AI Flight Simulations: Training Pilots Faster and Reducing Costs
The U.S. military is incorporating more AI in training and simulation exercises to reduce costs and to train pilots faster.
As global military spending continues to rise, the focus has turned to cost control and how new technologies like artificial intelligence (AI) can help. One area in military avionics that is being looked at is incorporating more AI in training and simulation exercises to reduce costs and to train pilots faster.
Challenging traditional pilot training
“Here is the challenge,” said Pete Morrison, CEO of Bohemia Interactive Simulations (BISim), a provider of training and simulation software. “You want to conduct realistic training exercises in live environments with large groups of troops, but it’s an expensive endeavor. There’s huge fuel, ammo, and personnel costs when putting together live training exercises.”
Getting around these “real-world” training costs for pilots isn’t as solvable an option as it is for aircraft mechanics, electricians, and others who might be able to use a plethora of less expensive online courses for their training. Training pilots is further complicated in military flight scenarios that can involve many complex combat decisions that aren’t in play when you are flying a commercial aircraft.
Recognizing these factors, the U.S. Air Force committed in early 2019 to overhaul its pilot training curriculum. “[There] are two things that are poised to make a revolutionary change in how well we train pilots and in how long it takes us to train pilots,” said Gen. Mike Holmes, commander of Air Combat Command (ACC).
“I want to see how fast and well I can produce experienced pilots,” he said in an interview with Military.com.
Training pilots faster and at a lower cost
To train more pilots faster, and to also more effectively cost-control spending on training, the Air Force is moving to more flight simulation exercises and less classroom learning. The focus will be on thorough flight training and the ability to make intelligent decisions during “real-time” fight operations, including those missions that involve combat.
In real-world flight simulations, the AI technology that facilitates these simulations won’t replace all live training, but it will be used in pilot training and simulations that can give trainees advanced knowledge and training that can expedite the real-world training that follows after — with a goal of producing more pilots faster.
The use of AI and simulations also has the potential to reduce training costs, which can be substantial.
BISim’s Morrison gives the example of military exercises that involve real-world aircraft.
“The cost of a helicopter in the air in Europe is roughly €10,000 an hour,” said Morrison. “Two jets for 15- to 20-minute joint terminal attack control exercises is between $15,000 and $60,000 depending on the airframe used. A single large-scale exercise can run in the millions of dollars depending on the size and duration of the event.”
Pilot Training Next Initiative program
By substituting some real-world flight exercises with AI technology and simulations in its pilot training, the U.S. Air Force has been able to reduce training costs while training pilots faster in its Pilot Training Next Initiative program.
“This program has achieved some stunning results,” said Morrison. “In some cases, the amount of time pilots need to learn to fly has been cut in half. Training times have gone from one year to six months. This has been done by using low-cost commercial virtual reality (VR) headsets and AI-
powered intelligent tutors to help pilot trainees hone their skills. This was achieved at a fraction of the cost of a typical full flight simulator, which costs millions of dollars.”
Nevertheless, there will always be costs that can’t be subtracted from the equation. One of these is the need for human expertise on the tasks to be trained. Also, even if AI is used, the AI still requires a human operator to give the instructions and to program rules and guidelines into the software.
“The work of a human operator with the AI can be arduous,” acknowledged Morrison. “Fully autonomous AI is not yet capable of replicating commander decision-making in realistic scenarios, because there are simply too many variables. This remains a significant area of research, and it is likely to be many years before this problem is solved.”
How AI and simulations work
Although AI and flight simulations can’t address every variable of combat flying, they can help develop pilot decision-making skills in stressful military scenarios.
The AI in avionics does this by using behavior trees for decision-making. The decision tree structure is also easy for instructors and experts to modify because they don’t have to change everything else that is already part of the decision tree, and they don’t have to be AI specialists to make the changes . This allows a non-AI specialist combat instructor or a subject matter expert to focus on the subject matter instead of on the technology behind it.
“Every flight behavior is created in the decision tree, which is easy to modify,” said Morrison. In this respect, AI differs from traditional simulation engines, which are accessible and modifiable only by users of the toolkit who have more advanced AI skills.
New AI breakthroughs
One significant improvement in today’s AI is new navigation technology that enables the AI to more efficiently navigate on maps and find very smooth paths in corridors and buildings. Older AI couldn’t navigate freely inside of buildings. Instead, it could move only along pathways that had to be predefined in an AI simulation model.
“With this new navigation technology, AI will be able to navigate all areas of the environment that a human player could reach,” said Morrison. “With future updates, the AI will be able to recognize changes to the terrain, which means it will still navigate correctly even after a map is modified in the AI editor, or when buildings are created or destroyed.”
In the case of military applications, AI can be delivered as desktop-based training through which units practice tactics against AI-driven opposition forces — or as virtual reality tools in which an air crew working on cockpit procedures is supported by an AI-powered intelligent tutor. “The AI can also simulate full missions where trainees are immersed in complex combined arms scenarios,” said Morrison. “The important thing is that users can customize AI scenarios to their needs.”
While simulations are being run, the AI can also efficiently and automatically collect performance data about them so instructors can determine how well simulation training is being absorbed by trainees and how simulation training content can be further improved.
However, one thing AI can’t do is simulate a human commander, given all of the complexities in the battle space.
“A human still needs to be involved at the high level to assign behaviors,” said Morrison. “The AI is not figuring out how to act or respond without human intervention. You have to give instructions like, ‘AI platoon, go defend that town.’ There is no tech that we’re aware of that can make the decisions of a commander.”
SOURCE: EE Times, By Mary E. Shacklett, President, Transworld
VIEW ORIGINAL: https://www.eetimes.com/ai-flight-simulations-training-pilots-faster-and-reducing-costs/#