[ODE] world update - best practices?

[Patrick Enoch]

No, the solver will become unconditionally unstable. Sounds nasty -  
and it is indeed.

You can identify this behaviour by looking at a single error-element:
it will spike very positive and then very negative. If you sum up the  
absolute value of all elements, the error will be monotonic  
increasing all the time rapidly.

--
It is okay for the error-sum to go up for several iterations, this is  
part of the SOR algorithm. A very large error in one element will  
spread across the other connected elements (e.g. connected by a  
joint), to find a global solution later on.

--
The default value for omega ("W") is 1.3 in ODE. And it appears to  
work well for everybody. You should leave it this way.


Patrick

On 25. May 2007, at 1:45 Uhr, Mark Williams wrote:

>
>> Note: you can set a "W" smaller than 1.0. This is not "foolish" as
>> stated in the linked articel - it is called "underrelaxation". But
>> you dont need that for constant matrices.
>>
>
> Thanks. Would there ever be a valid reason to set it above 2.0?
>
>
> Mark
> _______________________________________________
> ODE mailing list
> ODE at ode.org
> http://ode.org/mailman/listinfo/ode