From: Ranjit Nair (nair@usc.edu)
Date: Thu 17 Oct 2002 - 08:42:09 GMT
---------- Forwarded message ----------
Date: Thu, 17 Oct 2002 01:40:59 -0700 (PDT)
From: Ranjit Nair <nair@usc.edu>
Reply-To: rescue@r.cs.kobe-u.ac.jp
To: rescue@r.cs.kobe-u.ac.jp
Subject: [rescue:5552] Re: Proposals for RoboCupRescue 2003 regulations
I agree that having a common ACL is useful for comparison of teams not
necessarily from a competition perspective but certainly from a research
perspective. However, this can be done by having developers explain what
information they exchange through their individual ACLs as part of their
team description. I don't believe that the simulators should specify an
ACL because this restricts the creativity of developers in coming up with
their own information exchange mechanisms.
Another important use of having a common ACL that you mentioned is the
coordination of the Rescue effort with a team composed of agents
developed by different developers. Here a common ACL becomes crucial.
This is a very realistic problem which is seen in the real world. I
think this can be the basis of a very interesting Scientific Challenge
Problem at the forthcoming RoboCupRescue. If others find the idea of
a Scientific Challenge based on getting agents developed by
different researchers to collaborate, then we can discuss the formulation
of this problem here. Certainly deciding on the ACL to use will be part of
it.
Regards,
Ranjit
----------------------------------------------------------------------
Yoshitaka Kuwata wrote:
> I agree with Koto-san and Ranjit-san with issue (1).
> However, I think we should have standardized ACL in the near future.
>
> Reasons:
> One of the most important research issue is cooperations among agents
> with different purpose and different design. It will be very hard to
> compare the results from two teams with proprietary ACL. We SHOULD
> have common ACL for rescue simulation. With common ACL, agents
> become exchangeable, also.
> However, we have not reached group-consensus, and it seems to require
> much more discussion about this issue.
> I think teams with proprietary ACL *MUST* explain what kind of
> information is used, in order to understand the activities of teams.
> This is very important for both agent research and rescue-operation
> research.
>
> -Yoshi
> --
> Yoshitaka Kuwata; kuwatay@nttdata.co.jp
>
>
This archive was generated by hypermail 2.1.4 : Thu 17 Oct 2002 - 09:05:02 GMT