Sequential and Partial Updating of Parameters During Estimation

Two least-squares-type UD factorization algorithms are proposed which update only part of the covariance matrix and parameter vector during each identification interval. This reduces the computational load per interval but still provides good parameter estimates for slowly varying systems. The algorithms are particularly well suited for applications where a large number of parameters must be estimated, e.g., for MIMO models and/or direct identification of step-response parameters.