Re: <> MultiAgent Systems <>
Dr. Tuan T. Ho
> Dr. Tuan T. Ho wrote:
>
> > What is the best and easiest way to design the
> > process of communication, coordination,
> > negotiation, controls, etc. for the Multiagent
> > Systems (MAS)? I've used the Genetic/Evolutionary
> > Algorithms for this process before, and it worked
> > out great for several applications. I thought
> > about using the Simplex and Simulated Annealing
> > Algorithms as well. What about Fuzzy Neural
> > Expert Systems for the process? Any suggestions?
>
> That seems confused. Evolutionary computation
> techniques are essentially search optimization
> techniques, not multi-agent system coordination
> mechanisms. Neural nets deserve the same comment.
> Fuzzy systems are not so specialized, and are
> much more often used in the field of design and
> implementation of controls.
Yes, at firts it seems confusing. But the final goal is to achieve an
optimal system outputs for certain applications through the process of
Multiagent communication, coordination, negotiation, processing, etc.,
and the Evolutionary Algorithms can satisfy the final endeavors...
> Three examples of multi-agent systems that I know are
> particle swarm optimization (PSO), ant colony
> optimization (ACO), and the unique example code of
> "Sugarland", also modeled on ant behavior. The first
> two are also search optimization techniques, but
> they are also fairly primitive in the capabilities
> incorporated into their agents. You might well want
> something vastly more useful.
>
> Despite my comments above, one can, I suppose,
> attempt to use search optimization techniques to
> tune the parameters of multi-agent systems, but it
> seems (to me) a case of going far out of the direct
> path to solving a design problem to make them the
> core of then implementation of such systems.
Besides tuning the Multiagent System (MAS) parameters, we can optimize
the MAS structure, rules, knowledge, etc. This is why the search
optimization techniques are quite useful in the MAS process...
> I forget which one, but there is one US university
> in Iowa that seems to be a "center of excellence" in
> the field of designing and implementing multi-agent
> systems. An internet search should find it fairly
> promptly, at which point you might want to open
> off-line communications with some of the professors
> there to see if you can get the guidance you need.
Good idea...
Thanks!
Cheers!
TTH
> xanthian.