DataBeacon develops a complete end-to-end data solution for training AI tools for Air Traffic Controllers

As AI has advanced, so too have the number of AI R&D initiatives in aviation—including air traffic control. That said, deep learning algorithms still depend on reliable training datasets, and these can be difficult to acquire in many industries. In fact, in the majority of cases, manually annotated training data is the only option, meaning expensive, slow and inefficient exploitation of AI. It’s clear that we need a new way of ensuring reliable data if we want to take full advantage of the capabilities of AI, especially in aviation.

With the use of technology and data provided by FlightAware, DataBeacon has developed an advanced platform of data pipelines specific to air traffic control. This platform is capable of analysing air traffic data in real time and precisely annotating potential air traffic conflicts. Whenever a potential conflict is identified, the platform automatically annotates multiple features, resulting in a rich dataset that can empower deep learning algorithms. Simply put, this technology unlocks AI potential in aviation.

The DataBeacon platform can identify changes in aircraft trajectories, including heading or altitude, and annotates the commands given to different flights to resolve conflicts or optimise aircraft trajectories in the course of flight. Make no mistake—this will reorient how we think about AI in air traffic control today, especially given strict aviation safety standards and challenges of scarce capacity in high demand airspace.

How has this technology scaled? Currently, the platform is processing up to 7.000 flight data positions per minute, just within EU. On average, the platform processes a total of 5 million flight position messages and around 150k flight plan messages daily. By the time you read this, the dataset will have increased by 21.000 new data positions!

With the use of DataBeacon’s platform and data pipelines, DataBeacon can support an AI digital twin engine that powers our digital assistant, Victor5. When complete, Victor5 will be the first aviation digital assistant capable of supporting air traffic controllers with proposed conflict resolutions and flight trajectory optimisation based on state-of-the-art AI and machine learning techniques. Victor5 will soon be released in a closed beta version.