Turmites and "Langton's automaton"..

From: Chris Langton (cgl@t13.Lanl.GOV)
Date: Thu May 06 1993 - 15:31:05 UTC


  It seems that the fellow who developed turmites and I came up
with a similar automaton independently. It's not surprising,
really, because its an obvious thing to try. I implemented only
one of the possible FSM's mmoving over a 2D tape, while the
turmites system allows a wide variety of FSM's to be defined.
In principle, these really are Turing Machines, because a TM is
just an FSM augmented with an arbitrarily extendable tape. They
are interesting for looking at 2D languages, and, for my purposes,
for looking at collective phenomena, in which a number of simple
agents get caught up in complex dynamics which is mediated by
structures in the environment - the agents can modify the 
environmental structure, while at the same time the same structure
is partially determining the agent's behavior. When many agents
are modifying and being directed by a shared structure, individual
agents can be driven to behave much more complexly than they
could in isolation, given the same amount of internal state -
a property that has certainly been important in the evolution
of social organisms, such as ants, bees, wasps, termites,
and even us. 

My original published account of these automata (which I called
"vants" for virtual ants") was in Physica D, V22 pp 120-149 1986.
I discuss some extensions of this "virtual automaton" approach
in another paper in Complex Ssytems V1 pp 257-271. In the latter
paper, I explore the idea that, since the FSM's moving around
in the CA are themselves just state configurations, FSM's can
write other FSM's on the "tape", which themselves begin to
propagate around interacting with both static and other 
dynamic tape structures (other FSM's). Likewise, such FSM's
can read other FSM's because, again, they are simply state
configurations on the tape (albeit dynamic ones). They can also
modify each other - even erase other FSM's. This all requires that
the "rules" for the FSM's be embedded in the CA rules, rather
than being externally defined, as in the Turmites system.
Nonetheless, I think there are some interesting issues to be
pursued in htese kinds of systems.

At the Santa Fe Institute, we are building what might be considered
a generalization of the Turmites system, in which arbitrarily coded
agents propagate around in an environment with its own state and
dynamics. Agents can interact with each other and with features of
the environment. It is not a CA, although you could simulate
a CA within it. It is more in the spirit of Mitch Resnick's *Logo, 
but more general (agents can be expressed as arbitrary C code, for
instance). We are calling this the SWARM simulation system, for
obvious reasons. It is now in development, not ready for release, but
we should have a beta version within 6 months or so.

Cheers!

Chris Langton <- cgl@santafe.edu


This archive was generated by hypermail 2.1.7 : Tue Oct 14 2003 - 21:44:13 UTC