Artificial Intelligence for a Safe Space: Data and Model Development Trends in Orbit Prediction and Collision Avoidance

The importance of safety in space has been greatly emphasized during the recent years as a result of the continuous technological advancements, leading to the reduction of launch costs, the development of smaller satellites that use Commercial Off-The-Self (COTS) equipment, and the capability to deploy them in constellations. Space Surveillance and Tracking (SST) plays a critical role in ensuring the security of space infrastructure, and recent developments in Artificial Intelligence (AI) have shown great potential in improving SST applications such as Orbit Prediction and Collision Avoidance. This paper presents a thorough literature review of the use of Machine Learning (ML) and Deep Learning (DL) methods for SST applications in the context of safety in space. We analyze the trends in the adoption of AI technologies, identify the types of data and data sources used, and explore how classical methods are being replaced or enhanced by AI methods over time. We also discuss the types of models and architectures used in the papers that utilise ML/DL for Orbit Prediction and Collision Avoidance, as well as their data sources and the way in which they are commonly being processed before they are provided as inputs to the AI architectures. Our findings demonstrate the significant potential of AI in enhancing the safety of space infrastructure, as well as an evident trend in the adoption of relevant methods. Overall, this paper provides a valuable resource for researchers interested in utilizing AI for SST applications, especially those of Orbit Prediction and Collision Avoidance, and can act as an accelerator in terms of identifying data sources and AI architectures to adopt and extend.