Skip to main content
Weiter zur Homepage der Europäischen Kommission (öffnet in neuem Fenster)
Deutsch de
CORDIS - Forschungsergebnisse der EU
CORDIS

Methodological Framework for Developing Unerring Safety-critical Ai-Based Systems

Ziel

"Automation is one of the key technical levers to enable the vision of a Digital European Sky. To enable automation of complex tasks, algorithms based on Artificial Intelligence (AI), in particular Machine Learning (ML), are necessary. However, to develop reliable AI Solutions for safety critical tasks novel development methods are needed since traditional development approaches do not consider the particularities of AI development. The goal of the MEDUSA project is to create a model-based framework for the development of safe, secure and transparent-by-construction AI Systems based on data- and model-requirements specified in the EASA Concept Papers. Baseline of the approach are Operational Domain Models of the systems derived from informal descriptions of the use cases. The models are created using a dedicated tool with the capabilities to create structured and formalized domain models as representation. Based on these models, scenario descriptions are generated for the creation of training and test data. The generation of scenarios orientates itself on the fulfilment of the data requirements, such as completeness and representativeness. Explainability and robustness techniques are applied to analyze the performance and qualities of the ML-models in development. Results from the robustness and explainability analyses help to refine the Domain model of the ML-model and increase the safety of application of the AI model. The ODD model is complemented by a model-based security approach to develop a security assurance case of the AI-based application. The MEDUSA approach is applied to two selected use cases: ""AI-Based Optimization of Departure Profiles for Noise-Aware Operational Assistance in Air Traffic Control"" and ""Perception with Airborne Radar"". The output of the MEDUSA learning assurance framework are structured safety cases that use artifacts from scenario-based testing, as well as explainability and robustness analysis as evidence."

Wissenschaftliches Gebiet (EuroSciVoc)

Schlüsselbegriffe

Programm/Programme

Thema/Themen

Finanzierungsplan

HORIZON-JU-RIA - HORIZON JU Research and Innovation Actions

Alle im Rahmen dieses Finanzierungsinstruments finanzierten Projekte anzeigen

Aufforderung zur Vorschlagseinreichung

(öffnet in neuem Fenster) HORIZON-SESAR-2025-DES-ER-03

Alle im Rahmen dieser Aufforderung zur Einreichung von Vorschlägen finanzierten Projekte anzeigen

Koordinator

DEUTSCHES ZENTRUM FUR LUFT - UND RAUMFAHRT EV
Netto-EU-Beitrag
€ 749 511,13
Adresse
LINDER HOHE
51147 KOLN
Deutschland
Region
Berlin Berlin Berlin
Aktivitätstyp
Research Organisations
Links
Gesamtkosten
Keine Daten

Beteiligte (4)