Follow Datanami:
January 16, 2019

Algorithms Aim to Unclog the Skies

via Shutterstock

Researchers are churning out new algorithms initially designed to improve situational awareness in the sky while also helping pilots avoid dangerous weather and costly flight delays.

Knowing where you are on the clouds in relation to other aircraft is a fundamental flight safety requirement, and the stormy skies are not getting any less crowded. One data analytics effort is using deep neural networks and “memory networks” to improve air traffic controllers’ situational awareness.

A second seeks to develop an algorithm that could help pilots avoid dangerous storms while keeping flights on schedule. The latter project is part of European research effort at the University Carlos III of Madrid called TBO-Met, an abbreviation of the full project name “Meteorological Uncertainty Management for Trajectory Based Operations.”

The situational awareness effort at the Australian university QUT is developing an algorithm that can predict trajectories—initially aircraft but eventually any moving object. “In essence, it’s built to measure a trajectory in and predict a trajectory out,” said Prof. Clinton Fookes, who oversees vision and signal processing research in QUT’s science and engineering branch.

The goal is to determine the trajectory of a target object as well as nearby objects “to create awareness of what’s around the target,” Fookes said.

The algorithm draws on memory networks used to store trajectory data for a given location, essentially emulating how human memory works. “Those two sets of data are then analyzed by another sub-network that determines where the target will go next,” he added.

The situational awareness algorithm was trained using a variety of data sets, including air traffic control patterns and—for different applications—camera data from pedestrian traffic. Airport data included severe weather, allowing researchers to “test how well our algorithm coped in such a dynamic situation,” Fookes said.

The researchers said predictions generated by the algorithm were accurate since it was able to factor for how pilots reacted to similar conditions, then predict what an individual pilot’s next move.

Along with managing civilian airspace, the algorithm might also be used to manage drones or to identify the most cost-efficient flight paths.

Meanwhile, Spanish researchers said their algorithm would be applied to aircraft flight plans to keep air traffic flowing in the event of heavy weather. Those weather delays are costly. The researchers estimated that as many as 30 percent of air traffic delays in Europe are weather related, generating annual losses of as much as €200 million (nearly $228 million).

By factoring for weather uncertainties such as hail and lightning, the weather algorithm would help planners determine how many aircraft can safely fly through a given sector. The results would allow carriers to determine the best flight plans given weather conditions at different locations along a flight path.

Extreme weather was estimated to result in 2.1 million minutes of flight delays over Europe in 2017.

Recent items:

IBM’s New Weather Forecasting System Runs on GPUs

Machine Learning Model Tracks U.S. Spy Planes