Sequential parameter estimation using pseudoinverse

Abstract

This correspondence presents a method for solution of the parameter estimation problem in an entirely sequential fashion. In this procedure each measurement is processed as it is received, without regard to whether it and the prior measurements have established a minimal observable data set, or without the assumption a priori of a covariance matrix in the manner of the Kalman filter. During the period prior to the establishment of the minimal data set the estimate maintains the character of the pseudoinverse constraint, that of minimum norm. In addition at each measurement, a parameter is available for deciding whether this measurement has increased the observability of the system.