Could AI Soon Replace Aircraft Technicians, Designers…and Pilots?

The airline industry is currently in chaos. Extensive staffing shortages have led to countless flight delays and cancellations. In recent months, one out of every five flights was delayed, and 2022 has already had more flight cancellations than the entirety of 2021.

But could the answer to these disruptions be by artificial intelligence? Industry insiders seem to think so. AI has already made serious inroads in the aviation field, whether it's collecting and analyzing flight data, monitoring pilot reaction times, or predicting maintenance issues.

But some ambitious entrepreneurs posit that AI could soon fly the planes themselves. In August 2020, an AI algorithm defeated an F-16 pilot in a series of simulated combat scenarios. The AI won 5-0.

In a comparable simulated scenario in China, the human pilot was able to triumph, at least at first. However, the underlying premise of AI is that it learns as it goes. By the end of the military demonstration, the AI effortlessly defeated the pilot; it was able to observe his flying style, copy his tactics, and improve upon them.

Self-driving cars are starting to hit the streets, something that was previously thought impossible. So, exactly how far are we away from AI in aviation and autonomous aircraft? And could AI become a catalyst for a sorely needed overhaul of the aviation industry?

Well, it depends.

Implementing machine learning

Some degree of artificial intelligence is already in use in some aircraft, whether it's commercial or military. Onboard and ground-based systems are using machine learning (ML) to gather information as they operate and potentially learn from every flight. AI is deciphering patterns in air traffic data to find the best ways to mitigate congestion and allow planes to seamlessly maneuver around each other.

AI systems are also being used to reduce fuel usage and emissions, and thereby complying with the latest environmental regulations, by directing the plane on the most efficient flight path based on real-time conditions.

A bit further down the road, Airbus is trailblazing the way towards autonomous flight. The corporation recently completed its Autonomous Taxi, Take-Off, and Landing (ATTOL) project. In real-world scenarios, over 500 test flights, Airbus's AI system operated airplanes independently during take-off, landing, and taxiing without any assistance from pilots on board. Airbus said the project was intended to pave the way for autonomous technologies to handle routine aircraft operations and allow pilots more time to focus more on strategic decision-making.  Not to be outdone, Boeing is also sponsoring AI R&D, acquiring AI-based companies, and even cooperating with Airbus and industry leaders on independent AI research.

Improving safety

Data-driven algorithms have also gone a long way in improving safety regulations and protocols. AI is now being used to manage air traffic, identify maintenance issues, and even helping to design airplanes.

AI systems are capable of designing the most efficient and durable airplane structure based on the parameters set by an engineer: weight, size, capacity, etc. The AI can then test millions of potential aircraft shapes and structures and narrow them down to a few, thereby giving engineers a handful of final choices to select from.

The AI-human design partnership guarantees the most resilient of aircraft, both mathematically and in practice. It's no surprise then that 97% of aviation companies are investing in big data and AI as a means to improve air safety.

Yet the biggest barrier to implementing AI, whether as a pilot or engineer, is a lack of trust, which is somewhat ironic as AI can make flying substantially safer. In one research project, the Department of Defense tested seasoned fighter pilots in a flying simulator operated by AI tested seasoned fighter pilots in a flying simulator operated by AI.

The pilots had the option to take over if they thought the AI was underperforming. The project monitored the pilots' heart rate, facial reactions, and stress levels to see how much they trusted the machine, as well as whether they unnecessarily switched to manual control. The study found a significant trust barrier between pilot and machine.

Trust, or lack thereof, in technological adoption will always be a persistent barrier to progress. But that trust can increase when operators demonstrate compliance with rigid safety certifications, such as those in the DO-178C guidelines for avionics certification, which lays out 71 critical Objectives for software on airborne systems.

Pilotless planes?

The beauty—and perhaps biggest downfall—of AI is that it's designed to react in a split second before a human operator can even realize there is an issue. Imagine if an airbag was deployed only after the driver processed what was happening, or if the driver had to deploy it manually. The bag would deploy far too late to be of much use.

In a plane, this means the AI will make and execute a decision before a pilot can react; the action could be lifesaving, but, as of now, the AI also can't weigh context and morality. While AI systems may react faster, they can't make subjective decisions or understand moral choices. How does it decide, for instance, whether it would be better to hit the bird, highway, or endanger passengers?

AI piloting systems operate by comparing their programmed maps to their fields of vision in real-time. But this system isn't foolproof: a small harmless object may cause the AI to veer off its path.

So currently, commercial airlines are looking towards AI technologies to improve flight operations and performance, but not necessarily to reach autonomous flight, at least for now.

Although we may still be far off from entirely autonomous planes, AI is rapidly becoming a valuable co-pilot and engineer, making flying a little less taxing and a lot more smooth. Stand by for continuous new AI announcements which are improving most aspects of aviation.

Author:

Vance Hilderman, Aviation Expert, Author, and CEO of AFuzion