A note on feeding uncertain knowledge into recursive least squares

Abstract

The author presents a method to incorporate available uncertain prior knowledge into recursive least squares (RLSs) initial conditions. The simple procedure makes it possible to incorporate vague prior information about a linear combination of regression coefficients. It translates such knowledge as a guess of static gain both into a point estimate and prior covariance. The information about an imprecisely known linear relation of unknown parameters is the key result.<>