In pursuit of fleet-footed prey, the northern goshawk wings through thick forest canopies and underbrush at breakneck speeds, dipping and diving to avoid colliding with trees or other obstacles. But it can only go so fast, apparently obeying an unspoken speed limit dictated not by biology, but by the density of its environment — beyond a certain threshold, it is certain to crash into something. This is an important lesson for makers of drones and other flying objects, according to researchers at MIT and Harvard.
Most drones fly relatively slowly, especially at lower altitudes where they might encounter obstacles and require plenty of time to react. Biologists at Harvard and roboticists at MIT have been studying flight behaviors in goshawks and other birds, aiming to improve algorithms that would allow unmanned aerial vehicles to cruise more quickly through forests, urban areas or other cluttered landscapes.
Goshawks don't necessarily see everything ahead on their path, so they must judge the density of the forest and assume they'll find an opening. In an interview with MIT News, aeronautics professor Emilio Frazzoli aptly compares it to backcountry skiing. You don't always know where the next tree stands, but you cruise downhill anyway and assume you'll be able to navigate around it when the time comes. Beyond a certain speed, though, you might not have time to stop or turn before hitting the as-yet-unknown tree. So (if you're smart) you obey an innate, self-imposed, environment-dictated speed limit. Programming this into a robot is difficult, however.
Frazzoli and some grad students devised a differential equation expressing all the possible positions of a bird at a given location at a given speed. Then they developed a model of a forest, using statistical distribution models used by ecologists. Then the team calculated the probability that a bird would hit a tree while flying at a certain speed. They figured out that for any given density of trees (or other obstacles of choice), there exists a speed above which there is no "infinite collision-free trajectory," as MIT News explains. The bird will surely crash, because there's no way for it to avoid the obstacles. But below that threshold, things should be fine.
"If I fly slower than that critical speed, then there is a fair possibility that I will actually be able to fly forever, always avoiding the trees," Frazzoli said.
This theoretical speed limit equation could be extrapolated to any obstacle-filled environment — an actual forest, a city with tall buildings, and so on. So a drone could fly forever unimpeded, so long as a drone obeys its own speed limit.
to bad us silly humans don't obey the speed limit AND are to busy looking at our cells to wonder if we will crash or not eh?
*growls*If you troll or flare I WILL MAUL YOU!*growls*
More like too busy looking at our wallets!
It will only be a matter of time before they can speed drones up to ultra fast, just takes faster smarter AI and the landscape data known in advanced, in detail. This can be achieved to some resolution today with satellites.
Are you kidding, the last drone had like 3.5 million lines of code!
And I say too busy texting. (though I don't and can't, because I don't have a phone that texts)
There are three limiting factors - perception, cognition, and movement.
Increasing perception is great, and theoretically the easiest in robotics, so long as there are not changing variables. A drone with a 3D map of New York would have perfect perception - so long as it was not Thanksgiving and a giant Kermit balloon was there that day.
Cognition is the hardest for biological units (desciding what to do, which way to go) - but is easier for a computer which is willing to make choices without fearing potential consequences. However, each cognition resets the perception needs - like your GPS recalculating a route when you make an unrouted turn.
Movement, however, is a function of physics, an the faster something moves, the more difficult it is to turn (when relying on air-friction for turns). Thus, there is a speed at which turning requires not just drag, but energy - any non of our high speed craft (outside of space craft) work with that in mind.
Wow, it took a study at MIT to tell us that moving faster when you have to watch out where you're going is bad for you?
It's one thing to zip by at 200+ MPH on the Autobahn, it's another to try and do the same in a neighborhood street with cars, children, potholes and other things that you need to navigate through.
Wow! Speeding is dangerous. Thanks MIT!
How difficult is it to code a speed limit based on obstacle density? A 5 year old programer could do that.
Did they even ask a programmer for help?
@killerT "How difficult is it to code a speed limit based on obstacle density? A 5 year old programer could do that.
Did they even ask a programmer for help?"
The difficult part was coming up with equation that describes what the ideal speed is for a given object density, based on the drone’s performance.
@Oakspar77777 "Movement, however, is a function of physics, an the faster something moves, the more difficult it is to turn (when relying on air-friction for turns). Thus, there is a speed at which turning requires not just drag, but energy - any non of our high speed craft (outside of space craft) work with that in mind."
Lift causes an aircraft to turn, not drag (except for yaw control on the B-2). Drag is a byproduct of lift. I think you are trying to say: If the drone is going fast it will have to turn hard to avoid objects, and thus cause the drone to use more fuel.