Aerospace Applications of Noise Covariance Identification with Autocovariance Least Squares

Abstract

State estimation is necessary to support state feedback control applications whenever measuring the full state is prevented by environment or cost. Estimator performance depends on an accurate model of the stochastic process, however in practice noise covariances are not known and are hard to determine. The Autocovariance Least Squares (ALS) technique is a direct correlation method that identifies a system's noise covariance matrices. ALS poses a least squares problem using a linear or non-linear system model and experimentally observed measurement innovations. This dissertation addresses fielded vehicles for which ALS can be used to inform the design of state estimation filters. There are unique noise identification challenges with aerospace vehicles including the conditioning of the ALS problem, measurement of atmospheric disturbances, uncertainty in the dynamic model, and the cost and time required for data collection. Extensions of the ALS technique are developed to address these challenges. The extensions are applied to design state estimation filters for four aerospace systems for both simulated and experimentally collected data. The systems considered are the University of Washington's two gust load alleviation wind tunnel test-beds; the 3 ft x 3 ft Model for Aeroelastic Response to Gust Excitation and the Kirsten Wind Tunnel Large Wing test-bed, a simulated narrow body commercial aircraft, and flight test data from a wide body commercial aircraft.