Re: [Swarm-Support] 3-D Representation

[Rick Riolo]

This won't solve pauls' problems, but we have used
a Graph3d, a very simple java based 3D surface plot generator 
real time from RePast/Java models, so i imagine
you could use it from Swarm/Java, too.
To see a screen shot:
   http://cscs.umich.edu/lab/graph3d.png
(You can see text telling what can be done real time...
rotate, zoom, etc, but you can't make the screen shot do that!)
I have forgotten how fine-grained our plots are, but i think
maybe 50x50 or maybe 100x100 points are plotted, and it
goes pretty fast.  probably calculating all the z=f(x,y)
values is the bottleneck, not the plotting.

We actually can drive several of these at once, plotting
different things, and we can even change attributes in agents or sets of 
agents with sliders, and see the surface plot (which represents a 
preference landscape) change real-time, too.
so again, its pretty quick, if you don't ask it to do too
many points.

- r

-- 
Rick Riolo                           address@hidden
Center for Study of Complex Systems (CSCS)
4477 Randall Lab                
University of Michigan         Ann Arbor MI 48109-1120
Phone: 734 763 3323                  Fax: 734 763 9267
http://cscs.umich.edu/~rlr

On Thu, 17 Jul 2003, Paul Johnson wrote:

> Date: Thu, 17 Jul 2003 11:17:45 -0500
> From: Paul Johnson <address@hidden>
> Reply-To: address@hidden
> To: address@hidden
> Subject: Re: [Swarm-Support] 3-D Representation
> 
> Great. How about you look at the ballet dancing project Tina Yu and I 
> did this last spring.
> 
> http://lark.cc.ku.edu/~pauljohn/Swarm/MySwarmCode/Dancer/
> 
> There is a Swarm model with dancers and they follow certain aesthetic 
> rules that Tina created. I was using the 2d graph library to show the 
> dancers as ovals on a square. I outputted the information about their 
> positions and such to a data file, and Tina made a quick time movie of 
> the dance with a program called lifeform. That showed stick people 
> frolicking about.
> 
> I think it would be cool if you could write a 3d thing to directly 
> attach to swarm to make the same representation. Maybe you could use the 
> dancer model since it is already a working prototype.
> 
> A slightly abbreviated version of that paper was published in an art 
> journal called YLEM about a month ago.
> 
> That dancer model has one feature that i think is really interesting 
> from a Swarm scheduling point of view.  Each dancer chooses "steps" from 
> a repertoire.  Some steps take one unit of time, others take more.  So 
> this code features a sort of dynamic scheduling in which each dancer 
> chooses a step, and then only when that step is finished, will the 
> dancer choose another step.  All of the dancers are operating on 
> individualized time sequences, i suppose I mean to say.
> 
> Artem Baguinski wrote:
> > On Wed, Jul 16, 2003 at 05:35:34PM -0500, Paul E Johnson wrote:
> > 
> >>I was freaked out there for a minute. I thought you meant somebody 
> >>produced that 3d picture "graph.gif" with Swarm.  The 3dlib you refer to 
> >> implements positional record keeping, but does not do any of the work 
> >>of drawing the image onto a beautiful picture.
> >>
> >>Has anybody used Swarm to make a 3d picture?
> > 
> > 
> > i'm interested in using Swarm in real time 3D graphics software. 
> > 
> > being employed in a new media art lab, i'm interested in finding new
> > ways of creation "interesting" graphics, audio, text :-) and now and
> > then i spawn myself in a side projects when i have a hunch that
> > they can deliver something new.
> > 
> > so, I'm about to start such a new project, just need some time to get
> > accuainted with swarm. The idea is to visualize swarm based models in
> > fancy 3D. Since art is my domain I can sacrifice some accuracy of
> > visualization in favor of fanciness, may be my approach isn't that
> > useful for serious scientists.
> > 
> > 
> >>It seems to me that drawing a 3d picture like that during a simulation 
> >>would make it really slow, wouldn't it?  If you run a program to draw a 
> >>picture like that from a dataset, it is usually not speedy...
> > 
> > 
> > that picture is way simple for modern (graphics) hardware. although it
> > depends of course what for parameters are visualized. it is statistics
> > calculations that may take a lot of CPU, but I'm guessing, if you design
> > your model taking the stats your interested in into account, you can
> > optimize statistics calculation and make real time visualization
> > possible, wether in 3 or in 2 dimensions.
> > 
> > the application may be build that way that stats computations are only
> > made if stats-swarm is requested (if i got it right that's the way you
> > deal with GUI in swarm). then one can run simulation without stats /
> > visualization for some time, save results and restart with visualizer to
> > see what happenned to the model while (s)he was not looking. [here i
> > imagine a model that slowly changes in time, like ecosystem for
> > example].
> > 
> > 
> >>Makes me wish there were a blt 3d component we could  hook up.
> >>
> > 
> > 
> > GNU Maverik is a good GPLed virtual reality library. it's written in object
> > oriented manner in C, i guess wraping it, or part of it, in Objective-C
> > could provide swarm with nice framework for realtime 3D visualization.
> > 
> > http://aig.cs.man.ac.uk/maverik/
> > 
> > 
> >>If this were my project, I would be inclined to say that the simulation 
> >>should be in Swarm, but you should output the positional data and make 
> >>the beautiful 3d picture with some other program that is intended for 3d 
> >>pictures.  I have a really great commercial stat package called S+ that 
> >>can do that, and its free cousin R can do it, but not so easily.  I saw 
> >>some great graphs of that sort in a SAS presentation lately.
> > 
> > 
> > If stats you wanna see are really complicated, then this is definitelly
> > the way too go.
> > 
> > _______________________________________________
> > Support mailing list
> > address@hidden
> > http://www.swarm.org/mailman/listinfo/support
> 
> 
>