Autonomous Soaring

Methods of Reinforcement Learning

Autonomous Soaring System


From the beginnings of aviation until today, new technologies - primarily in the fields of aerodynamics and lightweight construction - have been investigated, developed, and tested within the sport of gliding. Due to an increasing number of applications using unmanned fixed-wing aircraft with long mission durations, the automation of tasks typical for soaring flight has become a focus of research in flight control. Besides, its tactical nature and easy accessibility make for gliding being an ideal "playground" for research into modern control engineering methods, especially those from the field of artificial intelligence.



Glider pilots use updrafts to achieve long flight durations and cover large distances in cross-country soaring. Flying without propulsion requires a constant trade-off between completely contradictory directions of action: On the one hand, the primary goal of traveling from A to B, for example, should be pursued. On the other hand, updrafts have to be found and exploited. At iFR, we develop methods to automatically localize and characterize thermals. Tactical decision-making in motion planning, characterized by a stochastic environment as well as long-term correlations between actions taken and goals achieved, are addressed using machine learning methods. A cooperative approach of multiple gliders makes the task even more challenging. Flight tests demonstrate the performance and feasibility of applying these methods in the context of flight control.


  • Contributing to eco-efficient aviation by developing algorithms for automatic localization and exploitation of thermal updrafts
  • Investigating artificial intelligence within the scope of aerospace guidance, navigation, and control and demonstrating its feasibility through real flight tests


  • Schimpf, F., Notter, S., Groß, P., & Fichter, W. (2021). Multi-Agent Reinforcement Learning for Thermalling in Updrafts. Proceedings of the AIAA Scitech 2021 Forum
  • Rothaupt, B., Notter, S., & Fichter, W. (2021). Autonomous Soaring Policy Initialization Through Value Iteration. Proceedings of the AIAA Scitech 2021 Forum.
  • Notter, S., Schimpf, F., & Fichter, W. (2021). Hierarchical Reinforcement Learning Approach Towards Autonomous Cross-Country Soaring. Proceedings of the AIAA Scitech 2021 Forum.
  • Notter, S., Groß, P., Schrapel, P., & Fichter, W. (2020). Multiple Thermal Updraft Estimation and Observability Analysis. Journal of Guidance, Control, and Dynamics, 43(3), 490–503.
  • Notter, S., Zürn, M., Groß, P., & Fichter, W. (2019). Reinforced Learning to Cross-Country Soar in the Vertical Plane of Motion. Proceedings of the AIAA Scitech 2019 Forum.
  • Groß, P., Notter, S., & Fichter, W. (2019). Estimating Total Energy Compensated Climb Rates from Position Trajectories. Proceedings of the AIAA Scitech 2019 Forum.
  • Notter, S., Schrapel, P., Groß, P., & Fichter, W. (2018). Estimation of Multiple Thermal Updrafts Using a Particle Filter Approach. Proceedings of the 2018 AIAA Guidance, Navigation, and Control Conference.
Flugdatenauswertung zum automatischen Lokalisieren und Charaktisieren mehrerer thermischer Aufwinde mittels eines Partikelfilters
This image shows Stefan Notter

Stefan Notter

To the top of the page