There is another option.   Simply don't put any kinematics in EMC and write
your g-code in terms of "axis space".  Then in whatever software creates
the g-code file, the kinematics is done there.   In the end, it does not
matter too much where the translation from joint angles to world
coordinates is done.

SOme advantages to doing the translation off-line are that with a high DOF
arm there are many times an infinite number of solutions and EMC would have
no way to know which is best while the off-line software might be smarter
or at least have more information available.     A typical problem that
takes more information is planning to reach around obstacles and avoid
collisions of the arm and fixtures or even the arm with itself.   The
motion planer needs to know the location of  objects in the work envelope
and the size and shape of the arm.

As a concrete example try this:  Grasp the edge of your desk with your hand
so that your hand can't move.   Keep your should stationary.   Now both
ends of your arm are fixed in location.   But I bet you can still move your
elbow.   With humans, your brain picks the best location of the elbow but
with a robot arm that location needs to be computed.   The question is
WHICH software is the best place for this computation. Is it in "path
planning" or should it be delayed to the real time system (EMC)?   The
answer depends on a lot of details

I think (an opinion) it is best to place the kinematics inside EMC in
simple cases, like say for a pick and place machine but at the other end of
complexity where the arm needs to reach into a refrigerator and grab a beer
bottle that is behind the milk carton  and tilt the bottle so as to get it
out of the tight space,  well that not something for EMC.

I would put kinematics inside EMC if the motion planning task is
trivial and put it outside if motion planning is nontrivial.

One more problem:  If you are building a robot arm that needs to react to
the environment, perhaps this is a palletizing machine that lifts a product
off a conveyor belt and ever one of the parts to be lifted is in a
different orientation and location.   Now you are forced to run kinematics
in real time.  These arms are very common and also there is a similar
problem of picking parts from a loose bin.


On Mon, Oct 29, 2018 at 8:43 AM yomin estiven jaramillo munera <
[email protected]> wrote:

> Hi, I need installing own kinematics for moving a manipulator arm, but i
> don't know  how can i put my kinematics in a file of linuxcnc.
> I first wannna try with only one axis  and then starting to up until to
> have the 6
> but still i dont know ¿how do i put my kinematic?
>
> if you have a link or book which i can see for this topic it would be great
> i have read that a ".c" is installed and this create a ".hal" file which is
> called from my principal ".hal" for executing the kinematics but
> ¿where are those ".c" files in my PC?
> I reload those files and then I executed genserkins.c with the follow
> command:
> sudo halcompile --install genserkins.c
> but it has a problem with genserkins.ko
>
> can someone help me ?
>
> _______________________________________________
> Emc-users mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/emc-users
>


-- 

Chris Albertson
Redondo Beach, California

_______________________________________________
Emc-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/emc-users

Reply via email to