nested swarms

[glen e. p. ropella]

Scott recently sent a message to address@hidden but didn't
include his return e-mail address.  So, in the hopes that 
someone knows him or that this will eventually get back to him,
I'm forwarding my bounced attempt to answer his question to the
list.

glen

------- start of forwarded message (RFC 934 encapsulation) -------
[... header detritus deleted....]
Scott, you should set your mailer to have a valid
email address in either the From: field or in the 
Reply-to field instead of address@hidden

> Hello, I am trying to put together a swarm simulation where
> multiple layers of agents interact.  Interaction can be within
> each layer or through inter-layer connections.  I am fairly new
> to swarm and I am trying to set this up so that each layer is a
> swarm (consisting of a list of agents, a world grid and a
> message grid) and their is a master swarm that does some
> initialization and keeps track of the individual layers.  The
> function of the master swarm is to make it easy to add more
> layers and schedule the layers to concurrently process
> information (not serial layer by layer).  The observer swarm is
> attached to the master swarm but I would like to be able to view
> specified individual layers activity.  I am have trouble with
> the buildactions methods of each swarm because I don't have a
> feel for how nesting these functions is handled in the activity
> source code.  Is their an example where one swarm gives rise to
> multiple swarms that do not directly have an observer swarm?  I
> would be thankful for any help/references that you could offer.
> Thanks Scott

The best example for you to look at is Chris' tutorial, 
esp. the simpleExperBug.  Here, an ExperimentSwarm instantiates
several modelSwarms and runs them serially.  If you're wanting
to start up several Swarms and run them concurrently, then it
will be done a little differently. 

This is a fundamental issue, updating your "space".  If any of
the elements of one layer are capable of interacting with any
of the elements in another layer, then the update can be done
concurrently, but, it will have to be based on the states at
the last cycle, and conflicts will have to be resolved (for
instance if two agents on one layer affect the state of a 
single agent on another layer).

Also, if the layers are completely transparent (any agent 
on one layer can affect any agent on another layer), then
I'm not sure the layers should be separate Swarms.  Ideally,
a Swarm should function as an object in its own right, which 
should involve some opacity.

glen
------- end -------