Online parameter estimation using matrix pseudoinverse

A method based on the matrix pseudoinverse is presented for the online identification of discrete-time systems of known order. Recursive algorithms are described which provide minimum-norm estimates of the parameter vector when insufficient data are available, and least-squares estimates with adequate data. These estimates can be updated easily with each pair of additional input-output data, as matrix inversion is not required. When the order of the system is not known, it may be determined offline using one of the two methods described. A recursive algorithm for calculating the residual error is also derived. A number of examples are given to illustrate the usefulness of the methods.