System identification, design and flight testing H-inf controllers for heave dynamics of RUAV

Abstract

Focus of this thesis was the improvement of the vertical flight performance of the small-scaled, unmanned, autonomous DLR helicopters HE-1 and HE-2 by employing an advanced model-based controller. The success of the final design was to be analyzed in flight experiments and benchmarked against the original flight controller. We cover the development of a basic three rigid body model for the helicopter HE-1. The heave (vertical) dynamics of the mechanical model was extended by the aerodynamic effect called heave dampening. Unknown parameters of the heave dynamics model were determined with the Prediction Error Minimization (PEM) system identification method. Necessary flight data was recorded in specially designed test flights. The decision for the controller concept was made in favor of H¥ control due to its desirable characteristics, such as MIMO capability and its robustness performance for systems subject to (parameter) uncertainties and external disturbances. Based on the final heave dynamics model, various H¥ controller designs, such as (i) S/KS/T mixed sensitivity optimization, (ii) signal-based H¥ control, (iii) 1 degree-of-freedom loop shaping design, (iv) 2 degree-offreedom loop shaping design were realized and compared in simulation. The most promising design (iv) was then implemented on the HE-2 on-board flight controller and benchmarked in flight experiments versus the original cascaded PID controller. Based on the analyzed data, suggestions on how to further improve the controller are given.