EUROCONTROL recently hosted a webinar to build on last year’s FlyAI Report to showcase some of the activities utilising big data, machine learning and AI. Presenters represented Airbus, EUROCONTROL, EASA and IATA.
Airbus kicked off the discussion with the topic of Data Sharing for Artificial Intelligence and three programs that are underway related to this topic. The first is aimed at improving operations for A380s by sharing information between Airbus and airlines who are operating this aircraft. Machine learning and systems expertise are being combined in order to develop prognosis and health management. The second program is working to improve aviation operations. Cloud, big data, AI, data analytics, data governance and ontology work together to support Skywise Digital Services. One key outcome is related to predictive maintenance, as AI can identify abnormal behaviors faster than humans reviewing data. The third program is developing more autonomous aircraft. Two topics are currently in evaluation stages. The first is speech-to-text ATC. Challenges still to be addressed include understanding of different accents, providing multi-lingual support, adjusting for background noise and fast speakers. The second is related to vision-based navigation for autonomous taxi, takeoff and landing (ATTOL). Challenges still be addressed here include the impact of severe weather, day/night lighting and sensor reliability.
EUROCONTROL members managed 11 million flights in 2019. That’s a lot of data that can potentially be used to improve operations. AI is a necessity in order to interpret this amount of data and prepare analyses that could be helpful. Considering how AI could help to do this is focused in four areas.
- Better flight predictions: arrival predictions, capacity planning
- Cleaning submitted flight plans
- Anticipating knock-on delay impacts: alerting to possible curfew blocks, decreasing passenger disruptions
- Cyber resilience: threat anticipation and detection.
This is a two-way street though. AI needs to be powered with data and in return solutions can help EUROCONTROL members with their business operations. There is an R&D archive with information from 14 million flights that is free to access for R&D purposes. Network Manager B2B provides live operational data to stakeholders. Collaborative walled gardens are being developed and used – ie Airbus Skywise and EASA d4S. The best practices that have been identified through this work are as follows:
- Share analysis vs bulk data
- Use a common language – SWIM for example
- Standarise the report requests and response process.
EASA spoke about the Data4Safety (d4S) proof of concept. D4S gathers expert knowledge including flight data, traffic data, safety reports and weather data resulting in a 'big data' pool of knowledge. Collaborative aviation experts and data analysts work within the analysis platform to provide metrics, studies, benchmarking and discoveries. In summary, ‘actionable safety vulnerability intelligence.’
IATA data exchange programs provide data governance, data management and collaborative applications. AI is used to help airlines with strategic decision making. Input includes levels of demand, customer digital experience measurements, network planning, predictive maintenance and people services. Output is improved safety, operational efficiency and sustainability.
These steps are moving us closer to the reality of AI in aviation. Although still working mainly behind the scenes with data analysis these are important steps to understand how/where AI can fit. EUROCONTROL has two trials underway to understand topics which could trigger a next step. Traffic prediction improvement at the MUAC. AI is assisting to help improve predictability of traffic, reducing ATCO resources needed for this and reducing the uncertainty of capacity planning. AI also assists in integrated flow management position (iFMP) to improve climb/descent predictions in 4D trajectory. The last one has just passed the prototype phase. You can learn more about these projects on the EUROCONTROL website.
