Learning-Based Motion Planning with Multiple Thermal Updraft Mapping for Autonomous Cross-Country Soaring

Solving the decision-making problem between pursuing the objective of covering distance and exploiting thermal updrafts is the central challenge in cross-country soaring flight. The need for trading short-term rewarding actions against actions that pay off in the long term makes for a hard-to-solve problem. Policies resulting from reinforcement learning offer the potential to handle long-term correlations between actions taken and rewards received. The thesis presents a learning-based motion planning scheme, which results in a control strategy for autonomous cross-country soaring. Real-world autonomous soaring flight requires an algorithm to infer regions of atmospheric uplift from in-flight measurements. The thesis presents a novel approach for the integrated mapping of multiple thermal updrafts. Flight test results showcase the feasibility and performance of the thermal updraft estimator proposed. Moreover, the flight test results validate the successful transfer of the learning-based control policy trained in simulation to real-world autonomous cross-country soaring. The thesis presents an autonomous soaring system capable of localizing thermal updrafts from in-flight measurements, harvesting energy by exploiting atmospheric uplift, and solving the long-term correlated decision-making problem central to cross-country soaring flight.