It seems like it should be relatively simple math to plan the most efficient route for an airplane. But for airlines trying to get thousands of airplanes to destinations around the world, on time and safely, it gets to be an extremely complex problem, especially when a single last-minute change to a flight can create a ripple effect in the routing of dozens of other flights.
GE thinks that the task of planning better flight paths is a task worth tackling, both to save airlines (its customers) money and to save on fuel burned.
With its $500,000 Flight Quest competition, GE has been working with Alaska Airlines to crowdsource solutions to the problem. The goal is to create a system that allows airlines and traffic controllers to make the most efficient flight plans and adjustments in real time. Today, it’s announcing the winners of the second phase of the competition, held through Kaggle, a platform that allows companies to challenge data scientists to solve problems.
“It would be a real far-reaching evolutionary step from where we are today,” says Giovanni Spitale, GE Aviation’s manager of the project.
The Flight Quest challenge asked entrants to come up with the most efficient flight routes, speeds, and altitudes, given a series of real-world constraints, such as expected arrival time, weather, wind, and airspace limitations. Out of 6,800 submissions, the winner, a Spanish statistician Jose A. R. Fonollosa, produced a model that proved to be up to 12% more efficient when compared to data sets from actual flights.
Eventually, the goal is to produce a set of software that would help pilots, airlines, and air traffic controllers make better decisions. If each scheduled flight globally reduced its flying distance by only 10 miles, the industry would save $3 billion a year and 380 million gallons of greenhouse-gas spewing jet fuel, GE calculates.
GE is also working with the model produced during the first phase of the Flight Quest competition, which involved making better predictions of actual flight arrival times. The winner of that contest predicted times 40% more accurately than existing programs, and now GE is working on prototype software so airlines can reduce gate congestion and give flyers better ideas of when they’ll get home.
Overall, GE wants to use data-crunching to allow airlines to optimize for different factors, whether that is fuel burned or on-time arrivals. “Our hope is to provide … a one-stop shopping center for algorithms that allow an airline to govern their fleet performance according to their business needs,” says Spitale.