How Artificial Intelligence Can Improve Pilot Safety

When people hear about Artificial Intelligence (AI) on aircraft, their first thought is usually about an aircraft flying autonomously from A to B, without any human input. For decades, this has been a horror scenario to pilot unions and training organizations alike. However, is rejecting further integration of AI on the flight deck always a good thing?

Next

Breaking Down the Problem

Let's start off this article by saying that no, AI will not replace human pilots on an aircraft anytime soon. Beyond the obvious confidence and redundancy issues, the technology is not yet mature enough. That does not mean AI cannot be used to increase aviation safety though, assisting the crew throughout the flight and providing feedback for improvement. If we break down the process of flight into smaller pieces we can compile a subset of problems where AI excels.

Suggest Options

The most important advantage of modern computers by far is their ability to quickly process large amounts of data. That means that, given an ample amount of correct data, AI can evaluate the consequences of actions in real-time, even before they are performed by the flight crew. An example of this is Airbus' ROPS (Runway Overrun Protection System) which evaluates data like runway conditions, airspeed and winds in real-time and alerts pilots when it seems the aircraft may not be able to stop within the available runway lengthi. Another innovative example is Xavion, an EFIS app by laminar research which uses GPS and wind data to calculate and display optimal glide paths to nearby runways in case of engine failure.

At this point in technology, it is important for this kind of systems to be purely advisory in nature, as opposed to taking autonomous action. A well-designed system can reduce the pilot's workload by helping him take decisions faster and with a higher degree of certainty.


Condition Monitoring

An implementation of AI that has been available in the automotive industry for a while is driver drowsiness detection. They record data from a number of sensors to detect changes in steering behavior over time. These measured changes are combined with a number of other parameters like time of day and trip length to calculate the driver's level of fatigue. When the fatigue level becomes excessive the car will alert the driver in increasingly annoying ways until the driver takes rest.

Systems like this can be readily modified to be used in airplanes. Although an airliner is normally unable to 'pull over for a break' and not all flights carry relief crews, safety benefits can be readily obtained when other crew members are aware of the affected pilot's fatigue. The affected pilot may also be able to take a walk, drink some water/coffee or perform similar actions which increase alertness.

Improvements in the field of machine vision also have come to the point where fatigue recognition may be performed visually instead. This would allow fatigue to be recognized during phases without crew input to the aircraft as well (extended periods using autopilot). Machine vision would also allow for other potential risks in a pilots condition to be detected early, such as stress, loss of concentration, sensory overload.

Identify Risks

When machine vision gets coupled with other data such as speech recognition and information from the aircraft's own sensors, it becomes possible to keep track of a crew's actions over time. This would allow for the aircraft to recognize and alert the crew to potential hazards such as checklist interruptions and missed items, loss of minimums, break-downs in crew coordination or breaches of sterile cockpit. The system could send out aural or visual alerts, alerting the crew to the potentially dangerous condition before they develop further. The alerts can also be used as a trigger for video recording, which can be used during debriefings.


Improve Crew Skills

When advanced systems like the one discussed in the section above get tuned to also observe the behavior of individual crew members, it becomes possible to evaluate a pilot's performance not only as part of a team but also independently. One implementation we, at D&V Flight Crew, are experimenting with is to identify areas for improvement and suggest supplementary training, and this is an area we may further discuss in a later article. When applied over time, it becomes possible to map the strengths and shortcomings of individual crew members. Such information can be used to suggest crew member pairing to create more balanced and optimized flight crews, based on each crew member's skills and abilities.

Challenges

Several limitations still remain before these implementations become widespread on airliner flight decks. An obvious one is a disconnect between the small companies pioneering and leading the development for this kind of technology, and the large aircraft manufacturers and airlines who have to end up using them. While there is definitely room for these small companies to become more vocal about their accomplishments, aircraft manufacturers are going to have to drop their 'not invented here' attitudes and open up their R&D to some of the brilliant minds outside their own companies.

Another role exists for regulatory agencies such as EASA and FAA to be proactive in developing soft and hard law surrounding AI systems on aircraft, and their integration into the aircraft's data bus.

Various legal and ethical questions remain, mostly surrounding privacy issues. Although most pilots won't object to smart aural or optical alerts, not all pilots will be happy if such data ends up getting stored beyond the end of the flight, especially when such records can affect the course of one's career. Airlines, governments and pilot unions will have to work together to develop laws and policies surrounding the use and storage of such data, so pilots can have confidence the artificial intelligence is there to support and protect them, not punish them.

Conclusion

While artificial is still very much in development, the current state of technology already for many benefits to be obtained, also within the aviation industry. While fully autonomous aircraft may not be possible yet, several technologies exist which may be readily implemented into production aircraft to assist pilots and reduce human error.

Also outside the actual aircraft, many safety benefits of artificial intelligence can be obtained. D&V Flight Crew is proud to be a pioneer in the use of artificial intelligence in pilot recruitment and training, providing our customers with better-trained pilots, selected to fit perfectly within their organization. If your organization is looking to grow or reduce costs while maintaining or improving your safety standard, we will be happy to help you find a solution tailored to your personal needs.

About the Author

Daniel is co-CEO at D&V Flight Crew, where he focuses primarily on business development and technology. Technologies he has worked with include artificial intelligence, machine learning and avionics design. He writes about technology and how it can make the aviation industry better and safer.