[ODE] simulator behaviour problem

[Russ Smith]

> The exact problem is the following: right after the morphology has
> been re-created again I got an "ODE Message 3: LCP internal error,
> s <= 0" error and the hexapod performs a bad move.

i've run your application, and i'm gradually tracking down this 
problem. at this stage it appears that there may actually be an obscure 
bug in the LCP solver - because i can't find any other problem with 
your application.

> It seems that I found a stable order initializing peripheral joints 
> first but this behaviour surprises me. Is joint creation order 
> supposed to make a difference?

it is not supposed to make a difference. but there is a mechanism 
through which it *might* make a difference. different joint 
creation orders result in a system matrix with a different structure,
so the LCP solver will make a different set of switching decisions as
it solves. the result at the end should be the same, but if there IS 
actually an LCP bug, it may be manifested in some cases and not others.

russ.

--
Russell Smith
http://www.q12.org