Alex (et al),
This is exactly my point... in a multi-agent system of autonomous
intelligent agents, from both a design and implementation perspective,
knowledge and rule partitioning (e.g. via modules) is highly desirable
(if not essential) in order to maintain autonomy. However, the primary
difficulty, I think, is not in maintaing the autonomy of the agents, but
in providing the 'lightweight' communication of facts between agents.
Adding modules to Jess would be one approach, but as others have pointed
out presents issues of efficiency, not to mention being shadowed by
other tasks of higher priority that already burden our friend Ernest. It
does seem, however, that there are a number of people that have this
same problem and have had to come up with their own communication /
shared memory mechanism (e.g. blackboards, tuple spaces, etc.) It would
be great if some of these approaches could be considered as 'user group
supported' add-on package(s) to Jess?!? Or perhaps at least links to
info on these subjects could be added to the "Jess Web Links" page?!?
(if those individuals have any web viewable info, that is).
Regards,
Alan Davis
CAD Research Center / CDM Technologies
Cal Poly State University
San Luis Obispo, CA
Alex George Bejan wrote:
>
> Al,
> Assuming you have an autonomous intelligent agent architecture, and that
> Jess provides the reasoning function in that system, I believe those agents
> should own both a Rete engine and a knowledge module. So modules in Jess
> would be useful in my opinion. Even more important though for such
> architectures would be the ability to run many Jess engines in the same JVM.
> Encapsulation for agents is a lot more demanding than encapsulation for
> objects - they require more independence of execution and decision making.
> As far as the execution is concerned, it's practically impossible to run one
> agent per JVM, and maintaining autonomy for multiple agents sharing the same
> process is complex, but possible. However it would be silly to have one
> (or a few, it doesn't make any difference) 'reasoning agent' where all
> agents would go to get a decision. Instead each agent should have, I think,
> its own reasoning power, i.e. Rete engine (in our case) and knowledge base.
> Agents can be very lightweight - in some recent tests we ran tens and even
> hundreds of them on the same JVM. The question that arises then is 'do we
> have an inference engine small enough to fit this paradigm?'. For this
> matter I would like very much to have a 'build your own Jess engine'
> capability, where one is able to create and work with very small and simple
> Jess engines, as well as larger and more sophisticated ones. Then we can
> have all kinds of autonomous agents, more intelligent, less intelligent..
> Best regards,
> Alex Bejan
>
---------------------------------------------------------------------
To unsubscribe, send the words 'unsubscribe jess-users [EMAIL PROTECTED]'
in the BODY of a message to [EMAIL PROTECTED], NOT to the
list. List problems? Notify [EMAIL PROTECTED]
---------------------------------------------------------------------
