ISOBAR – Research is building controller confidence in AI systems


Gilles Gawinowski, Camille Anoraud and Ramon Dalmau Codina, from the EUROCONTROL Innovation Hub and Stéphane Pierre, from the EUROCONTROL Network Manager Directorate talk about ISOBAR, winner of the SESAR Exploratory Research Award.

In March 2023 EUROCONTROL’s Innovation Hub ISOBAR programme was one of two Agency research projects to take home a SESAR 3 Joint Undertaking Digital European Sky Award, winning the Exploratory Research category.

Some research programmes have outcomes that are far more important than the basic project aims. The ISOBAR research work has been strategically important to EUROCONTROL for three main reasons: it gives controllers an important new tool in calculating the impact of severe weather incidents on capacity; it has moved from exploratory research to operational use at an unusually fast pace; and it is building confidence among operational staff that artificial intelligence (AI) tools can be usefully integrated into current systems and procedures.

ISOBAR – or "Artificial Intelligence Solutions to Meteo- Based Demand/Capacity Balancing (DCB) Imbalances for Network Operations Planning" – research has centred on the use of machine-learning to predict the severity and probability of convective weather events and their impact on en-route capacity.

“When convective weather began crossing Europe, we knew it would affect air traffic management capacity in Europe and that each sector would be impacted in different ways,” said Gilles Gawinowski, one of the ISOBAR research leaders based at EUROCONTROL’s Innovation Hub. “The simple idea was to use AI to predict the capacity reduction and detect how each area would be affected. In that way we could prevent chaos.”

It quickly became clear that if the system could predict expected traffic and required resource levels for convective weather, it could also be used for other capacity-impacting incident. A series of research objectives were defined (see ISOBAR: project aims, conclusions and members box) and ISOBAR work was launched in May 2020 as part of a Horizon 2020- funded SESAR JU research initiative. The budget for the programme was EUR 2 million. Essentially, the aim was to develop a system incorporating innovative AI-based components to predict convective weather up to 36 hours in advance and then forecast the impact of traffic demand on capacity while providing automated solutions.

May 2020, during one of the first peaks of the COVID pandemic, was hardly optimal timing for such a wide-ranging, complex, collaborative innovation programme, involving a network of stakeholders including engineers, data scientists and operational experts.

“We had to work remotely, via Teams, but, surprisingly, the process worked out very well,” said Stephane Pierre, from EUROCONTROL’s Network Manager Directorate, who worked with Innovation Hub colleagues on the project. “We had the validation at the end, in Bretigny, but all the rest was done remotely.”

Integrating AI tools within an operational air traffic management infrastructure had to be carefully planned, to understand how humans and machinelearning systems could interact in an optimal way – which was one of the most important challenges of the programme.

“The goal was to translate all the data into something meaningful for air traffic controllers because engineers speak a different language, which is fairly normal,” said Stephane Pierre. “So the researchers had to develop a new interface and design it in a way which would process the data and deliver different scenarios.”

“Operational staff currently calculate capacity reductions in their heads, based on their experience in the job,” said Camille Anoraud, the EUROCONTROL Innovation Hub researcher who piloted the solution. “And what we provided were real numbers. So during the validation exercise we wanted to see if our capacity reduction propositions matched their own views and experience. It was interesting to see whether the algorithm proposed to them aligned with their own forecasts.”

“The goal really was to provide them with realistic solutions, so they could also picture themselves using it in the short or medium term.”

“When you are designing a system which shows predicted capacity, you cannot do this in a too futuristic way because operational staff will neither accept nor understand it,” said EUROCONTROL Innovation Hub researcher Ramon Dalmau-Codina, who developed the solution. “We had to find a way not to transition from the proven technology to something entirely new. So we had to develop new algorithms which took a step-by-step approach in moving from legacy systems to this new technology.”

Currently, calculating demand and capacity balancing is done by humans in different ways in different countries.

“So for me the most difficult part of the challenge was to merge the current system of human calculations with a high number of cherry-picking measures which are completely automated by the machine and not determined by humans,” said Ramon Dalmau Codina.

Forecasting an accurate capacity figure as a result of weather incidents and presenting this to controllers in an optimal format was just one of the interesting challenges the team faced. Other issues included looking through the different timelines for the predictions to arrive at the most stable prediction through a defined time slot and, inevitably, ensuring that the weather forecasts were as accurate as possible.

“In all the work we did we had to rely on weather data,” said Ramon Dalmau-Codina. “So, even if we built a very good model that could map weather to capacity reductions, if the weather forecast was not good then the predictions were not going to be good, either. So the project underlined the fact that good weather forecasts are vital for ATM and we need to work very closely with the meteo agencies.”

“We had some pleasant surprises in the validation exercise,” said Stephane Pierre. “The weather predictions were fairly stable. In terms of capacity reduction calculations we had no idea whether they would correspond to reality, because it was so innovative. In fact, the figures that we got from the system were considered as fairly reliable and perfectly in line with what the controllers were expecting. So this has reinforced our confidence in these types of tools, which I think is very important.”

The simulations and final validation exercise took place with operational partners at the Innovation Hub and proved the technical feasibility and operational acceptability of ISOBAR’s new collaborative framework for a network of stakeholders which included the EUROCONTROL Network Manager, area control centres and airspace users.

“The research has finished but some of the elements of the project are being implemented now in our systems and this summer we have been testing some of the elements of the capacity-reduction AI tool,” said Stephane Pierre. “We have already started to implement the interface and we will be able to check all these tools against the reality for this summer. What we haven’t done yet is all the coordination aspects with the flow management positions (FMPs) and so on but that will come later. The outcome of the project is that the Network Manager is already starting to work with this capacity reduction calculation tool – very soon after the results of the first research outcome.”

ISOBAR project aims, conclusions and members

The ISOBAR research programme, which ran from May 2020 to December 2020, aimed to develop a new AI-based service to support the Network Operations Plan by integrating enhanced convective weather forecasts for predicting imbalances between capacity and demand. The programme exploited artificial intelligence (AI) to select mitigation measures at local and network level in a collaborative Air Traffic Flow and Capacity Management (ATFCM) operations paradigm. To achieve this vision, four objectives were set:

  • Reinforce collaborative ATFCM processes at pretactical and tactical levels into the LTM (local) and Network Management (network) roles integrating dynamic weather cells.
  • Characterisation of demand and capacity imbalances at pre-tactical level depending on the input of probabilistic weather cells by using applied AI methods and ATM and weather data integration.
  • User-driven mitigation plan considering airspace user priorities (and fluctuations in demand based on weather forecasts) and predicted effectiveness of ATFCM regulations, considering flow constraints and network effects.
  • Develop an operational and technical roadmap for the integration of ancillary services (providing AI-based hotspot detection and adaptative mitigation measures) into the NM platform, by defining interfaces, functional and performance requirements.

The ISOBAR project integrated accurate and probabilistic convective weather forecasts in the ATFCM process. These weather forecasts are an input to demand and capacity prediction, so that imbalances between capacity and demand can be better anticipated and more adequate mitigation measures can be prescribed to ensure safety and maximise efficiency and capacity. The overall outcome was an adaptive network plan, prescribing alternative solutions depending on the evolution of demand, capacity and weather. The main building blocks around which the work in ISOBAR has been articulated were:

  • An enhanced ATFCM process targeting weatherbased imbalances;
  • A meteo engine providing probabilistic forecasts of convective storms;
  • Data integration from diverse ATM and meteo sources;
  • Precise and granular characterisation of demand and capacity imbalances, including capacity decay prediction and the characterisation of potential airspace user demand responses to disruptions;
  • Automated DCB mitigation solutions boosting human operator capabilities;
  • Demonstration of the feasibility and usability of the ISOBAR Solution, addressing both an operational and a performance evaluation.

“The very promising ISOBAR concept should be pushed towards a higher maturity level and further improvements in performance/maturity of the AI-based components and DCB workflow” according to the conclusions.

Research consortium members comprised:

  • Centro de Referencia de Investigación, Desarrollo e Innovación ATM, A.I.E. (CRIDA)
  • Universidad Carlos III de Madrid
  • Cranfield University
  • Ecole Nationale de l’Aviation Civile
  • Direction Générale de l’Aviation Civile/Direction des Services de la Navigation Aérienne
  • Meteo-France
  • Agencia Estatal de Meteorología
  • Sopra Steria Group
  • Earth Networks
  • Swiss International Air Lines AG

Get our latest issue of Skyway

Explore our latest articles and download the full issues.

Latest highlights

LARA news item banner

EUROCONTROL releases new version of Standard Inputs for Economic Analyses


New call for operational challenges for our user-driven innovation projects

Press release

EASA & EUROCONTROL sign MoC to enhance cooperation for the safe and sustainable future of European aviation

Man flying drone

Continuing U-space ConOps development in the new CORUS five project

Eduardo Santander

Towards a unified sky vision: integrating ATM & AAM for future air transport

Julie Garland

The future of Europe's aviation sector: embracing innovation