Online Near Real-Time Online System Identification on Small Unmanned Aircraft Systems

Abstract

An online near real-time system identification system is developed for generating locally linear models of Small Unmanned Air Systems. Automated control surface excitation inputs consisting of doublets, triplets, and frequency sweeps are implemented and used to assure consistency in the excitation and to eliminate errors introduced by user applied inputs. To provide reliable data for processing, a high frequency data acquisition unit is developed and implemented. In addition, a real-time vehicle monitoring system is used to provide a human-in-the-loop model update capability, with a goal of ensuring safety of the vehicle. Flight tests and modeling are demonstrated on a fixed-wing Small Unmanned Air System, with locally linear models generated during flight.

Observer Kalman filter identification is used as the primary identification algorithm with adjustments made for real-time identification purposes. Identified models are both stored and sent to the ground control station for ground control operator for update verification. Results presented in the thesis show that the system provides a capability for generating accurate locally linear models that are suitable for real-time flight control design using model based control techniques and post-flight modal analysis.