A couple of years ago, I was riding on a train through the German countryside with my family. Out of the fields next to us rose a flock of what must have been several thousand birds, who then executed some of the most beautiful aerial maneuvers I have ever seen, shaping and reshaping a series of organic, flowing patterns against the grey background of clouds. I felt privileged to have been a witness.
So, imagine my delight when the latest issue of Physics Today showed up in my mailbox with the following cover:
There’s a lovely two page article on the STARFLAG project, which is modeling the 3-dimensional flocking behavior of starlings outside of Rome, using 3-d stereograms. The pictures are just spectacular, and give some sense of why I was mesmerized on that German train:
They’re coming up with interesting results about how the birds must be making individual decisions within the flock. They get the best match to the observed behavior by assuming that the birds track their 6-7 nearest neighbors, regardless of local density. (This idea is familiar to astronomers, who frequently use N-th nearest neighbor statistics when estimating density, rather than averaging over a fixed distance.) The distribution of nearest neighbors also seems to be anisotropic, with a tendency to find the nearest neighbors to the sides, where the birds eyes are. Biologists have reached similar conclusions based on models of schooling behavior in fish. However, the STARFLAG project is being led by theoretical physicists at the National Institute for the Physics of Condensed Matter in Italy, so you can imagine that the methodologies are a bit different from what biologists would naturally pursue.
Update: I’m daring to embed a video:
Update on the Update: There’s a longer, informative video here with somewhat better footage. It really picks up in the second half.
October 10th, 2007 at 12:55 pm
Unreal! Thanks for sharing it.
October 10th, 2007 at 3:21 pm
I’ve read somewhere that birds tend to flock in “V” formations – which is the optimal one where everyone sees the leader. Some of the pics you posted show a sort of “V” formation, but others don’t.
October 10th, 2007 at 3:29 pm
I think that the birds in the V-formation are drafting, much like you do in bike racing. The head bird at the apex of the V does substantially more work than the birds further down. Thus, you see this in migratory birds who are flying for the long haul. Canada Geese are probably the easiest to spot doing this around the northern US, but I’ve seen ducks at it as well. Flocking is different, and seems more like a behavior to cope with predators.
October 10th, 2007 at 3:56 pm
An almost related couple of questions Julianne:
I’ve often wondered would take in, say, a herd of wildebeest for the technique of stampeding the attacking predators to death rather than running away en masse (and hoping it’s not you that gets picked off) to emerge? Surely it would be viable through sheer weight of numbers!
Can they not overcome individual fear of the predators to act assertively?
Would an absence of predators be catastrophic in the end (eventually too many wildebeest for
land to sustain?)?
I suppose as it is it’s not a bad evolutionary strategy for the wildebeests and the predators.
October 10th, 2007 at 5:15 pm
Haludza, they do (to an extent):
http://www.youtube.com/watch?v=LU8DDYz68kM
That was buffalo, but you get the idea.
On topic, though, I’ve always loved watching flocks of birds here in Northern California. It’s really an amazing sight.
October 10th, 2007 at 6:26 pm
I got into flocking algorithms while working on the visual effects for that horrid film “The Day After Tomorrow.” It’s interesting looking back at how people investigating flocking often thought there was some central control or hive mind driving it. Craig Reynolds really opened the door to understanding what drives flocking in his 1987 SIGGRAPH paper. The concept of emergent behavior and how three simple rules (follow leader, steer to avoid, and keep distance) can result in beautifully complex patterns. This applies so much through out nature where ever we see simple iterative patterns. That was a hard one to explain to my evangelical ex-girlfriend
October 10th, 2007 at 6:33 pm
I was living with a bunch of Thinking Machines and MIT Media Lab folks at the time, and tagged along to the big SIGGRAPH video presentation right after Reynold’s paper. I remember flocking-CG-video after flocking-CG-video. That, plus someone had figured out how to make realistic jello.
October 10th, 2007 at 6:50 pm
Funnily enough, there was a wonderful film of flocking starlings on a new programme on the BBC tonight – the Nature of Britain. You can see a video of it at http://www.bbc.co.uk/earth/nature/uk/geolocatable/records/classified/derived~natureofbritain+video – click on the map.
I’ve watched starlings flock before but I had no idea they could make these dazzling patterns. They must be enjoying it surely.
October 10th, 2007 at 7:38 pm
Cool, love watching starlings (and other birds) flock
I wonder if they are feeding before bedtime, or just dancing around.
You’ve probably seen Vietnam Traffic
and a bird’s eye view of Traffic in Ho Chi Min City
October 11th, 2007 at 11:26 am
Arrgh! I’ve been scooped!
Six years ago ony my “finding myself” trip through Europe, I exited Rome’s central train station just at dusk to see _thousands_ of birds doing the most incredible coordinated maneuvers. I came back every night for a week to watch, jotted down some ideas on how to simulate the behavior using simple agents, and when I went home I tried to learn some graphics programming, but never got far with it. Then I went back to grad school and got busy with “regular” stuff and never came back to it.
I’m glad somebody is working on it, and it looks like they were inspired by the same view from the train station.
October 11th, 2007 at 11:27 am
Methinks it is like a weasel.
October 11th, 2007 at 11:53 am
[...] out this post on Cosmic Variance about a group of physicists who study flocking behavior. The video is [...]
October 11th, 2007 at 9:32 pm
Thanks for posting these wonderful pictures! Stephen Smale and other mathematicians have also been studying this type of behavior.
I live in an area where there are a lot of gnats during the summer, and while you would describe their behavior as more like swarming than flocking, they also exhibit remarkable group behavior. They often swarm in well defined globs in the air — usually at a height ideal for inhaling into your sinuses or lungs
— and the whole group will move up and down and horizontally almost as if they are of one mind.
This summer I’ve occasionally been encountering a family of three very large cranes, and am always amazed how fly in a tight perfectly synchronized unit.
October 12th, 2007 at 10:02 am
Just saw the video…WOW!!!
I wonder if this analysis can also be applied to bees? It makes you think where the arrow mark or question mark they come up with in cartoons can also be done in reality.
October 13th, 2007 at 3:43 am
Don’t fish do the same thing?
October 20th, 2007 at 10:39 pm
I’m glad to see that Brando mentioned Craig Reynolds who actually won an Oscar for his technical work on rendering flocks and schools using simple localized rules. He is the only person I’ve ever known to win an Oscar, and it is a testament to his drive and focus. When he arrived at the MIT Architecture Machine, the precursor the Media Lab, he wanted to model high school bands. Everyone in the band followed the leader and used a few simple rules to create all sorts of patterns even as they produced all sorts of music. That was his goal. Of course, in the early 1970s he had to do a bit of tool building. There were no 3D rendering packages written, so he wrote his own. There were no 3D modeling packages, so he wrote his own.
He moved out to Hollywood which is a nasty place to work, and he kept his eye on his goal. He was less interested in marching bands and more interested in the way flocks of birds, herds of animals and schools of fish moved. If you could render one bird, you could render hundreds. Sutherland demonstrated back in the 1960s with his hens and chicks. The problem was scripting them. That’s where Craig’s work came in, and he did a magnificent job. You’ve probably seen his algorithms in action in countless feature films and shorter works.
Interestingly, he was not initially inspired by great swarms of pigeons in Rome, herds of wildebeest in Africa, or schools of wrasse in the Coral Sea. He was inspired by the mechanism behind the good old fashioned American precision marching band.
November 5th, 2007 at 5:57 pm
[...] (Via Cosmic Variance.) [...]