Event-Oriented Computing and Reflection

[Chris Landauer]

hi, all -

i was interested in what glen wrote, since we have been designing systems with
reflective properties, and also experiments in cooperating agents, for some
time now

our framework is an integration infrastructure called "wrappings", that
includes explicit machine-processible information about every computational
resource in a system, together with processes that read and interpret that
information to collect resources and apply them to problems - these processes
are also resources and are also wrapped (so the system has a complete model of
its own behavior that it can monitor and analyze) - we think that interesting
agents can be made with this architecture (some papers are available)

also in this case, if the agents are also reflective, then they can access all
of their own behavior, and as much of the environment contents and behavior as
is made available by the server

in a different application area, we are using Virtual Worlds in the form of
MUDs as an experimental testbed, since the entry requirements are very small,
and the VWs themselves can have some fairly flexible semantics (simple
"softbot" vehicles can be found and adapted to carry computational processes
and thereby make agents called "characters", which interact using exactly the
same protocols that the humans do; they communicate through "interaction
items" that are redistributed by the MUD server using location-limited
multicast) - these MUDs also already have some well-defined information
gathering primitives, that can be used by any character to learn about its
environment - in this environment, the notion of tossing agents into the pool
is quite easily implemented

this places the MUDs (TinyMUD, MOO, ...) on a spectrum that is less about time
management than Swarm or most simulation support systems (or even the closer
to real-time DGD-based LPmuds), but still possibly interesting in this
context, since they can easily handle hundreds or even thousands of agents
(depending on your network bandwidth to the server, of course)

more later,
cal

Dr. Christopher Landauer
Aerospace Integration Science Center
The Aerospace Corporation, Mail Stop M6/214
P.O.Box 92957
Los Angeles, California 90009-2957, USA
e-mail: address@hidden
Phone: +1 (310) 336-1361





                  ==================================
   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.
                  ==================================