Machine-learning algorithms are being used to assist long-haul commercial flights that push the limits of long-range jets from missing their destinations.
Air New Zealand has occasionally had trouble operating its Boeing 787s to go nonstop from New York to Auckland. In the meantime, Qantas Airways is upgrading its Airbus A350s with additional fuel tanks in preparation for non-stop flights from Sydney to New York and London beginning in late 2025. The 20-hour flights are anticipated to be the longest passenger services in history.
Both airlines rely on data-hungry technologies to calculate flight paths that are fuel-efficient and steer clear of unforeseen and embarrassing refueling stops. The route-planning software can advise pilots to fly slower to use less kerosene, avoid bad weather, and catch a tailwind—anything to get more miles out of the tanks. The mapping software is made to improve with use, much like a mobile search engine that learns as you use it.
Aviation’s decades-old manual processes are being affected by artificial intelligence, which has an effect on everything from ticket sales to cockpit protocols. Although route planning is nothing new, as ultra-long flights increase and the enormous job of achieving net zero emissions by 2050 looms, AI offers carriers new ways to save money and fuel.
Every day, Flightkeys generates roughly 300,000 flight plans for clients like Air New Zealand, Southwest Airlines, and the American Airlines Group. The following is an edited interview with Vienna-based Flightkeys’ co-founder and head of innovation, Raimund Zopp. Former Austrian Airlines pilot Zopp is 67 years old.
Few people fully comprehend the process of flight planning. Everyone may believe that you enter your destination into the aircraft’s computer, just like you would in a car, and the computer will generate a path. That is not true. Finding the best path is extremely difficult when the aircraft’s systems are incapable of doing it. A system on the ground is required to gather a lot of data and then look for the least expensive answer. Machine learning must be used to properly apply the numerous limitations and parameters that must be taken into account.
There are several levels of freedom but also many limitations due to air traffic control, the military, the weather, and overflight fees that vary depending on the country. A minimum-cost trajectory can be quite difficult to find.
The features of the airplane are one of the important factors. The higher an airplane can climb, the lighter it is. A lighter aircraft often flies a little slower since the ideal speed decreases when weight is reduced. And of course there is the wind and altitude temperature. Avoid headwinds and make gains if there is a chance of a tailwind. To take advantage of the winds, you deviate from the shortest route.
Since we’re often at the aircraft’s maximum range, a flight planning system is more crucial the longer the travel. There is therefore always a compromise between the amount of freight we can carry aboard the airplane without compromising performance and the amount we must dump. On the one hand, you want to conserve fuel, but on the other, you also want to have enough to have a good chance of making it non-stop. A system that calculates quickly and accurately is therefore necessary because the result needs to be changed frequently.
The fact that sustainable aviation fuel is significantly more expensive has the most effect on the planning process. We fly a little bit slower to conserve gasoline, thus the planning is modified. The flight usually takes off a little bit quicker when gasoline is inexpensive since it makes little difference. Because you want to be extremely careful with this expensive resource, you may anticipate that SAF will cause aircraft to slow down.
