Recursive System Identification Using Fast Algorithms

摘要

This article shows how the class of so-called 'fast' (i.e. computationally efficient) estimation algorithms can be used in system identification. Commencing with a simplified derivation of a fast recursive parameter estimator, the discussion continues to cover the following points: The use of fast algorithms for the general model structures encountered in system identification from input/output data; The computational advantages and disadvantages of these allegedly computationally efficient methods; The theoretical basis for fast instrumental variable algorithms; and, The initialization and use of exponential memories in fast recursions. The article continues with practical examples of 'fast' identifiers at work in the least squares form (applied to feedback systems) and in the instrumental variable form.