Model Based Navigation of Delta-Wing UAV - In-Flight Calibration and Autonomous Performance

This paper presents the first practical results of a model-based approach to autonomous navigation applied to a small delta wing drone. The aerodynamic coefficients of the considered platform are unknown and need to be determined for the (vehicle dynamic) model-based navigation to work properly. The proposed approach uses post-processed INS/GNSS trajectory estimates of relatively high precision as observations to refine priors'' of aerodynamic coefficients via state-estimation (Extended Kalman Filter). Two methods to derive such priors'' (i.e., initial parameter values) are investigated. The first adapts coefficients described in the literature for an aircraft of similar geometry. The second performs regression analysis of flight data to estimate coarse values of the coefficients. Both sets of coefficients are further re-calibrated in-flight via state estimation. The accuracy of the coefficient calibration is evaluated by simulating a GNSS outage of several minutes, during which the trajectory flown under autonomous navigation is compared to that of the reference.