PathPilot robot
Youtube
I would love to have pre-made package for robot retrofitting. Is it true that Tormach pushes everything back to open source?
Please Log in or Create an account to join the conversation.
I don't see a lot of development going on in LinuxCNC when it comes to robot manipulators. One important step was the switch-kins branch but that was not accepted to merge into the master branch so you have to build your own. Unless there is a mechanism to switch between joint and cartesian kinematics in gcode it's fairly pointless to use linuxcnc with a serial manipulator. But, as machinekit, has shown using ROS to do the math is one way to do it.
[edit]
One reason that robot development is not really pursued here is that the manipulators themselves are expensive even without the controllers and the fact that they are not really all that useful for a hobbyist or small shop owner. Those things are built for _very_ repetitive tasks and having a CNC lathe or mill or plasma router will get you a _lot_ more return for the investment than a robot.
And yes I have TWO of them.
To play with.
Please Log in or Create an account to join the conversation.
- tommylight
- Online
- Moderator
- Posts: 19188
- Thank you received: 6430
Thanks.
Please Log in or Create an account to join the conversation.
Using vanilla python for programming the robot arm could be huge. If you could use all Python libraries from pip. Integrating tensorflow directly from operator console has great potential. Programming ROS nodes is not easy, this looks easy.
Please Log in or Create an account to join the conversation.
How exactly machinekit has plugged ROS and Linuxcnc together I don't know but really all you have to do is communicate the joint positions to LinuxCNC and possibly send the encoder positions back to ROS.
_Very_ true and a fact that many people don't realize, including me, until they get a chance to actually work or play with one. Just the fact that the kinematic itself comes with major stumbling blocks like singular points where the inverse kinematics math just folds is a very novel thing to have to work with.Industrial manipulators and CNC mills are very different beasts
user Grotius is using the KDL library to build a application that plugs into Linuxcnc.
forum.linuxcnc.org/38-general-linuxcnc-q...-c-c?start=80#191018
So since you have experience with ROS, can ROS or the Kinematic libraries do Tool Coordinate Moves?
forum.linuxcnc.org/10-advanced-configura...ics-to-robots#159066
Please Log in or Create an account to join the conversation.
People are using Arduinos to connect small tabletop robots to ROS ecosystem. This is using Ethercat. You can buy servo with ethercat from China for 500 dollars. Big brand robot controllers are whole package and expensive. I don't think company spending all the dough on state of the art Kuka or ABB will want ROS.There are drivers to connect ROS to various Robot controllers.
Above my pay grade. Machinekit is wrapper around Linuxcnc?How exactly machinekit has plugged ROS and Linuxcnc together I don't know but really all you have to do is communicate the joint positions to LinuxCNC and possibly send the encoder positions back to ROS.
So since you have experience with ROS, can ROS or the Kinematic libraries do Tool Coordinate Moves?
ROS path planning works differently to LinuxCNC path planning. You don't have a realtime motion control but buffer full of points in space and time. And you have more than one path planning algorithms at your disposal. You can create as many points as you wish but you don't have folly analytical path description.
Please Log in or Create an account to join the conversation.
Tormach is unveiling a new robot arm based on ROS+Machinekit:
I would love to have pre-made package for robot retrofitting. Is it true that Tormach pushes everything back to open source?
The look ahead trajectory planner was pushed back to linuxcnc.
fanuc subroutines are probably from Tormach.
We surely haven't incorporated everything they offer - there is no official process.
And the longer we are forked the harder it is to incorporate code.
If they are using Machinekit then that pretty much ends what we can use (IMHO)
i had voiced the opinion that we should have considered using their HAL code, but that was not popular among the few developers we have.
Linuxcnc can run robots directly but not in anyway you would usually run them for industry and it's not a focus of any real development.
Interfacing to ROS was a smart solution.
I would bet that someone could write code to connect our HAL to ROS.
But i still think it would be smarter to get out motion controller use their HAL code.
I think running real robots (indirectly) with linuxcnc is an exciting development - congrats to Machinekit and Tormach!
Please Log in or Create an account to join the conversation.
- tommylight
- Online
- Moderator
- Posts: 19188
- Thank you received: 6430
Having worked with industrial Kuka robots, the amount of parameters and settings and constraints and interlocks implemented is astonishing.
So being able to control such machine with LinuxCNC with a week or two of effort is also astonishing, and is a testament to the flexibility of LinuxCNC.
Of course it does not have everything the industrial ones have, but industrial ones do not have everything sorted out either, check "singularity" as it is still an issue with them.
Please Log in or Create an account to join the conversation.
- thefabricator03
- Offline
- Platinum Member
- Posts: 1130
- Thank you received: 533
I would bet that someone could write code to connect our HAL to ROS.
I think someone already has, github.com/tomlarkworthy/linuxCNC_ROS
Please Log in or Create an account to join the conversation.