递阶辨识原理与迭代方法族.pdf

Proceedings of the 25th Chinese Control Conference 7–11 August, 2006, Harbin, Heilongjiang 1, 2, 2 1. , 214122 E-mail: fding@ 2. , , 330034 E-mail: limingniat@, djiyang@ : .: :,, AXB = C , ,, , , , . AXB + CXD = F , , , . Hierarchical Identi?cation Principle and a Family of Iterative Methods Feng Ding1, Ming Li2, Jiyang Dai2 1. Control Science and Engineering Research Center Southern Yangtze University, Wuxi 214122 E-mail: fding@ 2. Key Laboratory of Nondestructive Test (Ministry of Education), Nanchang Institute of Aeronautical Technology, Nanchang, P.R. China 330034 E-mail: limingniat@, djiyang@ Abstract: In this paper, we extend the well-known Jacobi and Gauss-Seidel iterations to present a large family of iterative methods. The proposed methods are applied to develop iterative solutions to the matrix equation AXB = F and the generalized Sylvester matrix equation AXB + CXD = F by means of a hierarchical identi?cation principle. We prove that the iterative solutions converge to the exact solutions for any initial values. The algorithms proposed require less storage capacity than the existing numerical ones. The iterative methods can be applied to system parameter identi?cation problems. Key Words: Sylvester matrix equation, Lyapunov matrix equation, identi?cation, estimation, least squares, gradient iteration, Jacobi iteration, Gauss-Seidel iteration, hierarchical identi?cation principle. 1 (Introduction) , (Sylvester) (Lyapunov) ., : AX + XB = F , AX + XAT = F , AXET + EXAT = F , m × n , A, B, E, F T : : ,X . uct), mn (Kronecker prod- Ax = b ., , A ∈ Rmn×mn IEEE Catalog Number: 06EX1310 , : . . . , m = n = 100, mn × mn = 10, 000 × 10, 000 [1], , Syrmos [6], , ., [2, 3], Hessenberg-Schur 9, 10]; Borno [11]. , . [4, 5]. (SVD) , [7, 8, , [12, 13, 14, 15]. , 418 Ax = b (Jacobi) (Gauss-Seidel) [13]; Starke Niethammer (SOR) [16]; Jonsson and Ka¨gstro¨m - [17, 18]; Mukaidani [15]. , – ,. . 19, 20]. , , ., , , , [14, 15, :, , [21, 22, 23], . , X , 2Ga