Re: Stewart platform kinematics (genhexkins)

[Till Franitza <xfa-at-isw.uni-stuttgart.de>]

Hallo Andy!
What you talked about is referred to as Inverse influence in SIEMENS jargon.
When using spherical joints or specially designed kardanic joints you do not
have this effect. I have never written a kinematic transformation that
handles this, but i think it is not too difficult. The kinematics
transformations of emc have not been looked at for a longer time, while the
call to the function has changed. I recently implemented kinematic
transformations for my 2 machines
http://www.isw.uni-stuttgart.de/personen/t_franit/primodell/index.html and i
found i had to change the names of the variables in respect to older
versions. This is because emc supports 5axis-milling now.
Emc does the forward trafo quite often. This is very time consuming so you
might want to comment it out. EMC always displays the position in cartesian
coordinates that are calculated from the actual position values of the
motors. It would be easy to comment this out and just use the commanded
position for display. It is done like that on real machines that i know. On
my real-world machine ( linapod 3 on www.dynamil.de) the forward kins are
only calculated once when the motors are turned on. After that the control
just sends commanded values to the motor controllers and assumes the
controllers keep them to display precision.
for the callinc conventions of kins just look at trivkins.
Do you want me to send a sample of my trafo?
Till

Andy Anderson wrote:

> Questions about genhexkins:
>
> 1) Are the calculations documented anywhere? Specifically, what joints
> are assumed?
>
> This is a leading question. I imagine the calculations assume spherical
> joints. I'd like to confirm this.
>
> But of course many designs use universal joints to support the torque in
> screw/nut actuators. That's probably OK as long as the platform remains
> square to the world, but once the platform rotates, U-joints do not act
> like spherical joints.
>
> Is this reflected in genhexkins? (I assume not.) How is this handled in
> such machines? Is it feasible to create kinematics routines including
> the effects of U-joints? Is there another way to compensate for this
> (like feedback of the strut length, or actual platform position)? This
> is probably wandering OT, but I would appreciate anything that points me
> in the right direction.
>
> 2) What is the status of the forward kinematics?
>
> Skimming the archives, it seems that at one time the forward kinematics
> were disabled, and maybe broken? In the emc-2.1-16 source, it looks as
> though they are hooked up again. Is this correct?
>
> 3) What are the forward kinematics used for?
>
> Skimming emcmot.c, looks like the only applicable call has to do with
> 'free mode'. Free mode seems to be direct/manual input of the joint
> motion? So the forward kinematics are used to compute world coordinates
> when the someone/something is directly controlling the actuators? EMC
> can then display these, and knows where the platform ends up?
>
> The foward kinematics are apparently not used for any closed-loop motion
> control?
>
> Thanks,
> Andy Anderson

--
Dipl.-Ing. Till Franitza
Institute for Control Engineering of Machine Tools and
Manufacturing Units (ISW)
c.o. FISW
Rosenbergstr. 28
D- 70174 Stuttgart
Germany

phone: ++49 (0)711 22 99 2-28
fax:   ++49 (0)711 22 99 2-22
email: till.franitza-at-isw.uni-stuttgart.de
web:   www.isw.uni-stuttgart.de