[ODE] What is the best unit

Allan Simonsen Simonsen at rocketmail.com
Mon Apr 11 04:57:12 MST 2005

Hi, Gao Yang;

because of numerical stability, try to balance things
up so that most units, mass-rations, etc, stay around
1.00; maybe 1.7 units would be good? 

We also did some simulation of a character as a capped
cylinder for one of our games. Because players have
very specific expectations in terms of character
control (instant speedup / slow down, turning, etc),
you're going to end up hacking a bit to get a
character control module that feels good. 

We ended up with a light-mass character object, with
high friction and high impulse forces being generated
when the player presses any keys. We initially used a
constraint to keep our capped cylinder from falling
over, but found that sufficiently large forces (like
falling of a cliff onto a sharp, interpenetrating
edge), would break the constraint (this was with
Renderware's Karma). As a fix (*cough..hack*) for
that, we ended up resetting the player's rotation
matrix each frame (get the position and Y axis
rotation, generate a new rotation matrix, translate by
position, set new matrix). It worked pretty well..

There was a talk at GDC 2 years ago, at the physics
round-table, where two of the Havok guys was talking
about having built a custom player controller for
1st/3rd person characters. It was built around a
particle, with a contract generating cylinder round
it. Has anyone experimented with something like that?

Best of luck,
Allan Simonsen

--- Gao Yang <gaoyang at inerworx.com> wrote:
> Hi all,
> Does anybody know what is the best unit for setting
> up humanoid sized 
> objects (e.g. 1.7m) to be simulated in ODE? meter?
> diameter? centimeter? 
> or as big as it can be?
> Thanks
> Gao Yang
> Adam D. Moss wrote:
> > gl wrote:
> >
> >> Anyway, I recently thought that the 'explosions'
> resulting from deep 
> >> or unintentional interpenetrations (eg.
> 'spawning' objects 
> >> overlapping one another, or very high velocity
> collisions) could be 
> >> solved by just storing the 'error correction'
> velocity seperately 
> >> from the velocities due to the physical
> interaction.  This would be 
> >> cleared every frame, so as soon as objects are
> moved apart, they 
> >> would no longer fly off into the distance.
> >
> >
> > This seems like a good idea and has crossed my
> mind before, but
> > I've dismissed it because it seems so obvious that
> I think it is
> > either flawed in important-but-subtle
> physics-engine ways or is
> > already implemented as such. :)
> >
> > --Adam
> > begin:vcard
> fn:Gao Yang
> email;internet:gaoyang at inerworx.com
> tel;work:62445717
> x-mozilla-html:TRUE
> version:2.1
> end:vcard
> > _______________________________________________
> ODE mailing list
> ODE at q12.org
> http://q12.org/mailman/listinfo/ode

Allan Simonsen                    simonsen at rocketmail.com
ICQ# 16606984

Do you Yahoo!? 
Yahoo! Small Business - Try our new resources site!

More information about the ODE mailing list