A Systematic Review of Machine Learning for Situation Awareness

Abstract This systematic review analyzes the application of machine learning (ML) techniques for Situation Awareness (SA), structured according to Endsley’s three-level model: perception, comprehension, and projection. Following a PRISMA-based methodology, we examine 93 primary studies published between 2010 and 2025 across multiple domains, including healthcare, cybersecurity, autonomous systems, and critical infrastructure. Our analysis reveals a clear imbalance in the maturity and adoption of ML techniques across SA levels, with perception being the most extensively studied and projection remaining comparatively underexplored. In particular, while deep learning approaches dominate perception tasks and hybrid models improve contextual reasoning at the comprehension level, projection-level capabilities are still limited by challenges related to temporal modeling, data availability, uncertainty quantification, and cross-domain generalization. Furthermore, the review highlights significant heterogeneity in datasets, evaluation protocols, and experimental settings, which limits reproducibility and hinders systematic comparison across studies. Based on these findings, we identify key research gaps and outline future directions toward the development of integrated, multi-level SA systems that combine data-driven and knowledge-driven approaches, enabling more robust, interpretable, and proactive decision-making in dynamic environments.