Complexity and Data Science for ATM Performance
Proposals should select data-driven methods that allow to identify so far unknown patterns, correlations or even cause-effect relationships in ATM data or allow for improved predictions on different levels (e.g. trajectories, sectors, and network). Proposals may suggest to use model-based approaches as a complement to data-driven methods, including a ‘data-twin’ representing reality.
Research activities should consider potential sources of uncertainty (e.g. partial observations, inaccurate information, incomplete knowledge inherent random nature, etc.) and their impact on the potential conclusions. Research activities may aim to provide improvements to the whole data workflow including data acquisition, cleaning, processing, and analysis.
Research activities addressing this topic should aim at providing new insights into the performance of the ATM system and investigate how the findings can be used to support strategic or real-time decision making in order to improve ATM performance. They should develop specific case studies and aim to identify actionable indicators or develop innovative visualisation techniques for complex data to support decision making.
As a part of their activity, projects may also define and make available standardised data-sets for training statistical models and a 'gold standard' (e.g. best theoretical classification of input data).
For example, proposals could offer to study how data science can be used to provide new insights into optimising airspace management (e.g. sector design and configuration, demand and capacity balancing, separation management), for example by studying delay sinks and amplifiers.
In case research activities investigate safety critical applications, potential safety issues (determinism, certifiability, etc.) raised by the selected method must be addressed.
Due to the rapidly developing techniques of data science it is possible to analyse large data sets and to detect new correlations and relationships. In fact, data science techniques are already having a huge impact in many application areas. The application of innovative methods of data science in the ATM domain could provide new insights to measure and improve the performance of the ATM system.
Successful research in this topic should provide new insights into ATM performance and identify specific use-cases to show how these insights can be used to support strategic or real-time decision making in ATM.