Data Strategies for Training AI Solutions in support of Space C2

Rapid advancement in the field of AI provides opportunities to apply AI capabilities to the space C2 domain to gain and maintain space superiority. Many operational tasks that have traditionally been time- and labor-intensive can benefit from increased automation through AI. The space domain awareness data used to train AI models comes in many forms and fidelities, making it difficult to train reliable models. By utilizing common space C2 use cases, we will demonstrate an approach to effectively train AI models with varying fidelity of both simulated and real-world data. We will evaluate sufficiency of training data along the dimensions of data quality, density, geometric diversity, precision, accuracy, and timeliness relative to our selected use cases and to the anticipated quality of operational data. We will present an iterative, extensible, and repeatable approach to training AI models. Our Phase I effort will inform the design of a consistent digital training environment for training and maintaining models.