Dear Friends,
Regarding the discussion about the way some simulators
model some phenomona, I would like to add some points
which are my personal ideas and not directly inspired
by any set of competition rules:
Simulators in RoboCup Rescue Simulation environment
can be divided into two major groups. Some simulators
aim to simulate and impose the rules underlying urban
flows like the traffic rules which are simulated by
'Traffic Simulator'. I personally think that all the
details about the modelling of the events by these
simulators should be known by the teams and at least
there should be some means to find these details other
than studying the source-codes.
On the other side, we have disaster simulators which
simulate the disasters and their direct/indirect effects
on the environment. Unlike the first group, I think that
no direct information regarding the way they simulate
phenomena and events should be available and none should
be used by the participating teams in their agent
development process. The main reason for my opinion is
the fact that in the real world there is no exact
information available about the way many natural phenomena
happen. For example, the way an earthquake happens or the
way a burning building spreads fire does not have exact
and deterministic rules and even the best high-precision
simulation methods available do not give ideal results.
Furthermore, having information about the exact way some
simulators work may reduce the problems of RoboCup-Rescue
Simulation environment to some really easy ones.
By the way, in most of the previous competitions there were
teams that made use of such information by studying the
source-code of some disaster simulators and certainly, these
teams would not be the last ones doing so. And logically, it
cannot be said that these teams have done anything against
the rules as there were no official rules clarifying this
point. May be some of us can argue that they have wasted
their time reverse engineering instead of conducting multi-
agent and AI research but the fact is that according the
rules they are allowed to waste their time ;-)
Here, the main concern would be the fairness of the competitions
in case of having teams thinking differently about what is allowed
and what is not allowed to do as a participating team in the
competitions which, as I think, is one of the current important
problems of RoboCup Rescue Simulation. To achieve the fairness,
There are two solutions:
- Documenting the details of all simulators for all teams
officially. This is an official sign indicating that this
kind of information is allowed to use. Furthermore, teams
are no more needed to study and reverse-engineer simulator
source-codes and the distinction between the participating
teams will not be their skills in studying Java/C/C++ codes.
- Hiding the details of disaster simulators from the teams.
In other words, we can use the term 'secure'-ing the simulators
from being reverse engineered directly. Here, we also have
two ways:
* Changing open-source policy for disaster simulators.
This solution has a lot of problems both theoretical
and in action. Furthermore, this kind of solution i.e.
'Security by Obscurity' is proved to be a flawed and
incapable approach in various aspects of security-related
issues such as cryptography.
* Designing and implementing disaster simulators that are
themselves secure. I would like to define the term 'secure'
for a disaster simulator as 'the state of a simulator
the behaviour of which cannot be directly/easily simulated
by just studying its source-code'. Of course, there are
ways to do so and I think this is an important property
that all simulators should ideally have.
As a conclusion, I would like to say that currently, we have
some disaster simulators which fail to achieve accepted levels
of security. What I want to suggest the participating teams
to model the behaviour of disaster simulators just by observation,
data acquisition, data mining, and learning methods instead of
studying source-codes.
At last, by all these words, I wanted to open the discussion
about this subject in the community. Please, in case of having
ideas or suggestion, express them in the mailing-list.
Regards,
Arash Rahimi
> Degrading humanoids health point is miscsimulator's job and based on a
> random seed in config.txt
> effect of fire, buriedness, ... on humanoids' damage is computed by
> misc simulator in each cycle and then new hp will be set to [old hp -
> damage]
>
> have fun,
> Mohammad Mehdi Saboorian
> SBCe_Saviour
>
>
> On 1/26/06, Sarvapali Ramchurn <sdr@ecs.soton.ac.uk> wrote:
>> Thanks again.
>>
>> Now, does anyone have the 'real' model of how a humanoid's healthpoints
>> degrades over time according to its buriedness/damage/fire etc..? I've
>> noticed different teams implement different estimation functions of
>> these variables. Could someone point me to the code that implements the
>> degradation of the humanoid's health otherwise?
>>
>> Regards
>>
>> Gopal
>>
>> Mohammad Mehdi Saboorian wrote:
>>
>> >There is no point in saying something to civilians (for now).
>> >maybe in future rules and implementations changes make use of this
>> feature!
>> >
>> >have fun,
>> >Mohammad Mehdi Saboorian,
>> >SBCe_Saviour
>> >
>> >On 1/25/06, Sarvapali Dyanand Ramchurn <sdr@ecs.soton.ac.uk> wrote:
>> >
>> >
>> >>Thanks for this.
>> >>I now cannot see the point of telling a civilian agent something if it
>> >>is not able to 'act' upon what it hears. As far as i can see, a
>> civilian
>> >>can only act based on it's knowledge of its buriedness, refuges in
>> >>sight, or fieriness of buildings... but does not have a means to
>> >>transform a heard message into meaningful knowledge right?
>> >>
>> >>Regards,
>> >>
>> >>Gopal
>> >>
>> >>
>> >>On Tue, 2006-01-24 at 20:39 +0330, Mohammad Mehdi Saboorian wrote:
>> >>
>> >>
>> >>>Dear friend,
>> >>>KA_HEAR includes KA_HEARSAY and KA_HEARTELL. so every HEARSAY message
>> >>>is duplicated as HEAR message.
>> >>>thus, there is no need for civilians to change their hearing policy.
>> >>>
>> >>>have fun,
>> >>>Mohammad Mehdi Saboorian
>> >>>SBCe_Saviour
>> >>>
>> >>>On 1/24/06, Sarvapali Dyanand Ramchurn <sdr@ecs.soton.ac.uk> wrote:
>> >>>
>> >>>
>> >>>>Hi,
>> >>>>Can someone please tell me if 'civilian' agents can hear 'say'
>> messages
>> >>>>from other agents. Right now, from the code in ver 0.48, it seems
>> >>>>civilians can only understand KA_SENSE and KA_HEAR but not
>> KA_HEARSAY -
>> >>>>which results in "Unknown header2. 0x54".
>> >>>>
>> >>>>It would make more sense if they could hear such messages when
>> called
>> >>>>out by other platoon agents. Will this happen in the competition?
>> >>>>
>> >>>>Cheers,
>> >>>>
>> >>>>Gopal
>> >>>>
>> >>>>
>> >>>>_______________________________________________
>> >>>>robocup-rescue-s mailing list
>> >>>>robocup-rescue-s@mailman.cc.gatech.edu
>> >>>>https://mailman.cc.gatech.edu/mailman/listinfo/robocup-rescue-s
>> >>>>
>> >>>>
>> >>>>
>> >>
>> >>
>>
>>
>> --
>> Dr. Sarvapali Ramchurn
>> Intelligence, Agents, Multimedia Group t: +44 (0) 23 8059 3270
>> Electronics and Computer Science f: +44 (0) 23 8059 2865
>> University of Southampton e: sdr@ecs.soton.ac.uk
>> Southampton SO17 1BJ, UK.
>> http://www.ecs.soton.ac.uk/~sdr
>>
>>
> _______________________________________________
> robocup-rescue-s mailing list
> robocup-rescue-s@mailman.cc.gatech.edu
> https://mailman.cc.gatech.edu/mailman/listinfo/robocup-rescue-s
>
_______________________________________________
robocup-rescue-s mailing list
robocup-rescue-s@mailman.cc.gatech.edu
https://mailman.cc.gatech.edu/mailman/listinfo/robocup-rescue-s
Received on Sun Jan 29 08:03:04 2006
This archive was generated by hypermail 2.1.8 : Sun 29 Jan 2006 - 07:03:04 GMT