Impact of correlation errors on the optimum Kalman filter gain identification in a single sensor environment

Abstract

The impact of errors in the innovation correlation functions evaluation, related to the suboptimal filter, on the identification of the optimum steady state Kalman filter gains are investigated. This issue arises in all real time applications, where the correlations must be calculated from experimental data. An identification algorithm proposed in the literature, with formal proof of convergence, is revisited and summarized. Based on this algorithm, equations describing this impact are developed. Simulation results are presented and discussed. As contribution, experimental results of the identification algorithm, applied to estimate the states of a position servo systems, are presented.