SafeOPS will focus its research on “from prediction to decision”, a common decision-making paradigm in digitalization and predictive analytics. It builds upon and complements the technical work from the recent project SafeClouds.eu which demonstrated the power of big data analytics as a means of providing information-driven analysis for hazard identification in aviation. SafeOPS will investigate how predictive AI tools can be used in ATM as decision support technologies, to facilitate controllers making complex decisions. The project aims to pave the way towards the implementation of AI-based solutions by assessing their contribution to the security and resilience of the ATM system. The work performed in SafeOPS is split into three pillars which we call the Operational Layer, Predictive Layer and the Risk Framework.
Developing a decision-support tool powered by AI
to help air traffic controllers make complex decisions
in the context of go-arounds.
AI-driven decision support tools in ATM
AI/ML predictive analytics and Human interpretability
Decision intelligence and Safety performance
The Operational Layer (OL) directly addresses the ATM system. It aims to provide a thorough analysis of the ATM system as it is today, focusing on the technologies currently available and how a controller utilizes them. Based on these insights, ensuring user (ANSPs and Airlines) alignment, user stories and subsequently the requirements on how AI based decision support can bring benefit to the ATM operations, will be developed. Thereby, the OL will also investigate the operational usage of the probabilistic information from the Predictive Layer and evaluate through the Risk Framework its impact on the ATM operations.
The Predictive Layer (PL) addresses all big data and AI related tasks. SafeOPS will leverage on DataBeacon – a platform for research, development, and deployment of AI applications in the aviation safety environment. It will focus on two main objectives. The first one covering all the related actions for the creation of the necessary automated data pre-processing and preparation pipelines. The second focuses on the AI/ML solutions for the predictive analytics that will be chosen and trained with a special focus on the human interpretability aspect of the solution. The trained AI/ML solutions will be deployed, delivering the predictive analytics to the Risk Framework (RF).
The Risk Framework (RF) of SafeOPS addresses the challenge of integrating the information generated by the Predictive Layer in the Operational Layer. The goal of the RF is twofold: first, evaluating the risks associated to the go-around scenario, for example the risk of losing separation. Second, determining how Human Factors affect the safety performance in go-arounds, that is, how Air Traffic Controllers elaborate the available information to make decisions and take consequent actions.
During the Associated Partners Workshops, we take the chance to present the work done in SafeOPS as well as discuss and receive feedback on our work with interested stakeholders and discuss on ML based tools in ATM in general.
If you are interested in participating, please write a mail to Lukas Beller (firstname.lastname@example.org) to receive the invitation for the coming workshops.
D1.2 Final Project Results Report
D3.2 Integrated risk framework
D2.2 Impact evaluation of the developed decision support tools for ATM
D2.3 Guidelines on decision intelligence for automation
D4.2 Human interpretability framework for the selected user stories
D5.2 Dissemination and communication actions