Robo-drones learn to land by going bug-eyed
Let it bee
Landing a flying object can't be that hard: even bees can manage it. That's why researchers from Sweden, Germany and Australia have looked at what bees could teach about landing strategies for unmanned aircraft.
Of course, if your UAV is a quad-copter with a human operator or good automation, landing isn't so hard: return it to its take-off point, and let it down gently. A horizontal landing on an airstrip is a different matter, and a host of different landing aids have been developed over the years.
According to this study published in PNAS, the aim of the group, which conducted its experiments at the Australian National University's ARC Centre of Excellence in Vision Science, was to identify how bees approach the landing problem, and use that as a basis for a mathematical model for a “universal landing strategy”.
Moreover, they say, their model “does not require knowledge about either the distance to the surface or the speed at which it is approached”.
And it's quite simple, really: bees use purely visual cues. They rely on the patterns of motion of what they see, in particular, the apparent rate of expansion of the surface they're approaching.
If it were you or I sitting in the pointy end of an airliner approaching a runway, we would see the runway expanding in our field of vision, and so long as we stayed dead-centre, the expansion would remain symmetrical as we came closer to our target.
Not only that, but as co-author and professor at the University of Queensland Mandyam Srinivasan explained to The Australian, a constant-speed approach results in the apparent acceleration of the landscape as you get closer. Bees use this to adjust their approach speed.
“If you keep the rate of expansion of the image constant, you automatically slow down and by the time you make contact you’re moving at almost zero speed,” he told The Australian.
It's easy enough to say “well of course, that's how bees would do it” – but the trick is in proving that, and turning it into a mathematical model. Returning to the paper, the group analysed: “the trajectories of honey bees landing on a vertical surface that produces various patterns of motion. We find that landing honey bees control their speed by holding the rate of expansion of the image constant.”
Considering the size of a bee's brain and its low-resolution vision, that's a trick that should be easily applied to robot vehicles, if it could be modelled. To help create their model of bees' landing strategies, the group then observed how the bees responded if their landing field was manipulated while the bees were approaching. They used a spiral pattern to confuse the bees and trick them into crash landing.
Professor Srinivasan's next plan is to try implementing the model in non-insect drones, to attempt landings armed based on the visuals from a simple aircraft-mounted video camera rather than GPS, radar or remote control – a strategy he says would be useful for military applications. ®