flocking autonomous agents
A while ago I completed my Mathematics BSc from Bangor University. Its a course they’ve stopped doing now, but at the time, it was a most excellent thing. For those that are interested I did my final year dissertation on Flocking Autonomous Agents. That is, the order that appears when groups of supposedly simple agents congregate.
Its the sort of thing that you see in the wild with swallows flying about. Theres no master plan of where they’re going, nor any central command to the system. So how do these simple creatures manage to form cohesive flocks that can travel, with what seems, singular purpose?
It was just this question I proposed to answer in my dissertation (and I thought I did that rather well) as well as explaining the deceptively complex maths behind it.
Long story short. The system works like this:
- The individual agents are only aware of whats going on in an area directly around them (their vision range).
- The individual agents will try to adjust their direction to that of their neighbors.
- They won’t always do this right.
- There is 4th rule.
It seems almost counter-intuitive that the fact that a agent makes mistakes in their headings would actually help the flock as a whole but, after some rather intense mathematics, you can prove that, up to a point, it helps the system.
The crux of the matter is this – If there is group of agents flying together in sync and none of them ever makes a mistake as to their heading or position, their relative positions will conceivably never change.
If however they do make mistakes something wonderful happens.
Mistakes that have a negative impact on the flock as a whole get naturally canceled out while mistakes that bring the flock together are reinforced. Over time a very tight, cohesive, resilient flock is formed.
I even wrote a lovely java applet for it and ill post it up here when i figure out how to embed it properly. Ill even give you the source code, cos I’m nice like that.
