Abstract
Machine learning (ML) stands as a pivotal technology shaping the future across diverse industries, offering unprecedented opportunities for automation, optimization, and decision-making. From aiding medical diagnosis and treatment to revolutionizing transportation with autonomous vehicles, ML has positively disrupted numerous sectors, significantly enhancing quality of life. The adoption of ML for on-orbit servicing stands to provide similar benefits, acting as an enabler for a variety of mission profiles due to improved levels of autonomy, precision, and reliability. This paper provides a concise survey of current ML technologies crucial for advancing on-orbit servicing, focusing on solutions that provide robust autonomy with respect to RPO, GNC, space robotics, and ground segment operations. The analysis addresses gaps found in real space data, explores solutions tailored for hardware with limited computational resources, and outlines methods to utilize data rich environments found on ground. Development and quality assurance strategies to deploy robust ML systems in such environments are discussed, accompanied by a roadmap for continuous innovation. The proposed roadmap outlines key building steps towards the next generation of more intelligent in-orbit services, exploring possible synergies among technologies, to ultimately contribute to the broader goals of space sustainability.
