Satellite intelligence improves AI training and use in low Earth orbit
Satellite Intelligence: Distributed Training, Collaborative Inference,and LLM Adaptation in LEO Constellation
Networking and Internet ArchitectureArtificial Intelligence
Summary
Satellites in low Earth orbit face big challenges running smart AI programs because they have limited power, memory, and connectivity. This paper surveys ways to train AI models across satellites and ground stations, perform real-time AI tasks collaboratively among satellites, and adapt large language models on orbit despite these constraints. The authors lay out the principles and approaches that help make AI work better in space by using distributed computing and resource-aware methods. They also highlight future challenges to advance intelligent satellite networks.
What this means in practice
- •For satellite systems engineers: Design distributed training systems that enable AI model updates shared between satellites and ground stations under strict resource and link limitations.
- •For edge ai developers: Build infrastructure for real-time AI inference that collaborates across multiple satellites and ground stations to improve responsiveness in space applications.
A survey. It maps existing work.
Authors
Zheng Lin, Zhanwei Wang, Yuxin Zhang, Yafei Wang, Ons Aouedi, Zihan Fang, Jinfeng Xu, Jing Yang, Vu Nguyen Ha, Yuanwei Liu, Mohamed-Slim Alouini, Symeon Chatzinotas
Abstract
The proliferation of orbital sensing data and the real-time demands of emerging global services are pushing low Earth orbit (LEO) satellite networks from bent-pipe relays toward intelligent onboard processing. This shift confronts a fundamental tension, where modern artificial intelligence (AI) training paradigms require intensive computation, abundant memory, and stable connectivity, all of which are scarce under the dynamic topologies, intermittent links, and severe size, weight, and power (SWaP) constraints of satellite networks. Addressing this tension demands rethinking AI workflows, from training and inference to the emerging deployment of large language models (LLMs), under a constraint-aware design philosophy. To the best of our knowledge, this paper provides the first survey that jointly covers all three, whereas prior surveys examine these aspects in isolation. We begin by identifying the deployment challenge of satellite intelligence (SI) and deriving design principles for SI. We then examine three advancing paradigms: i) distributed training, where federated learning (FL) and split learning (SL) enable satellite-ground co-training under resource limitations; ii) collaborative inference, which exploits single-satellite, multi-satellite, and satellite-ground architectures to meet real-time service demands; and iii) on-orbit LLM adaptation and serving, which tackles the frontier challenge of deploying LLMs under severe orbital constraints and resource-aware serving. Finally, we identify open challenges and future directions. By bridging the AI and satellite networks, this paper provides a research roadmap for the implementation of space-native AI technologies.