2026-06-02T15:22:10Zhttps://riuma.uma.es/rest/oai/request

oai:riuma.uma.es:10630/403452026-02-03T12:00:19Zcom_10630_2254col_10630_37959

Autonomous Multi-Agent AI Systems for Satellite Mission Design. Navarro, Tomás Stroescu, Ana Izzo, Dario Gálvez-Rojas, Sergio López-Valverde, Francisco Inteligencia artificial Sistemas de comunicaciones inalámbricos Aprendizaje automático (Inteligencia artificial) Proceso en lenguaje natural (Informática) Procesado de señales Satélites artificiales Exploración espacial The integration of Artificial Intelligence (AI) agents in supporting engineering design is rapidly gaining attention due to their potential to accelerate decision-making, optimise designs, and reduce costs. This paper presents a comprehensive evaluation of two different AI agentic systems, each system run by a different LLM (Large Language Model): DeepSeek-R1-70B and GPT-4o. The agents are evaluated in supporting satellite constellation design across key domains: market analysis, frequency filing, mission planning, payload feasibility, and cost analysis. Four distinct satellite designs were analysed per model, and expert evaluations were conducted to assess their effectiveness. This study highlights both the benefits and shortcomings of AI agents in satellite design, providing a comparative assessment and discussing implications for future AI-driven space mission planning. 2025-10-21T07:16:00Z 2025-10-21T07:16:00Z 2025-10-21T07:16:00Z 2025-10 conference output https://hdl.handle.net/10630/40345 10.1109/ICMLT65785.2025.11192855 eng 10th International Conference on Machine Learning Technologies (ICMLT) Helsinki, Finlandia 23-25 de mayo de 2025 open access IEEE