In nature, bees can fly from flower to flower and navigate between obstacles using a method known as optical flow- a method for measuring objects' speed within their FOV (field of Vvsion). Roboticists have tried to replicate that method for drones but have had little success over the years until now. Researchers from TU Delft and the Westphalian University of Applied Sciences have created an optical flow-based learning method that allows drones to estimate distances based on the visual appearance (shape, color, texture) of objects in their view.
"Our work on optical flow control started from enthusiasm about the elegant, simple strategies employed by flying insects," states Guido de Croon, professor of Bio-inspired Micro Air Vehicles at TUDelft. "However, developing the control methods to actually implement these strategies in flying robots turned out to be far from trivial. For example, our flying robots would not actually land, but they started to oscillate, continuously going up and down, just above the landing surface."
Optical flow has some drawbacks when it comes to small drones, such as noise, which obscures obstacles in the direction drones are flying, meaning the obstacles the drone will be most likely to hit are the ones that are the hardest to detect. Drones also have issues adapting to height when trying to land, which often results in hovering rather than attempting to land. Being able to identify the attributes of objects within their FOV led the drones to land quickly and without detrimental issues. The researchers state the new optical flow method could enable drones to be used for greenhouse applications, crop monitoring, and inventory trackers.