-
Natural sciences
- Game theory, economics, social and behavioural sciences
- Machine learning and decision making
- Neural, evolutionary and fuzzy computation
- Database systems and architectures
- Decision support and group support systems
- High performance computing
-
Social sciences
- Criminology not elsewhere classified
-
Engineering and technology
- Other computer engineering, information technology and mathematical engineering not elsewhere classified
Ports are vital gateways for global trade, handling hundreds of millions of tons of cargo each year. They are also key entry points for smuggled goods, making effective inspection crucial but increasingly complex. Current inspection practices face a structural dilemma: scanning every incoming ship, truck or container is infeasible, while selective checks based on static risk models leave critical blind spots. This creates a persistent tension between efficiency, cost control, and security. Existing approaches rely on static risk models, rigid sensing configurations, and no anticipation of adaptive adversaries. These are not just engineering shortcomings; they reflect deeper scientific questions about how heterogeneous information, cooperative sensing, and human behavior interact under uncertainty, i.e. when the available information is incomplete, conflicting, or strategically manipulated.
IM‑PORT‑ANT addresses this gap for Flemish ports by developing a privacy‑compliant, interpretable, and adaptive framework for anomaly detection and risk‑based inspection. The project integrates:
- hybrid, explainable risk prediction that fuses structured and unstructured logistics data, surveillance sensor analytics, and expert knowledge;
- cooperative, low‑power sensor fusion that sustains detection and tracking when some sensors fail or conditions change;
- adversarial modeling that anticipates offender adaptation through agent‑based and game‑theoretic methods.
IM‑PORT‑ANT goes beyond the state of the art by combining neuro‑symbolic AI, simulations, and adaptive sensor fusion into a coherent system that learns from both expert reasoning and data‑driven insights. This enables risk models that are interpretable, resilient to uncertainty, and robust against strategic deception. Digital twins accelerate innovation by simulating rare events and optimizing sensor placement without costly real‑world trials. To further enhance resilience, game-theoretic modeling is used to anticipate adaptive adversarial behavior, allowing inspection strategies to evolve in response to strategic manipulation. By prioritizing high‑risk shipments and dynamically reallocating sensing resources, the project maximizes detection efficiency per euro spent, reducing unnecessary inspections and operational disruption.
Expected outcomes after two years include:
- A validated proof‑of‑concept for explainable risk analysis and decision support
- A modular, cooperative sensor‑fusion prototype with energy‑aware deployment strategies
- Adversarial behavior models and resilience mechanisms tested in simulated, realistic scenarios
- Legal and ethical guidelines for AI‑driven inspection systems
Impact: For Flemish ports, IM‑PORT‑ANT strengthens security while minimizing disruption and cost, supporting sustainable and competitive logistics. For industry, it creates opportunities in AI‑driven risk analytics, sensor fusion, and privacy‑compliant data governance. For society, it enhances the resilience of critical infrastructure, fosters trust in AI, and aligns with EU priorities on secure and ethical digital infrastructure. By embedding legal foresight and stakeholder co‑creation, the project ensures that innovations are not only technically robust but also socially acceptable and ready for future valorization.