Slashdot Mirror
The Rules of the Swarm
Hugh Pickens writes "Researchers are starting to discover the simple rules that allow swarms of thousands of relatively simple animals to form a collective brain able to make decisions and move like a single organism. To get a sense of swarms, Dr. Iain Couzin, a mathematical biologist at the Collective Animal Behaviour Laboratory at Princeton University, builds computer models of virtual swarms with thousands of individual agents that he can program to follow a few simple rules. Among the findings are that swarm behavior has patterns common to many different species, that just as liquid water can suddenly begin to boil, swarm behavior can also change abruptly in character, and that just a few leaders can guide a swarm effectively by creating a bias in the swarm's movement that steers it in a particular direction. The rules of the swarm may also apply to the cells inside our bodies and researchers are working with cancer biologists to discover the rules by which cancer cells work together to build tumors or migrate through tissues. Even brain cells may follow the same rules for collective behavior seen in locusts or fish. "How does your brain take this information and come to a collective decision about what you're seeing?" Dr. Couzin says. The answer, he suspects, may lie in our inner swarm."
We are already "cells" of a thing called "economy", for an alien the earth may just look like a single living being.
When I read the summary my immediate response was that this was old news. It has been known for quite a while that cellular automata with simple rules can form universal Turing machines. That means that they can "form a collective brain" and "make decisions and move like a single organism". The interesting question then becomes; What "program" is your machine running, and how do you make it robust so that it works in the real world (with all the noise of nature)?
The article is a popular science article, but addresses this, more interesting, question much more than the summary. They discuss some of the rules involved in specific situations (ants), and even look at "human swarms" (although that bit is a little cheesy). There is no general theory posited about how to make these rule sets though, apart from trial and error (in simulation if you can). They say that the researchers are starting to see patterns, but don't talk about what those patterns are - pity really, as that would have been very interesting.
There are several very interesting optimization algorithms based on swarm behavior, such as particle swarm optimization and ant colony optimization. These methods have a similar ability for non-linear optimization (and pattern recognition) as neural networks.
- Demosthenes
cynicsreport.com
Cool to see Couzin on slashdot... I coincidentally saw a talk of his last week and gave him a brief lab tour. His own research is somewhat outside my area, but one of the most surprising things I recall from his talk is that marching locust swarms are apparently propelled by cannibalistic behavior. If I'm remembering correctly, baby locusts (before they've grown wings) in a region will feed in a pretty disorderly fashion. However, once salt and protein supplies start running low, they get hungry and start trying to eat each other. The researchers realized this when the locusts in their enclosure seemed to be mysteriously disappearing at a steady rate, due to being consumed by their peers. ;)
In any case, once they start eating each other, the locusts start trying to chase the locusts in front of them, while simultaneously avoiding the locusts behind them trying to eat them. The emergent behavior is that the entire swarm moves as a mass until a new area is found where salt and protein supplies are plentiful enough to cause them to switch out of cannibalism-mode. This presumably has a number of ramification on how to control migration of locust swarms, which are an immense destroyer of food resources in the developing world.
and that just a few leaders can guide a swarm effectively by creating a bias in the swarm's movement that steers it in a particular direction.
In human populations, we call those marketers.