Re: Probing

[Carl Mikkelsen <cmm-at-alum.mit.edu>]

At 11:57 PM 1/16/02 -0500, you wrote:

>On Wednesday 16 January 2002 12:03 pm, Carl Mikkelsen wrote:
>
> > No, I'm doing everything with homebrew software and Linux.  I wrote a
> > kernel driver that does 6-axis coordinated movement in hexapod leg
> > coordinate system, and fronted that with a user-mode process which breaks
> > linear 6-axis Cartesian movements (XYZPRW) into leg movements.
> >
>
>I thought everybody was using EMC for the hexapods.

Maybe "everyone" is, but I'm more of an EE and programmer than metalworker, 
so I started by going inside the kernel to do my low level motor controls.


> > > > This is a picture of my probe...
> > http://www.foxkid.net/cgi-bin/next-res.pl?home_album=/Processed/Carl%27s%20
> >Projects&basedir=./Processed/Carl%27s%20Projects/Stewart%20Platform/Truck%20
> >Signs&res=svga&name=PA250761.JPG&album_name=Truck%20Signs
>
>I saw this when I was browsing the web earlier.  Pretty slick.
>
> > It would be pretty simple within my software to take the Z-axis
> > interpolation I'm doing now and change it to output a normal vector to the
> > surface along which the cut should be made.  My Stewart Platform machine
> > has limits on the angles, depending on where in the working volume the tool
> > need to be.
>
>Is that the typical approach, to calculate surface normals and set
>yaw, roll , and pitch of the head?  XYZ and YRP are completely
>independant for the entire range of motion?

XYZ and PRW (as I call them, there are probably more common names for the 
axes.  I couldn't use Y for Yaw, since it conflicted with the Y axis) are 
independent, but the workspace has a complex shape.  The software has a 
trial mode which goes through the calculations to see if a limit would be 
reached.
The geometric configuration is driven by two files:  a geometry file 
containing the joint locations, leg length limits, and angular limits There 
is a geometry file that contains the limiting leg and joint parameters, as 
well as the geometry calibration information.

>After looking at your setup I have a few more questions...
>1) What are the aluminum sections on the 3 sides of the triangular head?

The aluminum sections contain spherical mounting sockets for laser diode 
modules.  I'm not currently using them very much.  Three point down to 
establish a reference point.  Three point left, right, and back to 
establish calibration points.

>2) Your angle limits are a result of the arms hitting the motor at certain
>      positions?

The legs can hit the motor, as well as each other.  They can also exceed 
the range of the gimbals.

>3) I assume your are using steppers as linear actuators,  this being the case,
>     while you are cutting across a flat plane all  steppers must 
> be  running..
>     Are there problems with resonance?

I have had no problems with stepper resonance making certain ranges of 
speed unachievable.

>4) I'd like to see a picture that shows the entire "cage".

Here is one fairly good picture of the whole thing.  It's also a good 
picture of some other junk in my basement.

http://www.foxkid.net/cgi-bin/next-res.pl?home_album=/Processed/Carl%27s%20Projects&basedir=./Processed/Carl%27s%20Projects/Stewart%20Platform&res=svga&name=P0004052.JPG&album_name=Stewart%20Platform

Generally, starting from 
http://www.foxkid.net/cmm/platform/project-notebook.html you can get to 
many pictures, as well as the design notebook with some calculations and 
sketches.  All of the pictures are stored in several resolutions.  If a 
picture isn't big enough, click on it and it will zoom another level.

> > Sounds like tk/tcl is the tool of choice.  Is there any code you'ld
> > recommend I read to learn it?
>
>Here's a good start:
>http://www.objenv.com/cetus/oo_tcl_tk.html

That looks quite helpful, Dean.  Maybe I can lift my head up from the muck 
of bits and chips (silicon and metal) to do something useful at the UI level.



> >
> > -- Carl
> >
> >
> > Carl Mikkelsen
> > cmm-at-alum.mit.edu
> > www.foxkid.net

Carl Mikkelsen
cmm-at-alum.mit.edu
www.foxkid.net