The ASTRA tool predicts potential airspace bottlenecks up to one hour in advance and suggests how to avoid them. The operators who tested it reported high levels of satisfaction, with both its performance and its clear, intuitive interface. The latter of which is the result of Deep Blue’s expertise in user-centered design.
Last August, the technical phase of the ASTRA project was completed. The project aimed to develop an AI-based tool capable of supporting Flow Manager Positions (FMPs), specialised operators responsible for managing air traffic flow in a given sector or region of airspace, in situations of high complexity.
“In aviation, a situation is considered complex when a sector’s capacity, is exceeded, making corrective actions necessary, such as altitude changes, route deviations, or speed adjustments,” explains psychologist François Brambati, company project manager and Human Factors (HF) consultant at Deep Blue. “Complexity assessment takes into account many factors, including aircraft speed, type, and size. Identifying these scenarios well in advance makes it possible to intervene before safety risks increase.”
Until now, air traffic hotspots were typically identified with an average lead time of about 20 minutes. With the ASTRA tool, this can be extended to up to one hour.
“One of the most innovative aspects of the project is certainly the forecasting time horizon,” Brambati continues. “The tests, carried out in a simulated environment based on real historical data provided by Eurocontrol and referring to Swiss airspace between 2021 and 2024, involved ten operators: three FMPs, three Air Traffic Controllers (ATCOs), and four ATCO Supervisors (SPVRs). All confirmed that the hotspots identified by the algorithm corresponded to realistic and potentially critical situations, and that the proposed solutions effectively helped reduce sector complexity.”
Strong performance, clear user interface
Another distinguishing feature of the project is the innovative user interface designed by Deep Blue to make the AI system intuitive, accessible, and genuinely useful for FMPs.
“We didn’t start from off-the-shelf solutions, but from an in-depth analysis of operators’ real needs through interviews and user workshops,” the psychologist emphasises. “These needs were first translated into technical development requirements and then led to the design of an interface that introduces, alongside the traditional histogram charts commonly used in this type of display, new visualisation modes such as radar maps and time sliders. These allow users to explore how air traffic will evolve depending on the actions proposed by the algorithm, helping FMPs choose the most effective and safest option.”
The algorithm not only suggests what to do, but also explains why, in line with the AI principle of explainability. Each proposed solution is accompanied by a clear explanation of how and why it helps reduce complexity (to what extent and with what level of ‘robustness’), strengthening user trust in the system. The operators involved in the tests responded positively to this graphical presentation, while also recognising the need for training, especially for users without an air traffic control background.
“One of ASTRA’s main strengths, as highlighted in the post-test questionnaires completed by FMPs, is precisely the human–machine interface design: the clear, clean visualisation and simple layout make the new operational context more efficient by supporting faster decision-making,” Brambati adds.
The suggestions mainly concerned changes in terminology, such as using symbols instead of letters to indicate trajectory modifications, as well as colour schemes and the layout of some of the elements displayed on the screen. It was also strongly recommended that the system be integrated with existing tools to avoid operators having to monitor an additional screen, which would otherwise increase operational complexity.
“Deep Blue’s contribution once again demonstrates the importance of integrating design, social sciences, and technology to make innovation understandable, usable, and truly valuable for the people who adopt it,” the expert notes.
From Theory to (Future) Practice
Overall, ASTRA has demonstrated the effectiveness of the concept from technological and human-centred perspectives, with positive feedback from operators. However, despite validating the concept, the project remains exploratory research at a low TRL (Technology Readiness Level). This means that no digital tool will be supporting FMPs in operational environments in the near future.
“The algorithm was implemented and tested locally in the Geneva area,” the Deep Blue consultant explains. ‘Scaling it up to a national or European level would require the input data to be recalibrated and the system to be upgraded, as well as the communication flows between control centres to be redesigned. For this reason, one of the project’s final recommendations is to shift future focus from local management to a broader network-management perspective.”
The effectiveness of systems like ASTRA largely depends on large-scale adoption, ideally by multiple Area Control Centres: a necessary condition to speed up communication and reduce negotiation time and effort.
“ASTRA represents the first step towards an air traffic management system that is increasingly oriented towards network management,” concludes Brambati. ‘If new research topics emerge in line with these results, the next step will be to move the solution towards higher validation levels, with the goal of turning a proof of concept into a tool that can be used within the future air traffic management system.”
