Re: Re: CNC robotic welding arms

Mark, Rab

While it would be possible to plot and compute the angles needed by a
multi-axis robot welder and turn those into steps and speed, I don't
personally think that this would be a very practical exercise -- at least not
twice. You can easily simulate the motion required of a robot arm using
your own arm. Put your fingers together into what might pass for the tip of
the wire welder and watch the motions of all of the joints while you try to
weld your way around the sides and across the face of your keyboard. The
problem here is computing where the tip of the fingers/welder will be and
what attitude it needs to take to make a proper weld. This kind of
computation is called kinematics.

Yes the EMC will do kinematic calculations for devices that are not
Cartesian. And it will turn the results into step and direction pulses for
up to six joints pretty much regardless of how they are strung together. The
distribution includes a file to do this for a Puma robot welder. It also has
a similar file for the NIST cable Hexapod. The EMC has been applied to
several Hexapod projects in Germany. I believe that it was applied to a Puma
robot at least once at NIST. The EMC task controller uses these kinematic
computations to produce both position from joint position and joint movements
required to change world position. These files are based on a set of tables
that define strut lengths and angles between struts and their relationship to
Cartesian space. The matrix math for these is available in two files
emc/src/emcmot/genhexkins.c and genhexkins.h. They were written by R. Brian
Register in 1999.

While each joint carries both plus and minus limits with it, we do not have a
very good way of defining singularities (a place where a platform lines up
with a strut in platform motion) or impossible positions of an arm.

EMC also has a number of rotational velocity issues similar to those pointed
out by Rab.

HTH

Ray

>    From: "rainnea" <rab@...>
>
> These would seem in theory to be ideal for plasma or laser cutting.
> Realising that calculating the movement coordinates can be pretty
> straightforwards set me thinking about these robot arms in the first
> place. Just generate G Code that contains all rotational coordinates,
> and most any controller would work in the same way that many folks
> use their controller's Y-axis to rotate a rotary table.
> However, how the controller handles the velocity calculations is
> another matter when you start using rotational axes, as the
> controller really needs to know the distance it's moving to calculate
> the velocity. ie. many controllers have problems with a large
> rotational move combined with a very small linear move.
> Can anyone shed some light as to how high-end controller handle this
> problem ?
>
> Rab
>
> --- In CAD_CAM_EDM_DRO@y..., "Mark Ehle" <ehlem@c...> wrote:
> > A while back I had entertained the idea of building a CNC plasma
>> torch that used rotary joints rather than linear stuff. I reasoned that
>> this would be a lot cheaper way to go if a person could work out the
>> math to use polar coordinates and translate this into step/direction.
> > I have since found a cheap source of linear bearnings/ways (the junk
> > yard) and have abandoned the rotary idea, but still think that this
>> has merit for a cheap way to go. Anybody else doing this? Does EMC
>> handle polar joints?
> > 
> > Later -
> > 
> > Mark Ehle
> > 
> > Dowliing, MI
> >