
Image of flight Path of a test flight with a Cessna 172.
This thesis focuses on the theoretical use cases and technical feasibility of reinforcement learning in solving complex problems in the aerospace sector. Its primary goal is to establish a theoretical foundation for further research in this field. The research aims to prove whether current-generation algorithms, with hyperparameter tuning and other performance enhancements, can be used in a new generation of autopilot systems for self-landing rockets.

Reward Function Visualization of a test flight with a Cessna 172 training to complete a waypoint mission.
This work aims to analyse existing research in the field of reinforcement learning. It will then build a platform for experimenting with reinforcement learning agents within the X-Plane 11 flight simulator. By providing sample use cases and environments for testing, the platform will facilitate the exploration of advanced manoeuvres, such as vertical landing. This focus on advanced landing algorithms serves to demonstrate the feasibility of reinforcement learning in such tasks.