Re: [Pyode-user] collision woes

[Marc K. <mki...@ga...>]

Sorry one more follow-up question,

What does setting the default axis in a Mass object do if it doesn't define
the direction of the associated body?  For example,

M = ode.Mass()
M.setCappedCylinder(10, *direction (x-axis = 1)*, *radius*, *length*)
body1.setMass(M)
body1.setPosition((*some position x, y, z)*)

I set the mass of my links such that they would be distributed along the
x-axis but is it true that this does not orient the body
along the x-axis?

Thank you again.

marc



On Mon, Jan 17, 2011 at 6:43 AM, Marc Killpack <mki...@ga...>wrote:

> Ethan,
>
> Regardless of the delay, I really appreciate the response.  Your solution
> was exactly correct and cleared up my collision problems.
>
> I was confused about the initial orientation given to created cylinder
> bodies.  Also, although rotation about the center of mass for each body
> makes the most sense, I was confused by the fact that all coordinates (both
> used for setting and returning a body's state) are in a global frame.  In
> any case, works great now, thank you.
>
> To explain my last point more, I'm interested in getting forces that result
> from collisions, but when I tried to use "setFeedback" with my contact
> joints, it threw the following error:
>
> ODE INTERNAL ERROR 2: Bad argument(s) in dJointSetFeedback()
> Aborted
>
> I'd seen this in other threads and thought this meant that getting reaction
> forces from the contact was not possible.  Also, I assumed that because the
> geometry has no "mass", that getting contact forces from the collision with
> a body and a geometry was also not possible.  So I made an extra body for my
> obstacle that is rigidly fixed to the environment with a fixed joint that
> will give forces induced by the collision somewhere along the body.  If
> there is a better/other way, I'd be happy to hear about it.
>
> Thanks again for the help and course correction.
>
>
> Sincerely,
>
> Marc Killpack
> Healthcare Robotics Lab - Georgia Tech
>
>
>
> On Sun, Jan 16, 2011 at 4:45 AM, Ethan Glasser-Camp <gl...@cs...>wrote:
>
>> Hi! Sorry about the delay -- it's hard for me to find free time these
>> days. I'm not really familiar with vpython so I may be vastly
>> misinterpreting the function of your program.
>>
>>
>> On 01/11/2011 03:17 PM, Marc Killpack wrote:
>>
>>> I've been trying to make a simple robot simulator which includes
>>> collisions.  I'm getting weird
>>> behavior though and wondered if I'm just making a dumb mistake.  I want
>>> some objects that are
>>> fixed in the environment.  From other examples and from archived
>>> questions on the list it seems like
>>> the best way to do this is using geometry objects that are not associated
>>> with a body.  The only
>>> collision that works so far is if I collide with the floor plane that is
>>> defined as a fixed geometry.
>>> I've attached my file that uses vpython for visualization and I had the
>>> following questions if anyone
>>> is willing:
>>>
>>> 1)If I set a tall square in place, different sizes will result in
>>> collisions while smaller sizes do not, why?
>>> (Try changing width and length of square to 0.1 in line 56 of .py file to
>>> reproduce this)
>>>
>>
>> It looks like body1, your robot arm, is not being rotated. If I understand
>> correctly, this means that the capsule is extending along the Z-axis,
>> whereas you are drawing it as extending along the X-axis. That is to say,
>> your joint is not swinging the capsule like the arm of a turnstile, but more
>> like the head of a hammer. I think the box obstacle is sufficiently wide
>> that the capsule hits it anyhow, but if it is only 0.1, it fits "inside" the
>> the radius of the hammer swing. To fix this, I put:
>>
>> body1.setRotation([0, 0, 1, 0, 1, 0, -1, 0, 0])
>>
>> Just after your body1.setPosition() call, before adding the geom. (A 90
>> degree rotation around the Y axis, to exchange the X and Z axes, to make
>> your visualization line up with the simulation.) Now a collision always
>> happens.
>>
>>
>>  Also, if I change the bounciness in the collision callback function (line
>>> 28), it seems to have no effect
>>> on the collision, why?
>>>
>>
>> I'm guessing the motor on the joint is sufficiently powerful to keep
>> pressing the body against the obstacle. Try changing MaxForce to 0 after the
>> first time step (at the end of the while loop), so that the body has an
>> initial velocity. You'll definitely see a bounce, and it will definitely be
>> different for different bouncinesses.
>>
>>
>>  2)I can place a geometry object that is a capped cylinder in the
>>> environment but how do I orient it?  I tried
>>> setting the rotation, but still couldn't get a collision to occur.
>>>  Change "OBSTACLE = 2" on line 9 to see this.
>>>
>>
>> This looks right to me, and once I added the rotation of the body (as per
>> point 1), this caused collisions.
>>
>>
>>  3)If I want forces from the collision, it seems that I have to have the
>>> obstacle be associated with a body
>>> as well, is that true?  If so, what kind of hinge or connection should I
>>> use?  I tried a fixed joint to the
>>> environment but that also did weird things.
>>>
>> I don't think this is true. Can you explain further?
>>
>> Ethan
>>
>>
>                                                   <#12d93c9ab7c40535_12d938cf63e95412_>
> <#12d93c9ab7c40535_12d938cf63e95412_>
> <#12d93c9ab7c40535_12d938cf63e95412_>       <#12d93c9ab7c40535_12d938cf63e95412_>
>