dependent coeffi- Manuscript received J anuary 31, 1992; revised August 28, 1992. S. J . Xu is with the CRPHT 6, rue Coudenhove-Kalergi, L-1359 Luxembourg, the LARAL CRAN CNRS UA-821, Route de Romain, 54400 Longwy, France and The Lab. of Flight Dynamics, Harbin Institute of Technology, China. M. Darouach is with the LARAL CRAN CNRS UA-821, Route de Romain, 54400 Longwy, France. J . Schaefers is with the CRPHT 6, rue Coudenhove-Kalergi, L-1359 Luxembourg. IEEE Log Number 9208724. cients, [4], [19]-[21] propose some useful methods to reduce the conservativity, but for those with nonlinearly dependent coeffi- cients, there is no general method to deal with this problem. One must concretely deal with his/her problem based on the specific situation. I n this case, the approach proposed by Wei and Yedavalli [22] and Pujara [23] may be useful, but, generally speaking, the conservativity cannot be eliminated. In this note, we give a determinant expansion of the sum of WO matrices, det(A +B), which is used to obtain the character- istic polynomials of uncertain state space systems. Kharitonovs theorem is applied to test the Hunvitz property of corresponding polynomials. Some useful results are obtained and several illus- trative examples are presented. 11. KHARITONOVS THEOREM AND EXPANSION OF det(A +B ) Let us recall Kharitonovs theorem. Let the interval polynomial be represented by 6(s) =6,) +6,s +6,s2 +.. + 8 n S n x, I 6, 5 y , . (1) ( 2 ) Lemma 1 (Khantonor 1978): The polynomial (1)-(2) is Hunvitz with if and only if the following four polynomials are Hunvitz: Sl (s) =y, , + X I S +x2s2 +y,s +y4s4 +x5s5 +X 6 S 6 +.. + a*($) = y o +y , s +x*s2 +X3S3 +y,s4 +y5s5 +X 6 S 6 +... + 6s) =xi , + X I S +y,s2 +Y3S +x4s4 +x5s5 +y,s6 +.. + 64(s) =x,, +yls +y2s2 +X 3 S 3 +x4s4 +y,s5 +y,s6 +... +. (3a) (3b) (3c) (3d) Definition 1: r; det ( A / B ) is defined as a sum of the combi- nation of determinants, in which the i rows of A are substituted by the corresponding rows of B. For example, let A ={arj > and B ={b,,] E R3x 3, then Pro06 The various proofs can be found in [24]-[27]. ri det(A/B1) =det a31 a32 a33 a31 a32 a33 +det[i i l ;:; ;;] a13 r:det(A/B*) =det a31 32 33 +det[: ;: ; : I . b,, 4 2 b33 Lemmu 2: Let A , B E R, then n- 1 det(A +B ) =det(A) +det(B) + r,:det(A/B ) (4) i = 1 0018-9286/93$03.00 0 1993 IEEE 1672 IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 38, NO. 1 I , NOVEMBER 1993 or n det ( A +B ) = r,: det ( A/ B' ) (5 ) i = 0 where det (.) represents the determinant of (.). lemma see Appendix A. For the properties of r: det ( A/ B' ) and the proof of this 111. ROBUSTNESS ANALYSIS OF UNCERTAIN STATE SPACE SYSTEMS Consider the following uncertain system: X ( t ) =( A +AA) X( t ) (6) where X E R", A is a constant matrix and AA an uncertain matrix with lAAl 5 D, D is a nonnegative matrix,)(.)l represents the module matrix of (.I. Theorem I : System (6) is asymptotically stable if the following polynomials are Hunvitz: 6I (s) =a, +bl s +b,s2 +u3s3 +ads4 +bss5 +b,s6 +... + (7a) S * ( S ) =U, +U, S +b,s2 +b3s3 +ads4 +asss +b,s6 +... + (7b) (7c) (7d) (8a) (8b) Proof: By using Lemma 2, the characteristic polynomial of S 3 ( s ) =bo +b, s +a2s 2 +u3s3 +bds4 +bsss +ahsh +... + S4( s) =b, +a,$ +a2s 2 +b3s3 +b4s4 +asss +U,S' +... + where ai =r;det[-(A +AA)/I J J ma b, =ri det [ - ( A +AA)/l'],in and I is an identity matrix. (6) can be given by det[sZ - ( A +AA11 =det [ - ( A +AA) +sZ1 = n r,: det [ - ( A +AA)/(sZ)'l r',: det [ - ( A +AA)/ l ' l s' 1 = O n = =det [ - ( A +AA)] i = U +rddet [ - ( A +AA)/Z1ls +... +r;-I det[-(A +AA) / I n- l l s n- l +s" from notation (8) and Kharitonov's theorem, the theorem is proved. Remark 1: The coefficients of the above characteristic poly- nomial are not independent, so the theorem gives a sufficient condition of the stability of system (6). Definition 2: Let A ={ai j } E R"'", then S det(A) E R is defined as a projection from RnX" to R, the Laplace expansion of S det ( A ) is defined as: n n S det ( A ) = (a,,lS det ( Ai j ) = (n,lS det (Ai,) (9) where A, , is the ij-th submatrix of A resulting from the deletion of row i and column j , for example, S det [ al l ] =lalll i = 1 j = 1 Sdet [ "121 =lull1 la*J +lUl*l lu211. a21 a22 The following relationships are obvious: S det ( A ) 2 det ( A ) (10) Sdet(- A) =Sdet(A). (11) Theorem 2: System (6) is asymptotically stable if the following 6'(s) =e, +fl s +f 2 s 2 +e3s3 +e4s 4 +fsss +f 6s 6 +... + polynomials are Hunvitz: (12a) s2(s) =e, +el s +f 2 s 2 +f 3s 3 +eds4 +esss +f 6 s 6 +... + (12b) ~3 ( s ) =f,, +fls +e2s2 +e3s3 +f4s4 +f 5 s 5 +e6s6 +... + (12c) s4(s) =f, +el s +e2s 2 +f 3s 3 +f4s4 +esss +e& +... + (12d) with det ( - A, ) +S det ( D, ) n - r - + e ,=1 fi =r,: det i n - r - - c j = 1 where A, and D, are the submatrices of A and D resulting from the deletion of the i rows and corresponding i columns, i.e., A, and D, are ( n - i )(n - i)-dimension matrices, respectively. Proof: From Theorem 1 and Lemma 2, we have U, =r,: det [ - ( A +AA)/I'l,,, =r,l det [ - A, - AA,l,,, =r; det(-A,) +det(-AA,) n - 1 - 1 i - i + r;-$ det [ -A, / (-AA, )' ] I = 1 <r,: det(-A,) +Sdet(D,) ,=1 b, =r; det [ - ( A +AA)/Z'l,,, =r,l det [ - A, - AA,l,,, det(-A,) +det(- AA,) IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 38, NO. 1 I , NOVEMBER 1993 1673 hence f, I b, I a, I e,, i.e., interval [f, e,] 2 [b, a,]. Therefore, if (12) are Hurwitz, which implies that (7) are also Hurwitz, according to Theorem I, the system (6) is asymptotically stable. For the low order systems ( n I 5), we have the following corollaries. Corollary 1: (n =2) The system (6) is asymptotically stable if f, >0, i =0, I. Where f o =det ( - A ) - S det ( D) - S det (A/dl ) - S det ( A / d , ) (14a) and fl =tr(- A) - tr( D) (14b) tr(.) denotes the trace of (.) and (A/d,) denotes the submatrix of A resulting from the substitution of the i-th row by, d,, the i-th row of D. Proof: We know that, for the four 2-order polynomials of (12), the stability conditions are that all of the coefficients of polynomials are positive, if f, >0 is true, then the above conditions are satisfied. Corollary 2: ( n =3, 4, and 5). The system (6) is asymptotically stable if f, >0 and the following polynomials are Hunvitz, respectively 1) for n =3 ~ ' ( s ) =e,, +f l s + f 2 s 2 +s3 (15) 2) for n =4 6I (s) =e, +f,s +f , s 2 +e3s3 +s4 S2(s) =eo +e, s +f 2s2 +f 3 s 3 +s 4 (16a) (16b) 3) for n =5 S1(s) =e, +f,s +f 2 s 2 +e3s3 +e4s 4 +s 5 (17a) S2(s) =e , +el s +f 2 s 2 + f 3 s 3 +e4s 4 +s5 (17b) S4( s ) = f o +e,s +e2s 2 +f 3 s 3 +f4s4 +s5 ( 17~) where f, and e, are defined by (13). corollary is proved immediately. Proof: By noting Theorem 2 and the results of [17], the IV. NUMERICAL EXAMPLES Example 1: Consider the system described by (6), where det ( - A ) =14, S det ( D ) =1.25, S det ( A / d , 1 =7, S det (A/d,) =4.75 tr ( - A ) =9, it is obvious that the condition of Corollary 1 is satisfied, therefore the system is stable. tr( D) =2 Example 2: Consider the system (61, where where q1 and q2 are the uncertain parameter respectively, we will give out the allowable bounds of q1 and q2 which the system holds stable. We have tr ( - A) =6, tr ( D) =lq21 the conditions for which Corollary 1 is true are: 1411<10, l q2l <6. V. CONCLUSIONS In this note, we have given an important determinant expan- sion of sum matrices, det ( A +B). Some useful robustness crite- ria are derived for uncertain state space systems. The results obtained are still conservative, the level of conservativity relies on dependence of coefficients. The examples show that the robustness estimation of Example 1 is conservative because of the nonlinearly dependent coefficients resulting from uncertain matrix AA. But, in the Example 2, these coefficients are inde- pendent, therefore, the robustness estimation is necessary and sufficient. How to reduce the conservativity of robustness analy- sis for the polynomials with nonlinearly dependent coefficients, is still a problem to be studied in the future. APPENDIX A The properties of combination ri det ( A/ B' ) : 1) r; det ( A/ B' ) =r;-' det ( B/ A' ) (A-1) 2) ri 'det(A/Bo) =det(A) (A-2) (A-3) 3) r; det ( A / B " ) =det ( B) if B =I . then 6) det(A/I ") =1. (A-6) The proof of lemma 2: This lemma can be proved by the induction. I n fact, from the Laplace expansion of determinant of A +B directly, we know that the lemma is true for n =2 and n =3. We assume that it is true for n =k , and then prove that it is true for n =k +1. If n =k, then k - I det ( A +B ) =det ( A ) +det ( B) + r; det ( A/ B' ) 1 = 1 for n =k +I, the expansion of det ( A +B) based on the final row of A +B is given by 1674 IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 38, NO. 11, NOVEMBER 1993 where the deletion of the i-th row and j-th column of (.). Then {a,+,,,det[(A +B ) ~+I , I ] represents the ij-th submatrix of (.) resulting from k + l + l det(A +B ) =(-1) +bk+l , l det[(A +B)k+i ,i ]} +... + ( - l ) k + l + h { a k + l , h det[(A +B)k+i ,hI + ( - l ) k + l + k { a k + l , k det[(A +B)k+i ,kI +bk+l ,kdet[(A +B)k+i ,kl ) +~, +~, ~+~d et [ ( A +B)k+i ,k+i I +bk+l , k+l det[(A +B)k+i ,k+i I +bk+l ,h det [ ( A +B)k+l .h]} +'.. hence =r;+det ( A / B ' ) ... ... IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 38, NO. 11, NOVEMBER 1993 1675 REFERENCES V. L. Kharitonov, Asymptotic stability of an equilibrium posi- tion of a family of systems of linear differential equations, Difjer- entialnye Uracneniya, vol. 14, pp. 2086-2088, 1978. B. R. Barmish, Invariance of the strict Hunvitz property for polynomials with perturbed coefficients, IEEE Trans. Automat. Contr., AC-28, pp. 935-937, 1984. N. K. Bose and E. Zeheb, Kharitonovs theorem and stability test of multidimensional digital filters, IEE Proc. G, vol. 133, pp. A. C. Bartlett, C. V. Hollot, and H. Lin, Root locations of an entire polytope of polynomials: It suffices to check the edges, Math. Contr. Signals. Syst., vol. 1, pp. 61-71, 1988. B. R. Barmish, A generalization of Kharitonovs four-polynomials concept for robust stability problems with linearly dependent coefficients, IEEE Trans. Automat. Contr., vol. 34, pp. 157-165, 1989. H. Chapellat and S. P. Bhattacharyya, A generalization of Kharitonovs theorem: Robust stability of interval plants, IEEE Trans. Automat. Contr., vol. 34, pp. 306-311, 1989. Y. C. Soh, Stability of an entire polynomials, Inf. J. Contr., vol. 49, pp. 993-999, 1989. M. Fu and B. R. Barmish, Polytopes of polynomials with zero in a prescribed set, IEEE Trans. Automat. Contr., vol. 34, pp. 544-546, 1989. I. R. Petersen, A new extension to Kharitonovs theorem, IEEE Trans. Automat. Contr., vol. 35, pp. 825-828, 1990. A. Cavallo, G. Celentano, and G. de Maria, Robust stability analysis of polynomials with linearly dependent coefficient pertur- bations, IEEE Trans. Automat. Contr., vol. 36, pp. 380-384, 1991. I. R. Petersen, A class of stability regions for which a Kharitonov-like theorem holds, IEEE Trans. Automat. Contr., Y. C. Soh, Strict Hunvitz property of polynomials under coeffi- cient perturbation, IEEE Trans. Automat. Contr., vol. 34, pp. 629-632, 1989. Y. K. Foo and Y. C. Soh, A generalization of strong Kharitonovs theoremto polytopes of polynomials, IEEE Trans. Automat. Contr., vol. 35, pp. 936-939, 1990. -, Kharitonovs regions: It suffices to check a subset of vertex polynomials, IEEE Trans. Automat. Contr., vol. 36, pp. 1102-1105, 1991. M. B. Argoun, Frequency domain conditions for the stability of perturbed polynomials, IEEE Trans. Automat. Contr., vol. 32, pp. 913-916, 1987. S. Dasgupta, P. J. Parker, B. D. 0. Anderson, F. J . Kraus, and M. Mansour, Frequency domain conditions for the robust stabil- ity of linear and nonlinear dynamical systems, IEEE Trans. Circ. B. D. 0. Anderson, E. I. J ury, and Mansour, On robust Hunvitz polynomials, IEEE Trans. Automat. Confr., vol. 32, pp. 909-913, 1987. M. B. Argoun, On the stability of low-order perturbed polynomi- als, IEEE Trans. Automat. Contr., vol. 35, pp. 180-182, 1990. F. Kraus, B. D. 0. Anderson, and M. Mansour, Robust stability of polynomials with multilinear parameter dependence, Int. J. Contr., vol. 50, pp. 1745-1762, 1989. B. R. Barmish and Z. C. Shi, Robust stability of a class of polynomials with coefficients depending multilinearly on perturba- tions, IEEE Trans. Automat. Contr., vol. 35, pp. 1040-1043, 1990. J. Ackermann, H. Z. Hu, and D. Kaesbauer, Robustness analysis: A case study, IEEE Trans. Automat. Contr., vol. 35, pp. 352-356, 187-190, 1986. vol. 34, pp. 1111-1115, 1989. Syst., vol. 38, pp. 389-397, 1991. [26] [27] K. S. Yeung and S. S. Wang, A simple proof of Kharitonovs theorem, IEEE Trans. Automat. Contr., vol. 32, pp. 822-823, 1987. H. Chapellat and S. P. Bhattacharyya, An alternative proof of Kharitonovs theorem, IEEE Trans. Automat. Contr., vol. 34, pp. F. R. Gantmacher, Theory of Matrices, vol. 11, New York Chelsea, 1964. 448-450, 1989. [28] Nonlinear Control of Induction Motors: Torque Tracking with Unknown Load Disturbance Romeo Ortega, Carlos Canudas, and Seleme I. Seleme Abstrecf-In a recent note Ortega and Espinosa [SI presented a globally stable controller for torque regulation of a complete induction motor model with partial state feedback, i.e., no assumption of flux measurement. The result was established under the assumptions that both the desired and load torques are constant, that the former does not exceed certain bounds which depend on the systems natural damping, and that the motor parameters are known. In the present contributions we extend these results in several directions. First, by adding mechani- cal damping to the closed-loop system we relax the upper bound condition on the desired torque. Second, we use a new controller struc- ture that allows us to treat the case of time-varying desired torque. Finally, a new estimator is proposed to handle time-varying (linearly parameterized) unknown loads. I. PROBLEM FORMULATI ON We consider in this note the classical dq model [lo] of the induction motor (1.1) (1.2) with generated torque 1990. 1221 K. H. We1 and R. K. Yedavalli, Invariance of strict Hunvitz property for uncertain polynomials with dependent coefficients, IEEE Trans. Automat. Contr., vol. 32, pp. 907-909, 1987. [231 L. R. Pujara, On the stability of uncertain polynomials with dependent coefficients, IEEE Trans. Automat. Contr., vol. 35, pp. 756-759, 1990. 1241 N. K. Bose, A systemtheoretic approach to stability of sets of polynomials, Contemporary Math., vol. 47, pp, 25-34, 1985. [251 N. K. Boseand Y. Q. Shi, A simple general proof of Kharitonovs generalized stability criterion, IEEE Trans. Circults Syst., vol. 34, pp. 1233-1237, 1987. Manuscript received March 14, 1992; revised August 21, 1992. R. Ortega is with Genie Informatique, Universite de Technologie de Compiegne, BP 649-60206, Cedex, France. He was a Visiting Professor at the Department of Electrical Engineering, McGill University, Mon- treal, Canada, when this work was completed. C. Canudas is with the Laboratoire DAutomatique de Grenoble, ENSIEG, BP 46, 38402, Saint Martin DHeres, France. S. I. Seleme is on leave fromFaculdad de Engenharia de Joinville, UDESC, Brazil and iscurrently with the Laboratoire DAutomatique, de Grenoble, ENSIEG, BP 46, 38402, Saint Martin DHeres, France. His work was supported in part by CNPq/CEFI, Brazil-France. IEEE Log Number 9208709. 0018-9286/93$03.00 0 1993 IEEE