Robust Hybrid Kalman Filter for a Class of Nonlinear Systems

Motivated by real-world applications with intermittent sensor data, an extended Kalman filter is formulated as a hybrid system and constructive conditions on its parameters guaranteeing an asymptotic stability property are provided. The dynamical properties of the estimation error are first characterized infinitesimally so to yield bounds on the rate of convergence and overshoot that depend on the parameters. By recasting the problem as the stabilization of a compact set, robustness properties of the proposed algorithm in the presence of disturbances in the system dynamics as well as measurement noise in the output are established. The proposed strategy is applied to spacecraft relative motion control with position-only measurements.