langtonAntSpace-1.3.tar.gz

[Pietro Terna]

        Hi All,

        at ftp://ftp.santafe.edu/pub/swarm/users-contrib/anarchy/ you can find
langtonAntSpace-1.3.tar.gz, which upgrades langtonAntSpace.tar.gz

        Below you'll find the content of the README file. Many thanks for 
comments.

        Pietro

README ---
        Pietro Terna (August 29, 1998)

        langtonAntSpace - derived from simpleObserverBug2 - is a 
simple Swarm (1.3 v.) program developing a space where one or 
more Langton's ants can move. The space is a torus; its 
dimensions can be changed via a probe window (in Linux env. must be
a square; in Win32 env. also a rectangular one runs). The same model 
probe window allows us: (i) to modify the probability of finding 
ex ante traces (see below) in the space (put 0 to avoid traces); 
(ii) to establish the number of moving ants (each ant starts in 
the N, E, S, W direction on a random basis); (iii) to define if 
a single position of the space (a square) can contain many ants 
simultaneously (manyAgentsInASquare set to 1) or only one 
(manyAgentsInASquare set to 0).
        The observer window allows us: (i) to choose the display 
frequency (the program proposal is that of having a refresh of 
the display each 100 simulation steps; I'm doing so because I'm 
now working with a win32 Swarm installation and display actions 
are pretty slow); (ii) to modify the zoom factor.
        To admit the presence of more that one object in a space 
position I have substituted the Grid2d class used in 
simpleObserverBug2 with the Sven Thommesen's MoGrid2d class. I 
have introduced light modifications to this class (they are 
noted with a //pt sign).
        The behavior of a Langton's ant is the following: an agent 
moves only in the N, E, S, W directions, making one step in each 
simulation cycle; if the arriving position contains no trace (no 
sign, no pheromone ...) the agent turns 90 degrees to the right 
and leaves a sign in the position; otherwise, it turns 90 
degrees to the left and clears the position.
        I'm posting this file for two reasons: to receive comments 
about the structure that I'm describing below; to communicate a 
"surprise effect" of the simulation, due to a curious emerging 
property. If we run the program with manyAgentsInASquare set to 
0, we observe many reciprocal blocking situations, with agents 
assembling if few places of the space; agents seem to follow, in 
their movement, paths to attractors, written in the traces left 
by themselves of by other agents. The presence of ex ante traces 
- i.e., of noise - does not eliminate this effect, which may be 
is well known. For me it has been surprising.

        langtonAntSpace is a first step application in a project 
regarding the definition of a scheme of  Agents Environment 
Rules (AER) simulation models. Within this scheme I will rewrite 
in Swarm a neural network tool named CT [see, as an example, 
Terna, P.: A Laboratory for Agent Based Computational Economics: 
The Self-development of Consistency in Agents' Behaviour. In 
Conte, R., Hegselmann, R., Terna, P. (eds.): Simulating Social 
Phenomena. Springer, Berlin (1997)]. CT method has the goal of 
developing agents wired with soft rules (mainly self developed 
with leaning), but I'm preparing the AER scheme so that other 
solution could be easily interchanged. At present I'm thinking 
about Classifier Systems and rule based systems.
        In this perspective, the following protocol is adopted: the 
rules describing the behavior of an agent are managed by a 
specialized class of objects, named RuleManager; a ruleManager 
receives all the necessary data from each agent; we can have 
several ruleManager objects; the interaction with the 
environment is allowed only to the agents; the development or 
the modifications of the rules are performed by other 
specialized classes of objects (RuleMaker classes). (At present, 
we have not examples of ruleMaker objects).
        I've also to developed the possibility of writing rules
in script files, using Tcl: See tclTest-1.3.tar.gz



                  ==================================
   Swarm-Support is for discussion of the technical details of the day
   to day usage of Swarm.  For list administration needs (esp.
   [un]subscribing), please send a message to <address@hidden>
   with "help" in the body of the message.
                  ==================================