[ODE] lib question

Jon Watte (ODE) hplus-ode at mindcontrol.org
Wed Jun 14 10:01:00 MST 2006


That's surprisingly small time steps. You should be able to make it "go" 
with time steps more in the 0.01 range. Try using more CFM; setting it 
as high as 1e-5 or even 1e-4. Your ERP can probably go higher than 0.2 
without bad effects, too; try setting it to 0.5 or so. Other things that 
may help include allowing slip in contact joints, using "reasonable" 
friction values (0.3 - 3.0) etc to give the system more ways "out" of 
sticky situations.

Cheers,

          / h+


Bill Sellers wrote:
> Sometimes you really do need to use very low values for your  
> integration step. My horse simulation sometimes falls apart with an  
> integration step of 1e-3 but works quite happily at 1e-4 and when  
> I've been playing around with stiff springs I've had to drop to 1e-5.  
> This is double precision and dWorldStep using ERP="0.2" and CFM="1e-10".
>
> Cheers
>
> Bill
> On 14 Jun 2006, at 10:28, LR wrote:
>
>   
>>  Hi,
>>
>> Changing the time step does not resolves my mass problem, then i  
>> done a very
>> little simulation:
>> 2 boxes (designs a L) connected with a hinge and a plane for the  
>> ground:
>>
>> If the boxes mass = 0,050kg and torque applyed into the hinge is 1 -->
>> bodies take over the world (BANG !)
>>
>> If the boxes mass = 1    kg and torque applyed into the hinge is 1  
>> --> OK
>> If the boxes mass = 0,050kg and torque applyed into the hinge is  
>> between 0.1
>> and 0.5 --> OK
>>
>> It seems that if units are not in the same range, simulation cannot  
>> resolve
>> ... But with those so low values ...
>>
>> Is there a way to resolve it ? Or ... Does i have a problem anyway
>> ???????????
>>
>>
>> Laurent.
>>
>>
>> -----Message d'origine-----
>> De : ode-bounces at q12.org [mailto:ode-bounces at q12.org] De la part de  
>> Ignacio
>> García Fernández
>> Envoyé : mardi 13 juin 2006 18:47
>> À : ode at q12.org
>> Objet : Re: [ODE] lib question
>>
>> What you are experiencing is called numerical unstability.
>>
>> Let's do a test. Instead of increasing masses, sibstantially reduce  
>> the
>> stesize of your simulation. Say, divide it by 10 or by 20. And tell  
>> us if
>> this also solves the problem.
>>
>> The reason is that the frequency of the movement of the low mass  
>> body is
>> very high, and the stepsize is not small enough to describe the  
>> solution.
>>
>> Hope this helps
>>
>> Ignacio
>>
>> El mar, 13-06-2006 a las 14:46 +0200, LR escribió:
>>     
>>> Hy,
>>>
>>> I'm always trying to understant why low mass geoms crash the
>>> simulation.
>>>
>>> A question : I'm using a plugin system: core is running simulation  
>>> and
>>> plugins can act on the core and on ODE. i built ODE ReleaseLib then
>>> ODE code is embedded in each plugin and in the core.
>>> I know it's not a good solution but at this time it's easier for me.
>>>
>>> My question is :
>>> Does ODE keep "global" simulation variables (in this case it's normal
>>> my simulation crash : plugins variables does not contains good
>>> values), or does ODE keep all its variables inside an world object ?
>>> (in this case pfouuu i don't understant what happend).
>>>
>>> Changing the mass resolves all my problems (it seems ...) but it's  
>>> not
>>> normal i must enter something like 0,500 kg for a little robot wheel
>>> ....
>>>
>>> A precision : when plugins are mounted, core gives pointers to  
>>> created
>>> objects to the plugin
>>>
>>>
>>> Laurent.
>>>
>>> _______________________________________________
>>> ODE mailing list
>>> ODE at q12.org
>>> http://q12.org/mailman/listinfo/ode
>>>       
>> -- 
>> May the source be with you
>> ---------------------------------------------------------------
>> Ignacio García Fernández                  Instituto de Robótica
>> ignacio.garcia_at_uv.es                 Universidad de Valencia
>> http://robotica.uv.es/~ignacio/                Tlf. 96 354 3671
>>
>>
>>
>> _______________________________________________
>> 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
>
>
>   


More information about the ODE mailing list