Adaptive State Estimation with Subspace-Constrained State Correction

Abstract

In many applications of state estimation, it is efficient to confine the output-error injection to a prescribed subspace of the state space. This paper considers this problem by applying the unscented Kalman filter and retrospective cost state estimator (RCSE) to linear and nonlinear systems with subspace-constrained state correction. As an application of these techniques, parameter estimation is considered for linear and nonlinear systems with unknown parameters, where the output- error injection is confined to the subspace corresponding to the states representing the unknown parameters.