Re: new tests in competition

From: Mohammad Nejad Sedaghat (mnsedaghat@yahoo.com)
Date: Thu 10 Jun 2004 - 17:01:15 GMT


Dear All,

--- "H. Levent AKIN" <akin@boun.edu.tr> wrote:
> As you know in order to foster new avenues of
> research, we are trying to
> consider different issues in the agent competition.
> This year in
> addition to the classical competition we will
> consider cases where the
> abilities of the agents and the teams are reduced
> and see how the teams
> cope with these increasingly difficult conditions.

   Good idea! But I think announcement of such
modifications in the competition structure is too
late. I really agree to change the competition's
structure toward a more progressive one. But, this
sudden decision, only three weeks before the
competition, will just mix up our development plan.

> We tend to make the
> games increasingly difficult in each stage of the
> games: preliminaries,
> semi-finals and finals.

  Really good!

> We consider two different situations:
> -By changing initial configuration: The competition
> difficulty can be
> increased by increasing the number of fires and
> trapped civilians and
> perhaps by reducing the number of agents, or by
> placing agents trapped
> in buildings.

  Again it is a good idea! But, it is applicable if
the technical committee spends enough time to choose a
set of initial configurations that covers all aspects
of a rescue team.

> -By changing sensing properties: We also consider
> increasing difficulty
> by making the communication to become less reliable
> - easy would be 4
> messages of 256 bytes, medium is 2 messages and hard
> is zero messages,
> and we also consider limiting visibility.

   This may arise some questions. Do the agents get
aware of the new parameters? Which parameters are to
change? Are those only related to visibility and
communication? On what basis you decide which
parameters to change and how?

> We will also have learning: Here on the same random
> map three
> consecutive runs will be made and based on the
> difference of the score
> of the 3rd and 1st run teams will be ordered.

  I don't think that a team can learn using only three
simulations. The other points are mentioned wisely by
Sébastien and Michael.

  As another point, I think that this proposal needs
further consideration. For example, one of the
challenges in search and rescue operations is to
decide about centeralized or decentralized decision
making in such a multi agent system. Such limitations
in communication between agents, leads us to
decenteralized decision making. Which is not yet
proved to be a better solution.

Regards,
--Mohammad
Leader of Caspian Team

        
                
__________________________________
Do you Yahoo!?
Friends. Fun. Try the all-new Yahoo! Messenger.
http://messenger.yahoo.com/



This archive was generated by hypermail 2.1.3 : Thu 10 Jun 2004 - 17:01:29 GMT