Software-Defined Hardware Could Transform Defence Development, say Leonardo and PhysicsX
At the Resilience Conference in London, speakers said AI-driven, software-defined hardware could revolutionise defence manufacturing and drastically shorten aircraft development cycles
Leaders from Leonardo and PhysicsX say software-defined hardware could mark a turning point for the defence and aerospace industries, drastically reducing design and testing times for complex systems such as aircraft and jet engines.
Speaking during a panel at last month’s Resilience Conference, Leonardo’s VP of aircraft technologies Massimo Maroni and PhysicsX co-founder Jacomo Corbo discussed how artificial intelligence and simulation are reshaping engineering processes.
Maroni said software is “really shaping a new world for defence in two main aspects.” The first, he explained, is embedded software within products.
“We need very flexible software that is easy to be updated,” he said. The second is the use of software tools throughout the development chain. “The development is really based on software. We have model-based system engineering. We have model-based design. It’s a very important part of how an aircraft is built.”
Maroni noted that traditional aerospace development cycles are too slow to meet modern demands. Referring to the Eurofighter programme, he said: “There was a slide on the Eurofighter that said, plus five years – I will say that was optimistic. If I took the development time for the Eurofighter, I think it was double. And we can’t afford any more that.”
To shorten timelines, he said, “we need better tools, we need faster tools,” calling for “tool chains for development [to become] faster and faster.” This means progress on both hardware and software fronts: “The GPU revolution is already there. We need to develop software that is suitable to exploit the GPU revolution on one side, and on the other side, we need smarter software.”
Leonardo’s collaboration with PhysicsX, he said, focuses on reducing the time needed for computational flight and dynamic simulations, one of the most time-consuming parts of aircraft design. “What we have done with Jacamo is to build a different way to approach that… to build a much faster model based on deep learning surrogate so that we can use this surrogate model to do the preliminary phase of the design,” Maroni said.
Corbo described PhysicsX as “building a new engineering simulation software stack… to enable much more software-defined hardware innovation.” The company’s technology uses AI models trained on large datasets of numerical simulations to predict physical behaviours far more quickly.
“It’s very compute-intensive,” he said of conventional simulation. “It takes a lot of time. A lot of how engineers spend their time is around manipulating meshes.” By contrast, AI-driven inference can accelerate the process, making computational fluid dynamics simulations “10 to 50,000 times” faster, he said, and electromagnetic field simulations may run “closer to 800,000 times faster.”
PhysicsX, which raised $135 million this year, has positioned itself as a dual-use company serving both defence and commercial markets. Corbo said investors are increasingly comfortable with that balance. “Maybe that might have been an issue a few years ago,” he said.