Thanks Faust,
This is a good paper to see the kinds of features that are desired by
robotics researchers. We can't get the hardware they're using of course,
but we could easily build our own controller for the robot and keep those
features in mind.
I'm hoping the creator of beagleg gets tired of hacking on small, harmless
3d printers that can barely break the skin and instead gets excited about
our motoman k10s. It would be really nice to see a beaglebone black
replace a control cabinet the size of a refrigerator, while simultanously
granting the robot exponentially more flexible capabilities like vision..
I stand ready to implement any hardware requirements of any programmers
interested in implementing a 21st century control strategy for the arm.
This means bypassing the ERC and all its geometry and acceleration
calculations, and communicating directly with the motor controllers and
position encoders.
-jake
On Wed, 10 Sep 2014, johanna faust wrote:
not that this is my field or anything, but this thread
was pretty interesting, and poking around I found this:
-------------------------------------------------------------------
The Fast Research Interface
for the KUKA Lightweight Robot
Günter Schreiber, Andreas Stemmer, and Rainer Bischoff
"Abstract—The KUKA lightweight robot (LWR) provides many unique features for robotic
researchers. To give full access to these features, a new interface
was developed that gives direct low-level real-time access to the KUKA robot
controller(KRC) at high rates of up to 1 kHz. On the other hand, all
industrial-strength features, like teaching, motion script features, fieldbus I/O and
safety are provided. Using standard UDP socket technology, the user
is not limited to one specific runtime system. This paper describes the capabilities of
the interface,the practical realization within the LWR control
architecture, and first applications of the interface. "
-------------------------------------------------------------------
PDF here
http://www.wies-pro.de/pages/publications/KUKA_FRI_from_WS_Proceedings_ICRA…
-------------------------------------------------------------------
because i dearly love to find things sometimes,
especially if relevant
On Tue, Sep 9, 2014 at 11:12 PM, Jake <jake(a)spaz.org> wrote:
My friend Hao of Dorabot wrote me an email that clarifies what we can expect from
the robot. Of course this assumes no hacking, and we can
hack.
one thing I was interested in was "remote" mode through the serial port.
I believe that it can be set to be controlled by commands like "let
me send you this job" "start that job" which would allow us to
effectively control it in real time.
but that will take some reading of manuals and experimenting. As for the latter, i
have connected the robot's serial port to the desktop
computer we use for 3d printing, and I can provide logins for that machine to
anyone who wants to experiment. I confirmed that with the
null-modem adaptor (thanks Somebody!) it does communicate, although i haven't
had any meaningful conversations with it.
of course there is always the radical alternative - the entire YASNAC control
system could be replaced with a homebrew solution that allows us
to do whatever we want. The problem is that it would take a bit of programming to
drive the axes at speed without overshoot, and there's
dynamic gravity and inertia dependant on the posture of the thing and the mass of
whatever it's holding.. could be a lot of math to get it
perfect.
however, if someone were to create a control program for it, a lot of people around
the world might like to use it. Hell, people would
probably pay for it since it would easily replace an entire YASNAC ERC unit.
the motor controllers we have are the CACR series as described here:
http://spaz.org/~jake/robot/CACR-SR-servopinouts.pdf
and they are pretty simple to control - you send them an analog voltage to tell
them which direction to turn, and how fast. You send an
analog voltage to say how much torque to exert (seperately for each direction),
and it sends out analog voltages telling you how fast it is going, and how much
torque it's exerting. And of course the encoders are always
telling you the actual position for all the axes so you can decide whether to make
changes to get what you want.
in the meantime though, we should keep playing with it as-is and see if we can
"hack" to get control of it without anything as radical as a
replacement controller.
-jake
---------- Forwarded message ----------
There are several models of MOTOMAN robot that can be controlled over serial port,
and some even supported by ROS Industrial community. If
your Robot controller is DX100, then you are very lucky.
However I think your robot is super old so probably not supported by ROS
Industrial. You can check it out here:
http://wiki.ros.org/Industrial/supported_hardware
Anyway, usually an industrial robot can be used in following ways:
Teach & Replay by teach panel
offline programming (generate a tool path offline and upload it, like 3D printer)
modern robot can be used in these ways, in addition to previous ones:
online program uploading (you generate end-effector/tool path and upload to the
robot control box via serial port/ethernet, and the robot
executed it right after it receives the path.
very few robot support this: real-time control of the robot joint, including
position, velocity, acceleration and so on. As far as I know,
only Universal Robot and maybe Baxter support this. Kuka/DLR's LBR(Light Weight
Robot) might support it too, but it's not for sale to general
public.
_______________________________________________
sudo-discuss mailing list
sudo-discuss(a)lists.sudoroom.org
https://lists.sudoroom.org/listinfo/sudo-discuss
--
Be seeing you.