The DIALOG project is one of several SESAR projects related to artificial intelligence (AI) in air traffic management (ATM). It started with defining use cases and objectives and now comes to the realisation phase. I discussed the status of the project and the introduction of AI in air traffic controller’s (ATCO) tools with Yannick Migliorini, the project’s technical leader.
There are many ways AI could be introduced in ATM systems and DIALOG picked a specific use case to test: supporting ATCOs when replying to pilot requests. This might seem simple, but a successful implementation needs to fulfil multiple tasks:
- Understand the request based on what the pilot said – typically a change of flight level or a radar vector to avoid bad weather
- Analyse the traffic situation to determine if the request can be accepted, or must be denied, or if an alternative can be proposed
- Finding the right moment to present the ATCO with the result of the analysis.
Recent progress in speech recognition supports the first step. The second step is the easiest of the three, especially given the power of modern computers. The hard part is to find how and when to present the results of the analysis. From a human perspective it is very easy, but not from a computer’s perspective. We instinctively know when to speak to someone, especially a colleague, or when it is better to let them finish what they are doing, or if its better to not say anything at all.
As Yannick summed it up: timing is an important part of teaming. A suggestion coming too late isn’t helping, and if it comes too early, it can be a nuisance, especially if the ATCO’s cognitive load is too high. The DIALOG team decided to have AI suggest actions using variations of the existing radar HMI and not with a synthetic voice, to avoid risking sensory overload.
Evaluating the ATCO’s cognitive load remains the key challenge. Traffic complexity is not a good proxy. For example, a complex situation which has already been solved is not creating a high load. The DIALOG team decided to use sensors capturing bodily signals as a basis to evaluate the ATCO’s cognitive load. Eye tracking devices, heartbeat sensors, and even head-mounted electro-encephalogram sensors. This full instrumentation is not usable in real-life operations and the team hopes to replace them with lighter sensors in the future, maybe something like a connected watch. At the moment, DIALOG is aiming to achieve Technical Readiness Level (TRL) 2, and extra instrumentation is acceptable.
The project is at a critical and exciting stage as proof-of-concept validations will run in March 2026 and the results will be analysed and presented in June 2026. The team focuses on executive ATCOs and planners could be considered too at a later stage. Any change in the tasks of one the two ATCOs managing a sector together impacts the others. Future deployments at the scale of a full centre will open new possibilities, with a single AI assistant servicing all sectors at once, therefore facilitating inter-sector co-ordinations.
Yannick encourages air traffic controllers and other professionals to embrace and shape the progress of AI, emphasising that it is a tool that, if designed well, can be used to improve safety for controllers, pilots, and passengers.
