Modelling and Control of a Liquid Rocket Propulsion System using Machine Learning Algorithms.

Institut

Lehrstuhl für Raumfahrtantriebe

Typ

Masterarbeit /

Inhalt

 

Beschreibung

The space sector is currently gaining more and more interest due to the commercialization of Space (OneWeb, SpaceX Starlink) and upcoming human missions such as the return to the lunar surface and the goal to land humans on mars. The remoteness of these missions and their high complexity demand high autonomy in the spacecrafts control systems.

Spacecrafts, especially the propulsion systems, are complex machines made up of several subsystems. Accurate engine control is essential for mission success, especially for landing manoeuvres. The current traditional industry standard uses a hierarchy of conventional control algorithms, mainly due to their predictable behaviour and the ability to prove stability.

Recently, advanced machine learning and control algorithms such as reinforcement learning (RL) have shown promising results for autonomous control of complex dynamical systems, with real live implementations such as autonomous driving or plasma control in fusion reactors. At the same time these approaches show challenges in areas such as the enforcement of boundary conditions and guarantees for predictable behaviour.

The goal of this thesis is to implement an engine model to train a reinforcement learning algorithm for engine control in complex flight scenarios This ties into the ASCENT research project at the chair of space mobility and propulsion where a vertical takeoff and landing rocket hopper is being developed to function as a research platform for advanced control algorithms. The implementation of the engine model and the RL controller provides a foundation with which future investigation such failure tolerant and adaptive control will be enabled.

Möglicher Beginn

sofort

Kontakt

Jan Kayser, M.Sc.
jan.kaysertum.de

Ausschreibung