Assessment of Reinforcement Learning for CubeSat concept generation
B. Krijnen (TU Delft - Aerospace Engineering)
J Guo – Mentor (TU Delft - Space Systems Egineering)
More Info
expand_more
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.
Abstract
The growing need for CubeSats could present strong demands for the use of automated systems during the early stage of the design cycle. Automated design tools that are able to incorporate the entire design space offered by the commercial-off-the-shelf (COTS) components for CubeSats may potentially improve the design of a CubeSat, compared to manual methods. This thesis sets out to develop and assess such a design tool. The design tool that is developed makes use of reinforcement learning (RL) for automated CubeSat concept generation. Concepts are generated by selecting components from a manually created hypothetical components database. The capability of the design tool to create feasible CubeSat concepts is tested through a case study, where the results from a manual approach are compared to the results from the design tool. It is investigated whether the RL-based design tool shows promise for automated CubeSat concept generation.