[robocup-rescue-s] A discussion about current problems in preextinguishing...

Dear all,

I would like to discuss with everyone the following issue in the fire
simulator (and the kernel), which regards the effect of using
pre-extinguishing in the fire brigade strategy. Pre-extinguishing (that is
putting water on buildings that are not yet on fire, but which are right
next to buildings that are on fire) is a practice that is commonly used by
real fire brigades in real world fires.

In the roborescue game this is allowed, but putting water on a building that
is not actively burning right now, has a severely discounted factor of how
much cooling the evaporation of water provides (there is a coefficient
factor gamma set to 20% at this point in almost all competition maps that
have been played I believe), compared to a building that is burning (state
= 1, 2, 3). I haven't been able to find a single explanation (or reason) of
why there is such a large difference in the energy constants of water
between these two cases, and why 80% of the water should magically
disappear in this case. This also leads to strange phenomena; for example
when a building is about burnt out (but is still state=3) and you put water
on it and the cooling effect is full, then the next turn it goes to state 8
(completely burnt out) and whatever water you put on it is only 20%
effective. This difference is not justified by how a fire normally spreads,
nor by the physics of water evaporation. I can't quite understand how this
came to be in the fire simulator.

In fact at this point pre-extinguishing seems to be a viable strategy only
because of a bug that exists in the kernel and which sets the ignition
time, whenever the state of a building changes from 0 to something else. In
this case, putting water on an undamaged building -which does little good
from a fire extinguishing point of view- has the effect of setting the
building ignition time (incorrectly) which allows the agents to see this
building's fieryness from far away, even before said building catches fire.

I believe that this is not how the fire simulator should work, so I propose
the following two changes:

1. The BUG IN THE KERNEL MUST BE FIXED. This could be done by only setting
the ignition time when a buildings fieriness changes for the first time to
a burning state (1, 2, or 3).

2. The EFFECT OF PREEXTINGUISHING MUST BE INCREASED. However, in this case
there is one more issue that must be taken into account. We also don't want
to have a strategy (which would be effective in the competition) which is to
put water on buildings that are far away from the fire and then let it be.
So I'm proposing a middle solution, which is also supported by what
normally happens in a fire (I believe). In every turn, right after the
extinguish requests are processed by the fire simulator (and before the
next cooling effect on that cycle), you have some small percentage (like
2%) of the water that has been put on buildings that are not on fire now
(state = 4 to 8) evaporating without providing any cooling, then if the
building is hot you have the cooling effect as it is now (but without the
gamma=20%), and then at the very end of the cycle some more water (if there
is any left), like another 2% of what is left is evaporating again without
providing any cooling effect. This means that if a building is not getting
really hot (because there is no fire close to it), the water put on it will
be evaporating continually and being wasted (which is desired for the
competition and is I believe the reason why the current rule exists with
that coefficient that cannot be logically explained now); in fact with
2%+2% wasted evaporation per cycle (game round), 50% of the water is wasted
in exactly 17 game rounds. On the other hand if a building is right next to
a fire and getting hot, this is now a reasonably effective strategy because
only a small amount of water gets wasted; given that real fire brigades will
try to hose down buildings that are hot enough to catch fire, this change
makes this strategy viable for the competition.
I would like to point out that this change WILL NOT AFFECT THE EXISTING
STRATEGIES ABOUT EXTINGUISHING FIRES, but they WILL GIVE EXTRA STRATEGIC
OPTIONS TO THE AGENTS BY MAKING PRE-EXTINGUISHING A REAL VIABLE STRATEGY.

I would appreciate hearing everyone's opinions about this issue.

Take care,
Ioannis.

_______________________________________________
robocup-rescue-s mailing list
robocup-rescue-s@cc.gatech.edu
https://lists.cc.gatech.edu/mailman/listinfo/robocup-rescue-s
Received on Thu Apr 26 11:44:37 2007