AI-Enabled ATC Decision Support Systems: A Systematic Review Of Trajectory Prediction, Conflict Detection, And Human-AI Teaming In The Loop

Abstract

This systematic literature review (SLR) synthesises recent evidence on artificial intelligence (AI)-enabled Air Traffic Control (ATC) decision-support systems and examines how AI is used to improve operational performance, safety, and resilience in increasingly complex airspace. Research on machine learning, optimisation, and automation has expanded rapidly, but evidence remains fragmented across application areas and validation settings. A Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020-guided protocol was used to search Scopus and IEEE Xplore for English-language journal articles published from 2021 to 2025. Following screening, full-text eligibility assessment, and final reconciliation of study type, scope, and duplicate records, 30 eligible primary studies (n = 30) were included in the qualitative synthesis. The abstract level reporting completeness appraisal yielded a mean score of 80.28% and a median of 83.33%; 16 of 30 studies (53.33%) scored at least 80%, including six studies (20.00%) that scored at least 90%. These percentages describe reporting completeness rather than algorithmic performance. Because the included studies evaluated different operational tasks, datasets, outcome measures, and validation settings, their performance results were not directly commensurable; combining them into a single pooled model-performance percentage would therefore risk conflating non-equivalent metrics. Performance evidence was instead synthesised comparatively by application area, validation setting, and operational trade-off, so that study level performance was interpreted within the metric and dataset used by each primary study rather than averaged across non-equivalent tasks. Three themes were identified: (1) AI-enabled ATC/Air Traffic Management (ATM) operations, including delay and capacity management, conflict resolution, trajectory prediction, and traffic-pattern analytics; (2) aviation safety, surveillance, and resilience, including unmanned aerial vehicle (UAV)-related risks, Global Navigation Satellite System (GNSS) interference, sensing, monitoring, and safety assurance; and (3) aviation-connected smart mobility and intelligent transportation systems (ITS), including UAV logistics and related surface-transport applications. Comparative synthesis indicated that prediction methods were increasingly connected to operational decision support, whereas conflict-resolution and resilience approaches continued to rely substantially on simulation, experimental datasets, or bounded validation environments. Overall, the evidence showed increasing operational relevance, but important gaps remained in explainability, real-world validation, human-AI interaction, robustness, and system integration before wider safety-critical ATC deployment.