Optimizing for Edge-AI Based Satellite Image Processing: A Survey of Techniques

Higher autonomy in satellite operations is considered a significant game-changer for the space systems market in the coming decade, with numerous agencies and startups focusing on integrating machine learning into space technology. However, the adoption of artificial intelligence onboard satellites is limited by computational resources and power constraints. Therefore, AI researchers have developed several optimization techniques and tools to optimize AI models deployed on the edge. This survey reviews techniques and strategies for optimizing AI models for satellite edge deployment, including hardware acceleration, lightweight models, and compression techniques. It serves as a comprehensive reference and a starting point for further research into optimizing satellite EdgeAI deployment.