Stability and Control Derivatives Identification for an Unmanned Aerial Vehicle with Low Cost Sensors Using an Extended Kalman Filter Algorithm

Abstract

Unmanned aerial systems (UASs) tend to be autonomous vehicles. Thus, they require control algorithms. More advanced control algorithms can be developed when high quality UAS dynamic models are available. It is common to develop dynamic models for UAS using low fidelity theoretical methods. In this thesis, a higher fidelity approach which has been used for manned aircraft over the past 40-50 years is applied to the SkyHunter UAS. That approach is system identification. In this approach, the aircraft dynamic model is developed based on flight data. This thesis focuses particularly on identifying the longitudinal stability and control derivatives of the UAS. Such derivatives are important in developing UAS dynamic models. An extended Kalman filter (EKF) algorithm was used to identify the derivatives. The algorithm is appealing since it can potentially allow online system identification. The SkyHunter analyzed in this thesis weighs about 10 lb. (4.5 kg) and its wing span is about 82 in (2 m.) Like many UASs, the SkyHunter uses relatively low-cost sensors. Therefore, the data contains high noise levels. Several flight portions from three different flights were analyzed and the results are presented. These flight portions were selected carefully based on criteria that make the flight data more suitable for system identification. The identified derivatives showed reasonable results in several instances. However, a large degree of variation was observed when comparing derivatives identified from the different flight portions. The inconsistency is caused by unsteady aerodynamics, sensor noise, inability of the EKF to capture aircraft dynamics due to the use of simplified equations of motion, along with other reasons discussed in the thesis. The unsteady aerodynamics were investigated through: (A) Calculation of reduced frequency and (B) Measuring the effect of the propeller on empennage aerodynamics. This is relevant since the propeller is directly in front of the empennage.