[ODE] Joint instability with small bodys

[Patrick McColgan]

Thanks for the reply, I have now managed to sort my problem.

I read through the faq and then realised that I had ERP set as 1.0, I 
don't know why I had left it as that because I knew it wasn't 
recommended.  Anyway I then found that reducing this fixed it, and the 
smaller the body the lower it needed to be for stability.  I'm not sure 
why this is.

As for your problem; I think its taken as fact that -9.81 isn't enough 
for gravity (obviously if you are doing a simulation you'll still want 
it), I think its due to projection but basically it doesn't look right 
and should be higher (maybe someone else can confirm this).  Using the 
proper value makes things appear to fall slow, also you may want to look 
at the contacts you generate on collision.  When I started using ODE I 
used contacts described in the samples and later had problems because 
they were 'soft', if you take out bounce parameters and such you'll make 
hard contacts so the surface will react stronger when the box hits it.

Hope that helps.

Darabos Dániel wrote:

>You might try the experiment without fixed joints, they do not work
>very well for me (they keep wiggling with QuickStep). For creating a
>fixed body use the method described in the FAQ.
>I don't know however how the problem could be related to the size of
>the objects.
>
>BTW I might have a related problem. I changed the units in the
>boxstack demo to have 9.81 as the gravity and ever since the boxes
>fall into the ground as if they were falling into a swamp. I was
>completely unable to figure it out. I changed around the CFM and ERP
>parameters and now its tolerable, but still worse than in the original
>demo.
>Also increasing the number of iterations helped, but that's pricy.
>There might be a dark secret lurking however...
>
>Daniel
>
>On 6/7/05, Patrick McColgan <patrick at torcinteractive.com> wrote:
>  
>
>>Hello, I'm having difficulties with using the standard ODE joints with
>>'small' objects.
>>
>>Whilst bodies with dimensions around and below 0.2 are fine colliding as
>>rigid bodies, creating constraints between them produces constant
>>instability.  My basic test has been to have a static body floating in
>>midair (using a fixed joint) and then attach another body to it using a
>>ball or hinge, the joined bodies do not collide with each other and are
>>not close enough to collide with anything else.
>>
>>Has this come up before?  I searched the archives but found nothing I
>>felt was relevant, I know it has been said here before that measures
>>should be kept around 1.0, certainly constraints prove to be fine with
>>higher values but has anyone come across this before and maybe found
>>ways to deal with it?
>>
>>I'mn using double precision but is this likely to be a numerical
>>inaccuracy problem?
>>
>>
>>_______________________________________________
>>ODE mailing list
>>ODE at q12.org
>>http://q12.org/mailman/listinfo/ode
>>
>>    
>>
>
>_______________________________________________
>ODE mailing list
>ODE at q12.org
>http://q12.org/mailman/listinfo/ode
>
>
>  
>