Onboard System Identification for Improved Flight Control of UAS

Abstract

A new technique is proposed to select and estimate the significant aerodynamic parameters of micro unmanned aerial systems from ight data to improve the dynamical qualities of an indirect adaptive ight control system. The aerodynamic variables are estimated in the frequency domain using the angle of attack and sideslip air-ow angles which in turn are estimated using an extended Kalman filter. Parameter estimation and selection procedures of significant aerodynamic parameters are based on linear regression model structures with forward orthogonal least square (OLS) and error reduction ratio (ERR) methods. When combined, the methods can be applied to create an indirect adaptive ight control system. This new approach is verified by comparing the results with those obtained from conventional sensor data, including air ow angle measurements. Performance comparison of the system identification methods show that the proposed technique can obtain the same quality of ight performance as if the airow measurements were available. The new methods are demonstrated in simulation of a benchmark ight performance experiment on an Aerosonde UAV.