摘要：AbstractThis paper summarizes a set of Matlab tools called CoBRA to identify discrete-time multivariable systems, possibly operating under closed-loop or controlled conditions, on the basis of a subspace method that explicitly uses covariance functions estimated from the time domain data. The use of covariance functions is motivated by the approximation of compact support for covariance functions, allowing the size of the matrices required in the subspace method to be limited, even when a very large number of data points is available for identification. In addition, the CoBRA tools allow information of possible pole locations of the multivariable system to be included via user-defined convex regions in the complex plane. The constraints on the pole locations are created using linear-matrix-inequality regions that are solved with existing Matlab tools for Semi-Definite programming integrated within the CoBRA tools.
