[ODE] Previous timestep significance?

[Adam Moravanszky [Novodex]]

Hi,

> I understand now, thanks a lot. Implementation in ODE of the example
> you proposed behaved very much like you predicted.

Cool.

> This was accomplished by using the formulae
>   ERP = h k_p / (h k_p + k_d)
>   CFM = 1 / (h k_p + k_d)
>     h = timestep, k_p = spring constant, k_d = dumping constant
> ( see http://opende.sourceforge.net/ode-0.039-userguide.html#ref25 )
> , and trying to keep spring and dumping constants really being constant.
>
> Doing that made the system behave "well", or at least seem to do so.
> Have you tried this kind of solution?

I don't use a cfm variable in my code so I don't know about this.  In any
case it doesn't sound like a good idea to add even more terms which have 1/h
in them for the numerical accuracy reason I mentioned, unless you disallow
very small time steps.

> ( http://q12.org/pipermail/ode/2002-April/005135.html )
> seems to me very useful.

> Do you know the fate of this idea?

I do something along these lines that does work nicely, but it alone doesn't
completely solve the problem we're discussing -- it solves a different one.
But it depends on the exact implementation, so why not try it and find out?

--  Adam