Ph.D
Role details
Job location
Tech stack
Job description
Motivation Future space missions increasingly rely on autonomous decision-making due to communication delays, uncertain environments, and mission complexity. Applications such as planetary landing, autonomous docking, formation flying, and on-orbit servicing demand control policies that are adaptive, data-efficient, and robust to uncertainties. Reinforcement learning has shown significant promise in these domains by enabling agents to learn complex con-trol strategies directly from interaction with the environment. However, standard reinforcement learning methods lack formal safety guarantees, making them unsuitable for safety-critical space applications where constraint viola-tions can lead to catastrophic mission failure. To bridge this gap, safe reinforcement learning methods have emerged, combining learning-based control with explicit safety mechanisms. Among these, shielding approaches, which enforce safety constraints by filtering or correcting control actions, offer a promising path toward safe-by-construction autonomy.
This PhD project aims to develop theoretically grounded and practically viable reinforcement learning frameworks for spacecraft systems, where safety is guaranteed at all times through shielding, while still allowing learning-based performance optimization.
Your Tasks ï Develop reinforcement learning solutions for spacecraft control tasks, like autonomous docking, planetary landing, and active debris removal (requires experience with reinforcement learning) ï Implement high-fidelity simulation environments that can be used to train the reinforcement learning agents (requires solid knowledge about the underlying spacecraft dynamics) ï Integrate shielding methods into the training and application process to provide formal safety guarantees, even in the presence of uncertainties (requires strong mathematical background) ï Contribute to the design and development of spacecraft test facilities for the experimental validation of rein-forcement learning approaches (requires basic knowledge about hardware, sensors, actuators, etc.) ï Design and execute experimental campaigns in a laboratory environment to evaluate the performance of the reinforcement learning agents
Your Responsibilities ï Publish research findings in high-impact international journals, present at leading conferences, and support in the preparation of research proposals ï Mentor undergraduate and master students ï Develop expertise and stay up-to-date with the latest advancements in this research area ï Support in establishing a new professorship ï Contribute to teaching and examinations for TUM students at the Professorship of Spacecraft Control ï Participate in administrative tasks at the Professorship of Spacecraft Control
Requirements
ï Above-average master-s degree in Aerospace Engineering, Mechanical Engineering, Robotics, Physics, Mathematics, or a closely related field ï Strong foundations in spacecraft dynamics and control theory as well as experience with reinforcement learn-ing ï Good programming skills in Python and/or MATLAB (C++ is a plus) ï Strong interest in independently investigating scientific research questions ï Excellent team-work capability ï Fluency in spoken and written English
