Scholarly work on multi-ant situations

JoshCarp's picture
Projects: 
Here's a link to a paper written by a few researchers at the 'Logic Systems Laboratory'. The authors describe the behavior of several 'interesting' rulesets of >2-state ants, and then discuss systems with >1 ants. link [pdf]

Comments

PaulGrobstein's picture

Yep, a useful version at http://ccl.northwestern.edu/netlogo/models/Vants. And is, as reported, the case that simply reversing direction will cause a single ant to clean up. So, two ants usually (always?) create worlds in which each encounters another's path in reverse? And usually (always?) do so at the same time so they end up in inverted starting positions? I think there HAS to be a why to those questions. They can't just be a Chaitin "that's the way it is" (can they?).
JoshCarp's picture

I think we ought to be cautious in trying to characterize universal (or even modal) behavior of multiple ant systems. To quote from the paper I linked to in the parent post: We have empirically been able to show that under some well-specified conditions these ant collections present complicated cyclic behaviors. It has also been shown that these regular patterns are very fragile as they disappear, giving way to unstructured motifs, if some parameters of the situation are slightly modified...Even a small change in the individual's behavior may cause large social effects at the level of the collectivity. [pg. 6, italics mine] People brought up reversibility of multi-ant systems in today's (Monday's) class. I'm not sure if you're referring to this property or to some more general periodic property of a system, but both are interesting--and, according to the paper, relatively rare. The authors report that the majority of single- and multi-ant systems degenerate into 'chaotic' patterns, with no coherent structure emerging after millions of iterations. However, there's nothing to say that chaotic systems can't be reversible, or can't exhibit some other properties of interest aside from road-building. I think that trying to enumerate 'interesting' properties that systems like these can have might be a useful way to eludicate what looks, right now, to be an unlawful, irreducible, mucky bog.