Knowledge extraction, Machine Learning and other AI approaches for secure, robust, frugal, resilient and explainable solutions in Defence Applications
KOIOS (the acronym means “Knowledge” in Ancient Greek) is a research initiative born and designed from an open and transparent partnership from academia (TUE, UNIBW, UNIBG); Research centers (BSC, FOI, ONERA, LNE, MARIN and FORTH); industry (CT, NTTDS) and SMEs (MITIGA, AIA, VOCAPIA) whose main interests are to promote European research collaborations in Defence.
KOIOS is an interdisciplinary project that will develop Frugal Learning methods, i.e., AI techniques to be applied in specific military use cases where:
- There is small or scarce data.
- There is a need for optimizing the computational or energy resources (like embedded IA).
- There is a need for rapid adaption to new scenarios not trained in the method.
The project represents an ambitious investigation initiative that will produce new AI techniques beyond the current state of the Art in different research domains (few shot /transfer learning, zero shot leaning, synthetic data for AI, semi or non-supervised learning, domain adaptation…).
KOIOS will seek to improve AI for military applications, spanning simulation, use-
cases, metrics, and real world experiments.
The EU has accelerated AI adoption, motivated by the objective of achieving European strategic autonomy. Strategic autonomy and technological sovereignty also reflect Europe’s aspiration to become a global leader in critical technologies.
The potential disruptions of KOIOS will be materialized in the following outputs:
1. Holistic Human centered approach: Human involvement in decision-making reduces risk of undesirable outcomes and is a critical enabler of system resilience.
2. Combination of High-Performance Computing capabilities with AI models to improve frugality and robustness.
3. Contribution to standardization of benchmarking and evaluation
methodologies for AI systems in Defence.
4. Contribution to AI capabilities enhancement: KOIOS will produce training materials to support non specialist end-users in rapid adaptation of AI methods.