Unscented Kalman Filter for Thermal Parameter Identification
Matthew W. Hazard · 48th AIAA Aerospace Sciences Meeting Including the New Horizons Forum and Aerospace Exposition · 2010
Autonomous thermal soaring promises to improve the loitering time and range of unmanned aerial vehicles. By making efficient use of the net vertical movement of air in a thermal, an intelligent aircraft controller can increase the total energy of the system without expending significant propulsive effort. Soaring techniques rely on accurate estimates of the properties of the thermal. In a novel application of the Julier-Uhlmann Unscented Kalman Filter, total energy measurements (from a variometer or autopilot) are fused with GPS position and velocity measurements to estimate the position, strength, and size of thermals using a four-state thermal model. The proposed method was exercised using a nonlinear thermal model, showing excellent convergence, steady state error, and phase lag performance for several typical flight paths. The derivativeless nature of the Unscented Kalman Filter simplifies refinement of the thermal model, and the reduction in computational complexity may enable its use in low cost or lightweight embedded systems.