The focus of the project is design of state estimators for physical systems based on partial state measurements. We assume that the measurements are corrupted by disturbances with (partially) known autocorrelation profiles. The goal is to develop tools fo r integrating model-based observer design with a particular type of signal processing. The signal processing consists of creating mixed measurements as linear combinations of current measurements and deliberately delayed measurements, to weaken the intens ity of the disturbances based on the autocorrelation profile. The mixing changes the output map of the system, and can affect its observability properties. The choice of observer gains must therefore be integrated with the choice of mixing strategy.
Opti mization-based tools will be created to determine a mixing strategy and observer gains, to minimize or reduce the RMS and variance of the estimation error. Linear time-invariant systems will be considered first; then extensions to time-varying and nonline ar systems.
Theoretical results will be applied to GPS/INS integration for estimation of vehicle velocity and position. Focus will be on marine vehicles, where GPS measurements are subject to wave-induced disturbances that are well-suited for attenuation by mixing of past and present measurements. Experiments will be carried out on equipment at the Norwegian University of Science and Technology (NTNU).
The project addresses a fundamental problem within estimation theory, using a novel approach. The resu lts are expected to be particularly useful for measurement disturbances that cannot effectively be modeled as filtered white noise in the Kalman filter framework. Results in GPS/INS integration will benefit industries like the Norwegian marine industry. T he project will provide research experience for the postdoctoral researcher, who will primarily be stationed at Washington State University (WSU), and will lead to joint publications involving WSU and NTNU.