Re: agents of agents

[Paul E Johnson]

Marcello Gallucci wrote:
> 
> Hello,
> it would sound strange, but I'm trying to build a world of agents composed by
> agents. That is , a hierachical model where each agent (SuperAgent) is
> represented by a given number of agents (Units).
This is not unusual at all. ITs kinda neat, once you get it worked out. 
I did a few little programs concerning the simultaneous scheduling of
separate swarms that exist within larger swarms.  in
http://lark.cc.ukans.edu/~pauljohn/Swarm/WorkingExampleCode/objc you
should find programs "pjrepeater" that show variations.  At a most
advanced level, it requires some fairly subtle things and I hope they
help.

Concerning the structure of your interactions, you can control that in
either of two ways. First, you can regulate where your agents look and
whether they interact with what they find there.  Or you can specially
design a new space object to hold that. I've been working on a Grid 2d
that can hold many objects in one cell, a generalization/extension of
the MOGrid2d that Sven T worked out a while ago.  It is pretty easy to
look at the Grid2d code and override its methods and add your own in
your own class.  I don't think you need to do that because, in Swarm, at
least it is not the usual thing to have the space find objects for your
agents to interact with. Instead, the agents ask the space if the
conditions at a cell are such and such, or if an agent of type so and so
is ther.

If your object is just to limit communication to agents within one of
your groups, the easiest thing to do is to create a list/container of
all agents inside that group, and give all agents access to that list.  

I think your project sounds neat and hope it works out nicely.
> 
> I'm  following  two alternative strategies, each with troubles somewhere.
> 
> 1) first, the simple way: a very simple swarm implementation (say like
> Heatbugs): An agent type, a modelSwarm, an observerSwarm. The only difference
> with a HeatBugs-like model is that the agents (units) belongs to different
> lists, each list being a "virtual" superAgent. 

-- 
Paul E. Johnson                       email: address@hidden
Dept. of Political Science            http://lark.cc.ukans.edu/~pauljohn
University of Kansas                  Office: (785) 864-9086
Lawrence, Kansas 66045                FAX: (785) 864-5700


                  ==================================
   Swarm-Modelling is for discussion of Simulation and Modelling techniques
   esp. using Swarm.  For list administration needs (esp. [un]subscribing),
   please send a message to <address@hidden> with "help" in the
   body of the message.
                  ==================================