Re: [CAD_CAM_EDM_DRO] Re: was, Steel plate for hexapod setup questions

R Rogers wrote:

>Just thinking about this hexapod principle and a few questions immediately come to mind. How does it know X from Y? Does the operator mount a workpiece and pick up different points of geometry to reference? For instance, all four corners of the top of a block. Seems that the block or workpiece would have to be of perfect proportion. How does it know when the axis of the spindle is at 90 degrees etc. to the work surface? Home reference hard limit on each actuator to reference zero's?
>

Short answer:
I have reference lines on the bed.

Long answer:
I calibrate the machine beforehand. This only has to be once, when the
machine is put together. I start by measuring the actuator extension
lengths with a dial caliper. I enter those into the RoboFac software.

I have a plate with various coordinates measured and marked on it which
is attached to the bed. Then I attach a pointy piece of metal with known
dimensions to the mobile platform. I move the platform so that my pointy
probe touches several points on the plate. At each point I make sure
that pitch and roll are 0 by having levels on the mobile platform. Yaw
is made 0 by measuring from a big straight bar on the platform to
reference plumb lines on the base.

At this point, I know the value of all 6 axes, so I enter them into the
RoboFac software. The software has calculated the actuator length at
each point because it knows the number of motor steps and the pitch of
the lead screws, like any open loop system.

After I have several points, I can run the calibration part of the
software. RoboFac makes a series of guesses about what the geometry
might be: how long the actuators are, and what the endpoints of the
actuators are relative to the reference points. Eventually it comes up
with a good guess for the geometry. I've tested the machine with a pen,
and had encouraging results.

It occured to me this week that this general technique could be used to
calibrate any CNC machine.

>Is there a special software that calculates the pulse train to each actuator for just a linear move? Seems like it would be a nightmare to program. Alot of calculations even to do simple movements.
>

You know the locations of the endpoints of the actuators on the base;
they never move. You know that the other endpoints of the actuators on
the mobile platform never move relative to your tool. You then just use
the Pythagorean theorem to calculate the lengths the actuators should
be. Rotation complicates things a bit, but not too badly.

Take a look at the source code and see. The algorithm is in the file
HexapodGeometry.java, in the method computeInverseKinematics().

>Of course, the 3-D for modeling capabilities etc. would make it a worthwhile endeavor. Something to consider, if this software could be used in reverse and replace the actuators with linear scales it would be a rather crude ...tadaaa!!! CMM machine. :-)
>
>Ron
>
>

The forward kinematics problem (figuring out the platform position from
the lengths of the actuators) is a lot harder, but not impossible.