Helsing이라는 회사가 만든 AI가 120만 시간의 BVR 전투 훈련 결과를 학습했고 이 양은 조종사 1명에게 140년에 해당하는 양이라고 하네요.
AI는 다음에 JAS-39E 그리펜 자체에 대해서 배우고 전에 배운 BVR 전투를 JAS-39E 그리펜에 맞출 것이라고 합니다.
아래는 12년 전 2013년 파리 에어쇼에서 발표된 것인데 BVR 전투 기동이 자동화하기 가장 어렵다고 하네요.
https://cafe.daum.net/NTDS/51Tp/144
https://aviationweek.com/defense/aircraft-propulsion/ai-agent-assist-gripen-pilots-bvr-combat
AI Agent To Assist Gripen Pilots In BVR Combat
Tony Osborne March 20, 2025
LONDON—Saab is gearing up to install a new artificial intelligence (AI) aid on its Gripen E combat aircraft that will assist pilots in beyond-visual-range (BVR) air-to-air combat.
The AI agent, developed with European AI defense company Helsing, has been trained on the equivalent of 1.2 million flight hr. of BVR training, Peter Nilsson, Saab’s head of advanced programs, told the Royal United Services Institute (RUSI) Combat Air Power Conference here March 20.
“Helsing has trained the agent in their computer system with loads of GPUs [graphics processing units] equal to 1.2 million flight hours, equivalent to 140 years of BVR training for a human pilot,” Nilsson told the conference.
Once trained, the agent would be taught about the Gripen itself, including the aircraft’s performance parameters and sensor capability, and then adapt the learning to the Gripen. “We are probably going to fly pretty quickly [after that],” Nilsson said.
Nilsson did not say how the AI agent would assist the pilot, but Aviation Week reported in November when the partnership with Helsing was revealed that it could support BVR engagements to help increase the probability of kill associated with a missile engagement, perhaps suggesting when to make certain maneuvers such as the so-called crank after the launch of a missile. The system would be in addition to pilot decision aids that already equip the aircraft through what Saab calls human-machine collaboration.
Saab has a 5% stake in Helsing, and the AI company has become involved in several defense aerospace programs in Europe including input into the Arexis electronic warfare system that has been adopted for the German Air Force’s Eurofighter EK aircraft for suppression and destruction of enemy air defenses. Helsing also is working on Project Centaur, a large-scale AI initiative to develop autonomous air combat capabilities for existing platforms and future European combat air programs.
Nilsson said the development is part of Saab’s wider activities into developing software-defined products with AI at their core allowing for rapid prototyping, modularization and scalability.
“AI is the biggest invention since the splitting of the atom. It makes knowledge free, and you don’t need to have been building 85 years to be where Saab is today with AI,” he said.
He said the company’s efforts in making a more open architecture computer system for the Gripen are helping accelerate Saab’s future combat aircraft developments for the Swedish government as it mulls what could replace the Gripen C/D and later the Gripen E in the 2030s and beyond. The Swedish KFS (Future Combat System) initiative is a six-year-long program started in 2024 that will help the government in Stockholm to decide whether to develop something domestically or seek a partnership with other nations.
Saab’s work on KFS includes technology maturation studies as well as demonstrations and concept work looking at options for both crewed and uncrewed concepts.
As part of those efforts, Saab is readying an uncrewed aircraft demonstrator previously codenamed Ruby, images of which were briefly displayed at the conference that showed the aircraft in an advanced state of assembly. The images depict parts of the airframe structure that appear to have a non-conventional appearance, having been produced using additive manufacturing. The General Electric J85-powered, 6-7 m (20-23 ft.) wingspan aircraft has been developed by engineers within Saab’s Rainforest technology accelerator. It aims to prove new-generation production methods and cost-effective avionics using technology based on cell-phone hardware. Nilsson did not say when Ruby might fly.