Remember meForgot password?
    Log in with Twitter

article imageSoaring like birds, AI-equipped glider learns to use air currents

By Karen Graham     Sep 22, 2018 in Technology
Poway - Large birds like the albatross or eagles seem to stay aloft for hours, not always flapping their wings - but using air currents. Scientists are now using artificial intelligence to learn their tricks, hoping they can teach aircraft to do the same.
When birds are riding air currents, very little energy is expended. This is why the albatross can sometimes travel for 10,000 miles at a time, But exactly how birds are able to ride air currents is not understood perfectly - but some scientists are finding out.
A group of researchers at universities in the US and Italy has reported in the journal Nature, published on September 19, that they were able to design an artificial intelligence (AI) system that has learned to take advantage of a particular type of air current.
“Soaring birds often rely on ascending thermal plumes (thermals) in the atmosphere as they search for prey or migrate across large distances,” according to the study. The authors then go on to explain how they were able to get an AI-equipped glider to do the same thing as a soaring bird.
This is not the first time artificial intelligence has been used to train an algorithm to control a glider to navigate thermals. Microsoft published similar work with gliders last year. One big leap forward is that this research used data from actual flights to update and improve AI’s performance in the field.
Actually, it is the very same algorithm used by Google’s infamous AlphaGo that learned to play the board game Go all by itself and then beat professional players. This could not have been possible with conventional programming techniques.
The researchers at the University of California, San Diego, and the Abdus Salam International Center for Theoretical Physics, Trieste, Italy, first trained the machine learning algorithm in a simulator. Then they took it all outdoors, performing almost 240 separate flights, staying aloft on average, about 3 minutes.
Graphic depicts: (a) A trajectory of the glider soaring in Poway  California; (b) a cartoon of the g...
Graphic depicts: (a) A trajectory of the glider soaring in Poway, California; (b) a cartoon of the glider showing the vertical wind currents and torque experienced by the glider; (c) the vertical component of the wind velocity (blue) and the vertical wind accelerations (red) experienced by the glider during a typical flight session; (d) the bank angle of the glider during the same flight session as (c) and the corresponding torque experienced by the glider.
Gautam Reddy
However, some of the flights lasted 45 minutes, soaring up to fly at the same level of eagles. The glider even got attacked by an eagle during one particular flight, according to ZME Science.
Of course, there is still more work to be done. Thermals are just one of the many types of air currents our bird friends have to contend with. But even so, the work is impressive and a definite first step. It also shows how reinforced learning works.
More about gliders, Artificial intelligence, machine learning, thermals, algorithms
Latest News
Top News