Dr. Tuan T. Ho
Guest
|
 Posted: Tue Aug 26, 2008 11:28 pm Post subject: <> MultiAgent Systems <> |
 |
|
Greetings!
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?
Cheers!
Rexsy (Dr. TTH)
Rexsy.com
<>*<> |
|
Kent Paul Dolan
Guest
|
| Posted: Thu Aug 28, 2008 7:09 pm Post subject: Re: <> MultiAgent Systems <> |
|
|
Dr. Tuan T. Ho wrote:
| Quote: | 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.
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.
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.
xanthian. |
|
