Bringing AI on board aircraft

Over recent months and even years now there has been much talk of Artificial Intelligence - especially generative AI - affecting all walks of life. But what are the implications for systems on board aircraft, how is AI currently being applied, and how can these technologies and capabilities be legitimately certified?

Firstly, how is AI disrupting legacy avionics technologies? Well, the world of certified avionics is naturally cautious, given how little it likes the unknown or rapid evolution. And yet AI brings with it many potentially positive disruptions, including continuous improvement and an unprecedented ability to manage huge amounts of data and to perform complex reasoning.

Moving forward quickly to gain experience and data

At Thales we believe that we need to move forward as quickly as possible with the first versions of AI capabilities to gain valuable experience and real data to create new value and accelerate the adoption of these new capabilities. In practical terms, today, on board aircraft, the systems developed by Thales draw solely on the use of fully explainable optimization-based AI. This is only applied to non-critical systems to serve as a decision-making aid with a view to, say, optimizing an aircraft’s trajectory or minimizing fuel consumption.

Looking forward, hybrid AI, drawing on a combination of knowledge and data, holds tremendous promise. The use of hybrid AI would mean that, for instance, when a pilot has to manage an unplanned diversion, whether due to a technical malfunction or weather event, the system can draw on the current situation along with past occurrences and results, applying the necessary algorithms and procedures to propose the ideal way to advance. This information could then be fed into the Flight Management System for pilot approval. This full-scale AI-powered pilot assistant concept was showcased at last year’s Paris Air Show (pictured below) and was particularly well received.

Explainable AI has also been put to the test on The Perlan Project’s scientific glider program. During a recent campaign, AI capabilities overseen at ground level by a Thales team delivered vital predictive data and brought unparalleled recommendations to the quest for stratospheric heights, significantly reducing the time required to reach extreme altitudes. This optimization allowed for more efficient use of oxygen and battery capacity, transforming the flights into safer, more efficient, and more predictable operations.

Many exciting advances are also being made in the military field where we are working on a tactical mission assistant. It will shortly be entering its development phase and could prove crucial in future high-intensity combat situations by recommending optimized multi-constraint mission plans in alignment with the high-level intentions provided by pilots.

Delivering the computing power

With regard to the use of AI on board aircraft on a mid-term basis, points to be considered include how to deliver the increased computing capabilities the systems require. Players such as Thales are exploring the use of high-performance platforms with optimized computing power, as well as how to incorporate Edge computing concepts to cater for these energy-hungry algorithms... or else how to develop AI capabilities that are far more “frugal”. ?

Meanwhile, a stringent certification framework is obviously essential. European initiatives covering AI at large have entered into force since 2021, with advances expected from early 2025 onwards with regard to AI applied to aeronautics. EASA, the European Aviation Safety Agency, has notably established a concept paper that will bring authorities and industry players together on standardization convergence.

At Thales we are proud to be an active contributor to European workgroups including those led by EUROCAE, with members representing a number of our units who are working closely together on AI. We are also collaborating with academia through programmes including DEEL (Dependable & Explainable Learning) led by the Saint-Exupéry Technological Research Institute and the Confiance.ai programme led by IRT SystemX.

AI is also a lever for in-house requirements

In-house at Thales we are focused not only on how AI can be deployed on the products we deliver, but how AI can be applied to our everyday activities to achieve greater levels of competitiveness. This is a subject which is key across the board for us.

Finally, applying AI to our field is also a means of maintaining our attractiveness to students and engineers who are understandably motivated by the application of these new technologies to the challenges we face.

Looking forward, we can justifiably feel excited about what AI can bring, whilst proceeding with caution and being aware of the associated drawbacks (including data property issues, short obsolescence cycles, etc.). As things stand, when AI is injected into onboard systems, it can only be used as a valuable aid… provided the technologies employed are explainable, transparent and ethical! Let’s work towards harnessing and developing those capabilities!