Improving Autonomous Soaring via Energy State Estimation and Extremum Seeking Control
Open Access
- Author:
- Daugherty, Shawn Christopher
- Graduate Program:
- Aerospace Engineering
- Degree:
- Master of Science
- Document Type:
- Master Thesis
- Date of Defense:
- None
- Committee Members:
- Jacob Willem Langelaan, Thesis Advisor/Co-Advisor
- Keywords:
- autonomous soaring
uav
sb-xc
autonomous
thermal
thermal soaring - Abstract:
- This research is motivated by the significant potential of soaring UAVs to efficiently accomplish both civil and scientific missions by atmospheric energy harvesting. Viable missions include surveillance, communication relay, and environmental monitoring. This thesis seeks to improve the utility of small, autonomously controlled gliders by extending the range and endurance of these vehicles. This is accomplished through the exploitation of energy from columns of warm, rising air known as thermals. Thermals occur naturally and are utilized by large birds and sailplane pilots to soar for several hours and cover distances of hundreds of kilometers without any source of propulsion. This thesis analyzes limitations imposed by current algorithms and describes solutions in the form of improved energy estimation methods and turn optimization through extremum seeking control. The thermal centering algorithm, based on Reichmann's method, uses the second derivative of total energy as a feedback term to remain centered around the thermal core. Due to the controller's susceptibility to latency, conventional filtering methods greatly restrict the capabilities of the centering controller. This thesis discusses an alternative estimation method, an asymmetric Savitzky-Golay filter that computes estimates of total energy, rate of change of total energy and the second derivative of total energy using polynomial approximations over a moving time window. Significant improvements were observed including: the ability to track a larger range of thermals, rapid thermal centering, and improved disturbance rejection. The problem of optimal thermal soaring was also addressed. Assuming a Gaussian updraft distribution, any given thermal has an optimal flight radius that can be computed for the specific aircraft. However, determining this flight path has proven to be a difficult problem that has not been adequately addressed. A solution is proposed; climb rate maximization using extremum seeking control with turn radius as the varying parameter. Simulations demonstrated steady turn rate convergence while driving the climb rate to the optimal value. A simulation environment based on a commercially available soaring simulator is described, with a low level aircraft controller implemented on an Arduino Mega 2560 single board computer. This environment was used for testing and validation of the aforementioned methods. The utility of the energy estimator and extremum seeking controller is demonstrated in this high fidelity simulation.