Font Size: 

Dynamic soaring refers to a flight technique used primarily by large seabirds to extract energy from the wind shear layers formed above ocean surface. A small Unmanned Aerial Vehicle (UAV) capable of efficient dynamic soaring maneuvers can enable long endurance missions in context of patrol or increased flight range. To realize autonomous energy-saving patterns by a UAV, a real-time trajectory generation for a dynamic soaring maneuver accounting for varying external conditions has to be performed. The design of the flight trajectory is formulated as an Optimal Control Problem (OCP) and solved within direct collocation based optimization. A surrogate models of the optimal traveling cycle and initial value problem capturing wind profile uncertainties are constructed using Polynomial Chaos Expansion (PCE). In each cycle, the polynomial model of the initial value problem is used to rapidly estimate wind profile parameters using a genetic algorithm. The PCE surrogate model is subsequently used to generate a new trajectory using the estimated wind profile parameters. In order to maintain altitude and velocity at the end of each cycle, a thrust control variable is augmented to the model to provide trajectory correction in the terminal phase of the maneuver. The required thrust correction is shown to be decreasing in each cycle with reduced uncertainty of the wind profile.