Hello, Back in the day I have done about 2000 walking skeletons for a demo using the osgCal plugin - i.e. no instancing, separate node each of them, software skinning (ok, skeletons don't have much skin :-p) they just shared some data. That was running at ~20fps on a GeForce4 back then. So it is certainly possible. Not sure how amenable the osgAnimation stuff is to that - to me it looked a bit unfinished and clunky when I worked with it last time.
The comment from Christian about billboards is certainly valid if you want to scale the simulation above some ~20k-50k characters - bellow that you probably will be able to drive fully articulated characters. With some clever LOD you can do even something like this: http://archiveweb.epfl.ch/vrlab.epfl.ch/Movies/Movies_index.html (scroll down for the "Real-Time Navigating Crowds: Scalable Simulation and Rendering" video, the paper is here: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.90.3149&rep=rep1&type=pdf ) Of course, the CPU is the bottleneck, as Christian said - you need to push as much simulation as you can on the GPU. Jan On Tue, Nov 20, 2012 at 3:54 PM, Christian Buchner < [email protected]> wrote: > > Are you sure your agents need to be full 3D models? > > I've done something using perspective billboards, in the osgrvo2 sample in > the osgRecipes repository maintained by Wang Rui. The billboards were > rendered in Poser using a python script that rendered it out from many > viewing angles. > > The GPU can easily render 10000 agents with different walking speeds at > interactive frame rates. The bottleneck is the agent simulation on the CPU. > ;) > > > 2012/11/17 Bruno Fanini <[email protected]> > >> Hello osgUsers, >> >> Within a osg-based crowd simulator I'm currently developing, I was trying >> to investigate a simple skeletal animation for a given set of agents. >> I'm digging the animtkviewer.cpp example (osganimationviewer) and it >> looks like a good solution for this case. >> >> In this context, I would just need a plain simple "walking" animation for >> my characters. >> >> I've a set of PositionAttitudeTransforms nodes to control position & >> attitude of a single agent, so my questions are: >> - is it possible to control the independent speed of the walking >> animation for a single agent? (i.e. another agent is walking faster, etc..) >> - what about instancing 3d models? is it still possible with this >> approach somehow? (i.e. every p.a.t pointing to the same 3d model) >> - what about the performance drop (estimate)? >> - is there any example out there applied to crowd simulations or similar? >> >> Thanks, >> >> >> -- >> Bruno Fanini >> http://BrunoFanini.co.nr >> http://DevHelio.co.nr (Software Development) >> http://www.facebook.com/DigitalDivadlo<http://www.facebook.com/digitaldivadlo> >> (Digital >> Art) >> >> >> _______________________________________________ >> osg-users mailing list >> [email protected] >> http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org >> >> > > _______________________________________________ > osg-users mailing list > [email protected] > http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org > >
_______________________________________________ osg-users mailing list [email protected] http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
