Thesis project... long winded.

[John Eikenberry [MSAI]]

Hello,

        I've begun to work on my thesis project and I thought I'd bounce
some ideas off the hive. 

        My area of interest is (primarily) automous agents. I find just
about every application/aspect of these little buggers to be incredibly
interesting. Though for my thesis project I'm concentrating on the 
network roamers (info gathering, scheduling, etc.). 

        The network roaming variety has several experimental 
implementaions already, notable example would include the Tocoma project, 
the Penguin perl module and the Agent TCL project. These projects are all 
interesting and hold promise, but they are all pretty much strictly 
concerned with the technical aspects of the agents. 

        I found myself thinking about the possibilities of what it would
be like if this thechnology really took off. That thousands of these
little guys would be wandering around various networks, interacting,
communicating, etc. Seems like it could be considered a complex system 
(humour me ;). 

        Now, if I were going to make a bunch of agents to do things for
me, the first thing I would do is consider what metaphor would I use to
model their behaviours, communications, etc. While there is some work in
this area, it seems to be an area in need of further research. And
(finally) this is what I want to do. My plans are to create a
network-space for swarm, then set about creating a couple of different
metaphoric-species of agents to inhabit it. Then to experiment with the
different species to try and find out their relative strengths and
weaknesses, and whether they hint at different ways of doing things that
may not have been thought of outside that metaphor. And since one of my
other related loves is alife, the metaphors I am currently considering
(and studying their biological counterparts) are wolves and bees. 

        The first stage in this process is to create the network-space. I 
am considering something similar to the way the above mentioned 
implementations do theirs. At each node (computer on the network), there 
would be an agent-space, and (possibly multiple) information space(s).
Now, since the idea of N-dimensional space really exist in a network
of this sort, I will probably just use multiple lists. Though the 
different nodes could be seen as a space of some sort, it wouldn't have
the same properties as normal space (each node would be considered one 
move away from the agents point of view, and they would have very limited 
information about that location).

Well, I guess I'd better stop before I write a small novel here...

I'd love to get feedback on this idea (which was kinda the point ;-).

Thanks,

John Eikenberry
address@hidden
______________________________________________________________
"A society that will trade a little liberty for a little order
 will deserve neither and lose both."
                                         --B. Franklin