Hi!
We did now quite a lot of public demonstrations of RoboCupRescue during
events, such as the "Science Days" in the EuropaPark Rust, the "Science
Market" in Freiburg or interviews with local TV stations and newspapers.
It looked like that many people are very interested in RoboCupRescue
research and thus we had many questions to answer. One question that you
ALWAYS will have to answer is: "How differs RoboCupRescue from a
computer game". The only way to argue here is to explain the world model
of an agent (autonomy) and the reality claim of the simulation. People,
particularly kids, understand very well if there are things not
realistic (for example the fact that 10 fire brigades can slip into one
building). Therefore I think a realistic simulation is not just a
feature. In fact, it is necessary for the purpose and existence of our
research.
The nature of a rescue operation in a real world scenario is basically
about dealing with unforseen consequences. People are certainly trained
for particular situations, however, they will always have to make
decisions within situations they even could not imagine before. In order
to come up to this point, the committee probably decided to randomize
maps and by this to prevent agents that solve the whole problem in advance.
If we increase simulation complexity it might happen that agents perform
worser than before (e.g. the map cannot be extinguished). However, since
we are working within a simulation, this can easily be solved by scaling
parameters (e.g. by saying: one fire brigade works now that much as
five). Hence we do not really have the same impact of rule changes as
for example in the F2000 league.
I think we should not start to write agents for computer games, even I
would very much like to:). Instead, we should continuously improve the
simulation towards reality and thus to develop tools that are
increasingly usable within real disasters.
Cheers,
Alex
Mazda Ahmadi wrote:
> Hi All,
>
> The rescue system is too complicated at this point. We had alot of
> complications added last year (e.g. range of firebrigade visions and
> water quantity) and still in the proposals there is a tendency to add
> more complexity. I understand that some people want more realistic
> simulations, but from multiagent research perspective this much
> complexity is not desirable.
>
> I strongly believe we need to make this simulation "LESS REALISTIC"
> and thus "LESS COMPLICATED" in order to have a more interesting
> research infrastructure.
>
> Best,
> --Mazda
>
>
> --
> Mazda Ahmadi
> PhD Student
> University of Texas at Austin
>
>
>
>
>
>> From: Mohsen Izadi <roboresc_mi@yahoo.com>
>> Reply-To: r-resc-l@usc.edu
>> To: r-resc-l@usc.edu
>> Subject: Proposal[Part #1: Simulation Environment]
>> Date: Sun, 14 Dec 2003 07:39:14 -0800 (PST)
>>
>> Dear Friends,
>> Here's our team's proposal for 2004 competitions.
>> Our proposal consists of several parts. In this
>> e-mail I have only attached proposals regarding
>> the simulation environment.
>>
>> Regards,
>> Mohsen Izadi,
>> S.O.S. Member
>>
>> -----------------------------------------------------
>> a) Simulation Environment
>>
>> a-1) In a real situation, when an earthquake occurs,
>> it is usually followed by a chain of [often
>> weaker] earthquakes. This chain also causes
>> building collapse and road blockades. We propose
>> that the blockades and collapse simulators support
>> this feature for 2004 competitions.
>>
>> a-2) Kernel should generate random IDs for units in
>> each simulation. This means that the ID generated
>> for a building in each simulation should be
>> different from the next one.
>>
>> a-3) Agents should be able to specify exactly to which
>> part of a road do they want to move. In other
>> words agents should be able to set PositionExtra
>> property for a target road.
>>
>> a-4) Multiple Nozzles: There are two points about the
>> concept of multiple nozzles :
>>
>> 1) Is there any limits on sum of the amount of water
>> in an agent's nozzles in a cycle?
>>
>> - Our proposal is: Yes! There should be such a
>> limit. But how much? We think that if it is
>> possible for an agent to have 'n' nozzles of
>> power 'p', then it should also be possible
>> for it to have a single nozzle of power 'n*p'.
>> It means that this limit should exactly be
>> equal to the limit for a single nozzle. In the
>> real world, if we consider a tank which
>> contains water and the robot applies a certain
>> amount of pressure to the water from inside the
>> tank to pour a certain amount of water outside
>> the tank, then two facts are obvious :
>>
>> ) First, there is a limit on the amount of
>> pressure that the robot can apply to water
>> from inside to pour it out and the amount
>> of water pouring outside is proportional
>> to amount of this pressure.
>>
>> ) If we have a tank with 2 holes then because
>> of the fact #1, maximum sum of poured water
>> from these 2 holes is not more than the
>> maximum sum of poured water from a single
>> hole.
>>
>> And our proposal for the amount of water is
>> 1000 [m^3/cycle](Like previous years).
>>
>> 2) Is there any limits on the number of nozzles?
>>
>> - Our proposal is: No! As long as the first
>> limit is applied, there is no need to limit
>> the number of nozzles.
>>
>>
>>
>> __________________________________
>> Do you Yahoo!?
>> New Yahoo! Photos - easier uploading and sharing.
>> http://photos.yahoo.com/
>
>
> _________________________________________________________________
> MSN 8 with e-mail virus protection service: 2 months FREE*
> http://join.msn.com/?page=features/virus
This archive was generated by hypermail 2.1.3 : Mon 15 Dec 2003 - 10:43:16 GMT