“Precisely because things went wrong with me, I discovered how nature deals with this problem”

“My research on extremely small flying robots is difficult. There is little room for large sensors and computing power in our drones, the smallest weighs only 3 grams. This scale makes them very useful, for example, for “search and rescue” operations in collapsed buildings or for pollinating flowers. So we try to draw inspiration from nature.

“Some time ago I was working for the European Space Agency (ESA) on landing means for these mini drones. For this, I drew on the biological literature. Biologists have claimed that when landing, bees use movements in their field of vision, called “optical flow”. When you look outside the train, you can see the closer trees moving faster than the mountains in the distance. Trees have a higher optical current than mountains.

“Because I could easily apply this principle to my mini drones, I quickly organized a demonstration during a major internal symposium at Esa. Even Esa’s big boss was present.

‘During my demonstration, the flying robot flew cleanly, everyone started clapping. But near the ground, the drone started rocking up and down, preventing it from landing. The whole audience laughed. It was a big failure.

“After this disappointment, I started thinking about this phenomenon. Other researchers have always had good results with this technique. So why didn’t I succeed? Maybe it was the air bouncing around as the drone approaches the ground or maybe it was the image processing.

“Even during my Christmas holidays in France, I kept looking for a solution. I studied the landing simulations with the minidrones. It turned out that the same problem occurred in the simulations. So it was neither the wind nor the camera.

“I again immersed myself in literature. That’s how I found out that failover still happens. Even pilots suffer if they have to fly without all kinds of heavy equipment. Another disappointment, apparently it was an unsolvable problem.

‘Then I thought: maybe we could use the fluctuations. If the drone detects these movements, it knows that the ground is near. According to the literature, the bees do not see the distance, but they all start swinging exactly 7 centimeters from the flower. It’s kind of a constant flight correction, am I going too hard, am I going too soft before I land? My finding was that this sway is crucial for animals to see distance and land correctly.

“The following years, I worked a lot with biologists. These biological experiments have proven to be much more difficult than the robot research. For example, an Asian wasp had eaten all of our bees during an experiment. Yet we have much more information about how bees land. This eventually led to publication in the renowned trade journal Intelligence of natural machines.

“Other researchers have tried for years to deny this problem, for example by putting long legs under their drone or landing very slowly. But precisely because things went wrong with me, I discovered how nature actually deals with this problem and my drones can also land better.