More than a century after the invention of the autopilot, aerospace engineers are still working to bring more automated processes into aircraft cockpits to enhance safety, increase efficiency, and reduce pilot workload. With the help of artificial intelligence (AI), autopilot technology has evolved from simple devices that maintain an aircraft’s altitude and heading to fully autonomous flight control systems capable of performing gate-to-gate operations without any human input.
One way or another, almost every aerospace and defense company is looking to exploit the potential for AI to improve their aircraft and other systems. This ubiquitous and rapidly morphing technology sphere has sparked extensive discussions at major international aerospace events recently, such as this year's Paris Airshow and the annual European Business Aviation Convention & Exhibition.
In May, EASA released its new AI Roadmap 2.0, which is intended to advance "the human-centric approach to integrating artificial intelligence in aviation. The updated document incorporates progress achieved in the field since the publication of the air safety agency's first roadmap was published in February 2020, drawing on AI use cases involving aerospace companies, research centers, and academics.
According to EASA, the new roadmap, "provides a comprehensive plan for the safe and trustworthy integration of AIN in aviation, with a focus on safety, security, AI assurance, human factors and ethical considerations." In May, the organization also published a new report on research it commissioned into machine learning application approval, which highlights approaches to evaluating and certifying AI-based systems.
But AI isn’t only changing the way airplanes fly—it’s transforming nearly every aspect of aviation on the ground, too. As AI and machine-learning technology have matured in recent years, the aviation industry has explored ways to capitalize on it by making processes more efficient and often safer.
For example, aircraft manufacturers and service technicians can use AI software and robots, including language learning models like ChatGPT, to streamline assembly and maintenance, repair, and overhaul (MRO) processes. Airlines and other operators can also use AI for fleet optimization, flight planning, and ground operations. Engineers developing aircraft can use AI tools to facilitate and speed up the design and certification of products before they even hit the market.
AI isn’t just the way of the future. The aviation industry already has used at least some primitive form of AI technology for years, particularly for manufacturing and MRO. Traditional AI relies on human programmers to define rules and algorithms for pattern-matching and decision-making processes, and it can analyze large datasets much faster than humans. For example, MRO providers might use AI to analyze data from the various sensors onboard an aircraft to predict potential maintenance needs before they arise.
Generative AI Is Changing the Game
Although the aerospace industry already widely uses AI for various applications, it has only just begun to make an impact. In the coming years, new applications will begin to emerge as companies find ways to take advantage of generative AI—models like ChatGPT that use machine learning and deep neural networks to generate outputs not predefined by human programmers.
“AI is already helping both manufacturing and repair and maintenance users to work with robots much better,” Rishi Ranjan, founder and CEO of GridRaster, told AIN. GridRaster is a software company that specializes in “extended reality” technologies, like augmented reality and virtual reality, that employ AI and spatial mapping software. It provides such tools for the aerospace and automotive industries and works with several top U.S. defense contractors.
According to Ranjan, the U.S. Department of Defense and its top contractors already use traditional AI tools, but generative AI has the potential to make a much bigger impact on defense applications, as well as on the wider aviation industry. “We strongly believe that generative AI will really start helping scale these things to the much bigger aerospace industry in the next two to 10 years,” he said.
ChatGPT and other language-learning AI models like it generally are adept at disseminating an astronomical amount of information to yield relevant and—mostly—accurate output almost instantly. It can tell one how to build and certify an airplane, provide tips for improving the aerodynamics of an airframe, and even generate maintenance schedules for specific aircraft.
But the greatest value of generative AI models like ChatGPT will come when aerospace companies begin to verticalize the technology, integrating it with their own intellectual property for internal use, Ranjan said.
While ChatGPT and other generative AI models can access all the information publicly available on the internet, they don’t have access to companies’ valuable, private intellectual property. Giving AI access to that highly protected information would open up a world of new use cases for AI across the industry.
“For the real use, what will happen is that companies will have to pay for an AI model like ChatGPT and start training it—whether ChatGPT enables that or someone will come up with a solution—so that you can take this massively large learning model and now start training it with proprietary data,” Ranjan said. “That will be true for pretty much every enterprise where IT is very important.”
AI Makes Faster, Better Digital Twins
While ChatGPT is a language-learning model that only outputs text, generative AI can also create images and 3D models. In aerospace, that can be particularly useful for generating digital twins.
Aircraft developers and MRO providers alike nowadays often rely on sophisticated virtual models known as digital twins to simulate products, like aircraft and their various subsystems, in a digital environment. Engineers can leverage digital twins to speed product development timelines by reducing the need to physically build and test things, thereby minimizing costs. MRO technicians use digital twins for predictive maintenance and to detect anomalies by comparing real-world sensor data to the data generated by digital twins.
While digital twins can help to save time and resources, they’re also expensive and time-consuming to create. But generative AI will soon make the process of building digital twins much faster.
“Traditional AI is still very manual, and a digital twin is an extremely manual process to build,” Ranjan said. “Large AI models like ChatGPT, once you can verticalize these for the aerospace industry, can remove a lot of that manual work. They can look at text data and image data and start helping you create a digital twin for these automatically.”
According to Ranjan, generative AI will soon allow companies to build digital twins for just a small fraction of what it costs today. For every $1 spent on building a digital twin with traditional methods, “in another three to four years you're looking at like 10 cents,” he said. “In another 10 years, it might be one cent.
“Now these expensive solutions will start getting into the hands of more users,” Ranjan added, noting that he expects just about every aerospace company to be using some form of internal generative AI technology within two to three years.
Will AI Take Our Jobs?
As with just about any other industry, the impact that AI will have on the job market is not yet clear. Robots have already taken over some tasks that humans originally performed in the aerospace industry, and new autonomous airplanes will reduce the demand for commercial pilots.
However, AI has the potential to create jobs that didn’t exist before. Those new roles might involve maintaining AI systems for both aircraft and ground operations, developing algorithms, and ensuring that AI gets used ethically and responsibly.
According to Ranjan, aircraft manufacturers and technicians don’t need to worry about robots taking their jobs. Rather, he believes that AI will change the way they work. “The human in the loop is always going to be there,” he said. That’s because AI, while good at pattern recognition and making predictions, will never improve on human perception, he explained.
“If you want the best efficiency in aerospace, because of the high [amount of] intellectual property and very large knowledge base that is needed to operate these things, it will always be a complementary relationship” between machines and human staff," he said.