II - Enhancing final approach management with AI
This research wouldn’t have been feasible even five years ago, now we have the data and the tools.
Jose-Manuel Risquez, ORCI project coordinator
Air traffic managers operate in some of the most high-pressure environments in aviation, and landing approaches are where both workload and complexity are at their peak. The project ORCI(opens in new window), supported by SESAR JU(opens in new window), aims to develop advanced automation support tools in the Terminal Management Area (TMA) domain. The goal is to equip controllers in final approach sectors with information on when to issue vectoring instructions, to ensure optimal spacing between consecutive arrivals in high-density, complex TMA operations. “We wanted to help air traffic controllers manage spacing between aircraft landings, which is cognitively demanding and highly dynamic, using enhanced capabilities that lead to improved operational efficiency, capacity and environmental performance,” says Jose-Manuel Risquez, senior ATM-IA expert at INECO(opens in new window). The project brings together partners from France (ISA Software)(opens in new window), Spain (INECO, ENAIRE)(opens in new window) and Portugal (NAV-PT)(opens in new window) with expertise in air traffic management, artificial intelligence (AI) and aviation technology. It also draws on recent advances in AI-machine learning and leverages large-scale air traffic data to develop practical solutions. “This kind of research wouldn’t have been feasible even five years ago,” Risquez explains. “Now, we have the data and the tools to train models that can actually assist human operators in meaningful ways.”
From prototype to real-world potential
So far, the team has developed and trained prototype models for two airports – Lisbon and Barcelona – each chosen for their contrasting approach layouts: the point merge system in Lisbon and the trombone layout in Barcelona. By testing across two very different configurations, ORCI aims to confirm that its solution can be adapted to a range of operational environments and layouts. Early results are promising. The model’s average margin of error in predicting separation distances is around 0.4 nautical miles (0.741 km), a figure considered operationally useful. “Controllers told us this was a good starting point,” Risquez notes. Work is now shifting to simulation and validation. The project team is integrating the AI model into a simulation platform ahead of further trials with air traffic controllers later this year. Their feedback will be essential. “Ultimately, we want to know if this tool reduces cognitive workload and helps controllers make faster, safer decisions under pressure,” he adds. Success will mean more than a working algorithm, says Risquez. It needs to be a practical tool that controllers trust, improving spacing accuracy, reducing the need for radio comms and enhancing safety. “If they say it helps, and the simulation backs that up with hard data, then we’ll know we’ve built something that matters,” concludes Risquez.