[ODE] ODE spatial inconsistency

[Martijn Buijs]

pilat at cpsc.ucalgary.ca wrote:
> I've run into a problem with ODE.  I simulate simple shapes joined with
> universal joints.  When i create my shapes with the same position, they
> have exactly the same behaviour.  Problem is, when I create them in a
> different position, the behaviour can very greatly - even if the
> difference in position is only 0.01 instead of 0.0 on one of the axes. 
> I've thought that my simulation might be broken, but after many hours
> spent on testing various things, I isolated the problem to ODE having
> inconsistent behaviour.

ODE is deterministic, each run of a simulation should have the same results if the initial setup is 
the same.

AFAIK, your problem has to do with floating point accuracy. Generally, the results will vary more 
and more as distance between the two positions increases.

The solution would be to create two independent instances of the simulation, run them independently 
of each other, with identical positions for all shapes.

Martijn