递推辨识中的奇异值分解方法

Singular Value Decomposition Method in the Recursive Identification

研究了递推辨识算法的一种实现形式—SVD分解算法。这样就可以在计算量基本不变的情况下,有效地保证方差矩阵的对称性,正定性,从而获得较好的数值特性。仿真计算结果表明在实现自适应控制时,能提高系统参数实时辨识的精度。

This paper suggests a realize form in the recursive identification methods, based on singular value decomposition. The method proposed here can not only guarantee the symmetry and definite of the covariance matrix but also get good numerical stability. The simulation results confirm the theoretical results.