Supplementing state and dynamics estimation with information from optimal control policies

In this paper, we address dynamically mismodeled systems, and how those mismodeled dynamics may be detected and recovered. We combine an optimal control process with an optimal estimator to yield an algorithm that outputs simultaneous state and control estimates. This new estimator (the Optimal Control Based Estimator) is shown to be a generalization of the Kalman filter with automatic dynamic noise propagation and smoothing properties. The control estimates it outputs contain information about the mismodeled dynamics. Complimentary algorithms that detect and reconstruct maneuvers from the information in the control policies are developed. A simulation in which these algorithms are applied to an astrodynamics tracking problem shows how this algorithm may be used to estimate a system's state, detect mismodeled dynamics, and reconstruct those dynamics.