|
HJ,
I just went through your code.
Right now what you have is a Stochastic Birth-Death Model. There is really
no need for agents as each moth is identical. If you are not going to
expand the moth definition, then you can simply keep track of the number of
individuals you have and calculate the survivor/fecundity N times where N is the
total number of individuals in the system. You can then have a second
variable N' that you add the new recruits to. At the end of the loop, N=N'
and then N'=0. This still leaves you (Yoo?) the option of adding
randomness to the fecundity.
Are you planning on expanding this? Even if
you are, I don't see why you can't have the "step" function in your modelSwarm
and then iterate through the moth list. As long as the moths are not
interacting with each other or different parts of the environment, I see no need
for an asynchronous schedule. In this model, it doesn't matter what order
you do things in, the results will be the same. As it is, the present simulation looks like it can be approximated
quite well by a simple deterministic birth/death process. (i.e. A certain
% survive and they give birth.)
Cheers,
D4
----- Original Message -----
Sent: Saturday, April 21, 2001 6:04
PM
Subject: Synchronous scheduling
Hello,
I'm in the process of developing a discrete-time
population dynamics model of an insect. I am building up from a simple
model, in which agents simply reproduce then die or die without reproducing,
in each time step. To schedule this, I used buildActions from
simpleObserverBug2 in the tutorial as a template:
- buildActions
{
[super
buildActions];
modelActions = [ActionGroup create:
self];
[modelActions createActionForEach: mothList message:
M(step)];
modelSchedule = [Schedule createBegin:
self];
[modelSchedule setRepeatInterval: 1];
modelSchedule
= [modelSchedule createEnd];
[modelSchedule at: 0 createAction:
modelActions];
return self;
}
[In MOTH.M:]
-
step
{
unsigned i;
if (survive == 1)
{
for (i=0; i < [myModel getFecundity];
i++)
{
[self
giveBirth]; // New agents are assigned a "survive" value, then get added to
mothList
}
[myModel removeMoth:
self]; // I get removed from mothList
[self
drop];
puts("I gave birth, then
died.");
fflush(stdout);
}
else
if (survive == 0)
{
[myModel removeMoth:
self]; // I get removed from mothList
[self
drop];
puts("I died before giving
birth.");
fflush(stdout);
}
return
self;
}
The step code has two problems, I think, which need to be
fixed. One, since offspring get added to the same list as their parents,
they too get sent the step message in the same time step as do their
parents. What I really want is for offspring to get added to the list in
the following time step. Two, the program crashes following the first
moth removal (see below). I'm less sure about what causes this -
possibly a conflict between the createActionForEach call and the call to
removeMoth from mothList?
0 [main] NCP
1141 handle_exceptions: Exception:
STATUS_ACCESS_VIOLATION
647 [main] NCP 1141 stackdump:
Dumping stack trace to NCP.EXE.stackdump
There are two possible
solutions to these problems which I can think of given my limited programming
knowledge. One, I can put offspring into a temporary list, then dump
them back into the main list at each time step. This doesn't seem
efficient, nor does it solve the crash problem caused by moth removal.
Two, the schedule can be made dynamic, kind of like the one used in
Mousetrap? I'm not sure about how to build this sort of schedule for
this situation.
Does anyone have advice on how to deal with this kind
of scheduling issue? Examples you can point me to? Or maybe
there's a much simpler approach which is escaping me?
Thanks very much
in advance.
Ho Jung Yoo
|