01
Time-resolved orbital behaviour
Analysis of manoeuvre sequences, proximity patterns, timing, recurrence and deviations from expected operational profiles rather than isolated orbital events.
NATO Science for Peace and Security
GUARDIAN AI-Based Behavioural Risk Analysis for Space Situational Awareness
A human-in-the-loop behavioural Space Situational Awareness overlay designed to identify unusual and potentially risk-relevant orbital behaviour under conditions of uncertainty.
Project duration
2026–2029
Programme
NATO Science for Peace and Security
Project reference
SPS G10241
Project overview
From tracking orbital objects to understanding behaviour
GUARDIAN transforms heterogeneous and uncertain SSA data
into interpretable, decision-relevant assessments of
orbital behaviour.
Space systems in Low Earth Orbit and Very Low Earth Orbit are essential to communications, navigation, intelligence, surveillance and crisis response. These orbital regimes are also becoming increasingly congested, dynamic and contested.
Current Space Situational Awareness systems are primarily designed for catalogue maintenance and conjunction assessment. While these functions are essential for collision avoidance, they provide limited support for understanding behaviour over time or identifying complex patterns that may indicate elevated operational risk.
GUARDIAN addresses this limitation through a scientifically rigorous and transparent behavioural SSA overlay. The system analyses sequences of motion, manoeuvre timing, proximity patterns and deviations from expected operational profiles while retaining full human control over interpretation and decision-making.
Behavioural analysis
Three foundations for trustworthy behavioural SSA
GUARDIAN combines time-resolved orbital analysis, physics-consistent simulation and interpretable artificial intelligence to identify situations requiring closer analytical attention.
02
Physics-based synthetic scenarios
Reproducible orbital scenarios reflecting ambiguity, delayed manoeuvres, repeated close approaches, observation gaps, noisy measurements and degraded observability.
03
Explainable risk indicators
Probabilistic assessments supported by confidence estimates and interpretable explanations that allow analysts to understand why a behaviour has been highlighted.
Operational concept
A behavioural overlay for existing SSA infrastructures
GUARDIAN is designed to complement existing civil and defence-related SSA capabilities rather than replace them. It adds analytical depth by converting orbital data into explainable behavioural risk assessments.
Multi-source SSA data
Public and operationally representative orbital data, including object trajectories, manoeuvre histories, proximity events and uncertain observations.
→
Behavioural analysis layer
Machine-learning models and physics-consistent simulations identify unusual temporal patterns, quantify uncertainty and generate interpretable, probabilistic risk indicators.
→
Human decision support
Analysts receive confidence-aware explanations that support closer examination and informed action without automated attribution or determination of intent.
Project objectives
Earlier, more reliable and more interpretable risk assessment
Develop machine-learning methods for analysing time-resolved orbital behaviour and complex manoeuvre patterns.
Identify risk-relevant behaviour earlier than conventional event-based SSA methods.
Reduce false-positive alerts in ambiguous and degraded observation conditions.
Produce calibrated confidence metrics and transparent explanations for each behavioural assessment.
Train and validate models without dependence on classified or protected SSA information.
Support interoperability with existing civil and defence-related SSA infrastructures.
Expected outcomes
Three principal project deliverables
SSA decision-support prototype
A validated prototype implementing machine-learning behavioural analysis, quantified uncertainty and explainable outputs suitable for integration with existing SSA systems.
Synthetic scenario framework
A physics-based environment for training, testing and stress-validating advanced SSA analytics under realistic ambiguity and degraded observability.
Quantitative validation results
Evidence of improvements in early risk detection, false-alarm reduction and decision relevance compared with baseline SSA analysis methods.
Performance assessment
Key indicators of project success
GUARDIAN will be evaluated through quantitative performance metrics and expert-user assessment in representative, scenario-based validation exercises.
Reduction in time-to-detection of risk-relevant behaviours compared with baseline SSA approaches.
Measurable reduction in false-positive alerts under ambiguous and degraded observation conditions.
Calibrated confidence metrics corresponding reliably with observed scenario outcomes.
Positive expert evaluation of interpretability, usability and relevance to realistic decision contexts.
Consortium
Participating organisations
The consortium integrates expertise in space science, artificial intelligence, validation, systems engineering and project coordination.
Türkiye
Istanbul University
NATO Country Project Director:
Prof. Dr. Özgün Erler Bayır
Ireland
Dublin City University
Partner Country Project Director:
Dr. Oisín Creaner
Belgium
ULB-CREST
Co-Director:
Dr. Claire Perfetti
France
CEA LIST / SSIA
Co-Director:
Jérôme Gauthier
