[ODE] Running ODE backwards in time?

[Adam Moravanszky]

Good point.  You would have to make sure that you never completely stop 
-- you would have to maintain a tiny residual velocity.  Of course with 
float rather than real numbers you would have data loss anyway as values 
get truncated.  Anyway, this discussion is only theoretical anyway, in 
practice there is no way you could pull this off.  In the real world you 
would have to record motion keyframes when simulating forward and 
interpolate backwards -- just like in games that have a rewindable 
instant replay feature.

--Adam

Alen Ladavac wrote:

>I'd say that this is insolvable for general case with dynamic friction. If a
>box slides on a horizontal plane and comes to rest, how do you calculate
>where it was supposed to come from, from its position only? ;)
>
>
>----- Original Message -----
>From: "Adam Moravanszky" <adam.moravanszky at novodex.com>
>To: <mtoussai at inf.ed.ac.uk>
>Cc: <ode at q12.org>
>Sent: Monday, November 15, 2004 10:36
>Subject: Re: [ODE] Running ODE backwards in time?
>
>
>  
>
>>    mtoussai at inf.ed.ac.uk wrote:
>>
>>Hi there!
>>
>>I need to run some simulations backwards in time (for some planning
>>algorithms..). I know I could do this by inverting all dynamic properties
>>((angular) velocities, accelerations, etc.) of the system and running ODE
>>forwards. But is there a simpler way? Just setting a negative time
>>    
>>
>interval in
>  
>
>>a step function doesn't work.
>>
>>Thanks!
>>Marc.
>>_____________________
>>
>>    
>>