Researchers have been taking flying lessons from birds, copying their aerial techniques to teach intelligent drones how to cover long distances without needing a lot of energy.
Migratory birds can cover thousands of kilometres between summer and winter nesting locations, often saving energy by hitching a free ride on warm rising air currents.
While these currents can be very turbulent, the birds have developed methods for exploiting the rising air without being overcome by its choppiness.
Researchers from the Salk Institute for Biological Studies at La Jolla, California, the US, are teaching these two tricks - thermal soaring and surviving turbulence - to long-distance autonomous gliders.
The researchers’ goal is to develop a glider that can cover great distances while requiring very little energy.
Method
To this end, Dr Terrence Sejnowski and his colleagues developed numerical models of atmospheric flows in a rising thermal and combined them with reinforcement learning algorithms for a glider.
The glider then learned by repeated exposure how to move through the thermal, gaining altitude for easy sailing and reducing risks to itself if it got too rough or the thermal weakened.
It learned from two cues - how fast the thermal was rising and how much the turbulence twisted the aircraft.
The researchers showed the glider could stay within the thermal where the lift was strongest but still survive the rough conditions.
The glider also learned how to stay out of trouble by following “risk-averse strategies”, the researchers report in the Proceedings of the National Academy of Sciences, published on Monday evening.
The authors say the results of the research could be used to develop autonomous gliders that can travel long distances with minimal energy consumption.
