Re: CNC robotic welding arms

Thanks for the info on the EMC files.
As I understand it, EMC can handle all these calculations and would
be the ideal method in theory, but has some rotational velocity
issues in practice ?
I'd imagine that it may just be easier and provide more flexibility
to use a 3D animation program designed and perfected to handle very
complex kinematics (and able to visualise them) to work out all of
the joint / axis movements and then from there, output the
coordinates that each axis has to move to in order for the toolpoint
for follow a subdivided toolpath.
From this, the CNC controller would be able to handle the translation
of these axis movement coordinates into step and direction signals
while being totally unaware of the configuration of the machine if
(and this is the problem) it could determine how fast each axis
should travel for each move.

One method that kind of works is to set a maximum feedrate per axis,
for example, when using a 3-axis controller to do rotary machining
where the Y-axis controls the rotation of the part and the controller
doesn't know how far the toolpoint is really moving but calculates
the feedrate based on the maximum feedrate set for the Y-axis as if
it were a linear axis. This at least gives smooth and predictable
movement but the axis feedrates would be the same whether the tool
was cutting near the centre or at the outer radius of a rotary part,
giving very different toolpoint velocities.

Another method, not without its disadvantages, but that may be very
effective is tagging the distance that the tool tip is actually
moving relative to the material onto each line of G-Code containing
the axis movements (as opposed to the tool movement). The distance
value would be used to calculate feedrate or it could even be
replaced with a feedrate or time value. Would this be enough for the
controller to generate proper constant velocity movement at the
toolpoint?

For example, take a hexapod machining over a cusp or sharp point of a
surface where the tool is changing angle rapidly but the toolpoint is
actually moving very little in relation to the material. Using the
above method would mean a series of six-axis coordinates for the six
hexapod struts to move to, along with corresponding values that
correspond to the distance that the toolpoint is actually moving. The
controller would move the hexapod struts, coordinating their
acceleration / deceleration, target and max velocity not knowing or
caring about the physical arrangement of the machine, but just how
long these axis movements should take. The same should hold true for
other 4 and 5 axis arrangements.
Could / should this method work ?

Rab