[ODE] Particle simulation

Ignacio García Fernández ignacio.garcia at uv.es
Tue Feb 27 04:03:39 MST 2007


El Lunes, 26 de Febrero de 2007 19:40, Jaroslav Sinecky escribió:
> Hi,
> I would say doing particles outside ode is a good idea. Simulating each
> particle as a rigid body is a big overkill and would give very poor
> performance (imagine a particle system emitting 3000 particles per
> second). Usually you need some particle management for very fast
> particle emitting, updating and destroying. The physics is very simple
> part of it: basically just updating velocity in each step applying
> gravity force and/or buoyancy.
> If you need collisions, you can use ode for detecting collisions (as
> suggested by Bram) and do collision resolution your self. Again this
> should be very simple, as a moving point, each particle will have max 1
> contact point at a time and as you don't want particle to "push" other
> bodies, just reflect particle velocity with some restitution factor.
>
> Jaroslav
>

Yes, this is more or less my idea. Even simpler in some aspects, as I will 
have a small numbre of particles. I even consider using geoms for collision 
detection, in some cases.

Thank you!

> Ignacio García Fernández wrote:
> > El Lunes, 26 de Febrero de 2007 11:25, Bram Stolk escribió:
> >> Ignacio García Fernández wrote:
> >>> Hi!
> >>>
> >>> Do anybody have any experience on simulating particles (punctual
> >>> masses) within an ODE simulation?
> >>
> >> I think you would just create a body without associating geometries
> >> with it.
> >
> > Yes, but what about the dynamic part. Application of forces would always
> > be done in the center of mass, but the equations related to the rotation
> > would still be there. This makes a considerable amount of unnecessary
> > computations. If you use true particle dynamics (only trslation, no
> > rotation, mass matrix, etc.) it would save time, and probably would give
> > stability to some simulations.
> >
> > I will do it outside ODE, with its own numerical integrator and so. I
> > wanted to know if any effort is ongoing in order to do this within ODE
> > (perhaps I could help).
> >
> > Also, any idea of how to do it can help me, and do it myself (i can't
> > promise anythig!!).
> >
> >> Your particles will not collide with your world though.
> >>
> >> You would need to write custom collision detection by intersecting
> >> a line-segment with the world. The line-segment would have endpts
> >> p0 and p1 which is the particle pos at t0 and t1.
> >> If the segment intersects the world, you need to do collision
> >> response by creating contact joints.
> >>
> >> Come to think of it... such a particle program would make a
> >> nice addition to the demo collection in ode. E.g. simulate
> >> a water-hose that topples some stacked objects by spraying
> >> particles onto them.
> >
> > Yes!! It would be nice!
> >
> > Thanks a lot!
> >
> > Ignacio

-- 
May the source be with you                                     
---------------------------------------------------------------
Ignacio García Fernández                  Instituto de Robótica
ignacio.garcia_at_uv.es                 Universidad de Valencia
http://robotica.uv.es/~ignacio/                Tlf. 96 354 3671



More information about the ODE mailing list