Evaluating Large Language Models for Space Operations

Abstract

The increasing number of orbital launches has led to a significant surge in spacecrafts in Earth’s orbit, making mission control centers more complex and challenging to manage. To address this issue, we investigated the potential of Large Language Models (LLMs) in space operations, focusing on tasks such as information retrieval, documentation, and quick decision-making. This paper explores two evaluation studies that tested the application of LLMs for specific use cases in a non-cloud environment due to sensitive data confidentiality requirements, including a scenario for the Columbus Flight Control Team to quickly retrieve information during operations. Our results demonstrate the promising capabilities of LLMs in supporting spacecraft engineers with answers, while also highlighting the need for parameter fine-tuning, prompt engineering, or even model re-training. This paper provides actionable insights into the potential integration of LLMs in space operations and outlines future research directions related to this emerging field.