Re: Sub-swarms

[Ken Cline]

Paul,

I do vaguely remember this, and no, I don't think my advice
works.  I didn't try implementing the reset functionality as
I described below but I did try something similar.  After
several days of countless failures I shelved to work as
extra functionality to completed at some later date (ie
catagory: "Bell or Whistle").

As I recall, I could never get the schedules/activities to
die quietly; I have a nack for making them core dump :). 
Anyway, in Sept. we started work on a different model and
implementing a reset hasn't been an issue yet (ie runs are
typically done in batchmode, with multi-run "studies" setup
with a Java applet, ala Drone).

Sorry I can't be more helpful.

Ken.

PS: I seem to recall Vinicius was able to get something
    working.  Vinicius?


On Tue, 24 Feb 1998, Paul Johnson wrote:

> Dear Ken:
>   Do you remember this?  I wonder if you think your advice works.
> 
>    Now I'm grepping through the swarm list trying to fill out answers to a
> list of programming FAQs that Rick Riollo and a few others have helped me
> formulate.  One question I certainly want to answer is "How can I make my
> simulation run over and over again," and your comments here are, as usual,
> excellent and understandable. I just wondered if they are "up to date"
> 
> 
> On 05-Aug-97 Ken Cline wrote:
> > 
> > Vinicius,
> > 
> > I gather you are trying to run several simulations
> > sequentially inside one.  I haven't tried this, but that
> > doesn't stop me from responding [grin] ...
> > 
> > Some thoughts:
> > 
> >    - ObserverSwarm's are (usually) a subclass of GUISwarm
> >         and, as such, have a control panel.  You may or
> >         may not want more than one control panel.
> > 
> >    - If you do want a control panel for each "run" inside
> >         the simulation then you may have to create a new
> >         type of control panel.  For example, if you click
> >         `quit' on any control panel I *think* this will
> >         end the whole simulation.
> > 
> >    - You might consider having only one ObserverSwarm and
> >         just creating a new ModelSwarm for each "run".
> > 
> > 
> > This is the way I am thinking you could do it:
> > 
> >    In main.m:  No Changes
> > 
> >    In ObserverSwarm:  
> >       Move all the code in buildObjects, buildActions, and
> >       activateIn that builds the ModelSwarm and related 
> >       analysis objects.  Put all this stuff in a new method 
> >       called `buildModelSwarm'.
> > 
> >       Now, on the displaySchedule, create an action that
> >       calls `buildModelSwarm'.
> > 
> >       Use `setStateStopped' to pause between each run.
> > 
> >    In ModelSwarm:
> >       Put in a method that accepts a list of agents and
> >       uses those agents to populate the model.
> > 
> >       Also, add a method that does a "checkToStop" and
> >       schedule this method on the model's repeating
> >       schedule.
> > 
> >       Finally, add a `stop' method that, when the model
> >       decides to stop, it evaluates the current population,
> >       passes the "most fit" agent's back to the Observer and
> >       drops the rest.
> > 
> >       Finally, the ModelSwarm  notifies the ObserverSwarm
> >       that it has completed so that the ObserverSwarm can
> >       drop the unnecessary objects and then create a new
> >       model using `buildModelSwarm' and the "most fit" from
> >       the last run.
> > 
> > 
> > Again, I've never tried any of this and I have no idea if
> > it would even work.  If you've already started on another
> > way then I would probably stick with that, if I were you.
> > 
> > (I thought someone had already built something like what
> > you're trying to do, but I could be mistaken.)
> > 
> > Does anyone else have another way?
> > 
> > 
> > Ken.
> > 
> > 
> > 
> > On Mon, 4 Aug 1997, Marcus Vinicius Pereira Pessoa wrote:
> > 
> >> Hi all,
> >> 
> >> I’m trying to call swarms within a swarm. I’m making an application in
> >> which I will evolve a group strategy. In a model similar to hello-world,
> >> I will use a variation from Merelo’s GA lib. So I have some questions:
> >> 
> >> 1) How can I call a whole application via the main-ModelSwarm?
> >> 
> >> 2) From which sub-ObserverSwarm method can I receive a parameter back?
> >> 
> >> 3) Will the matter previously discussed, about problems with the
> >> schedule, have any influence in the model?
> >> 
> >> I think that the evaluation function will be something like:
> >> 
> >> float evalFunc( id<Chromosome> a ) {
> >> 
> >>   int argc;
> >>   char ** argv;
> >>   PplObserverSwarm * observerSwarm;
> >> 
> >>   initSwarm(argc, argv);
> >> 
> >>   observerSwarm = [PplObserverSwarm create: globalZone];
> >>   [observerSwarm buildObjects: a];           //here I’m passing the group 
> >> of
> >> agents
> >>   [observerSwarm buildActions];
> >>   [observerSwarm activateIn: nil];
> >>   [observerSwarm go];
> >> 
> >> 
> >>   return "evaluated value";                  //a value must return
> >> }
> >> 
> >> Thanks in advance,
> >> 
> >> Vinicius.
> >> 
> > 
> > _________________________________________________________
> > Ken Cline                             address@hidden
> > SAIC                                 VOICE (410) 571-0413
> > Annapolis, MD                          FAX (301) 261-8427
> > 
> > 
> ----------------------------------
> E-Mail: Paul Johnson <address@hidden>
> Date: 24-Feb-98
> Time: 23:53:23
> 
> This message was sent by XFMail
> ----------------------------------
> 

_________________________________________________________
Ken Cline                             address@hidden
SAIC                                 VOICE (410) 571-0413
Annapolis, MD                          FAX (301) 261-8427



                  =================================   Swarm-Support is for 
discussion of the technical details of the day
   to day usage of Swarm.  For list administration needs (esp.
   [un]subscribing), please send a message to <address@hidden>
   with "help" in the body of the message.
                  ================================