collaboration with grl about iiwa_stack HOT 8 CLOSED

Hi Andrew,

thanks a lot for the interest, we heard already about your nice work from our colleagues at JHU.

I come right away at the point about the KONI port : by design that port is meant to be used only for FRI, correct. Without any modification that port is handled directly from the Real Time OS on board of the KUKA controller, the Java calls for FRI (with that .setOverlay on a motion) do the magic so that commands send though FRI are executed. But that also means, to my knowledge, than you cannot send other than FRI messages using that port, as we don't have direct access to the RTOS.

The modification that we describe in our wiki, basically assigns the KONI port to the Windows7 Embedded on the cabinet, so that it is accessible from Java. This change comes from the original work of Stanford : https://bitbucket.org/khansari/iiwa that we initially followed. But in theory is not necessary, as one could simply use the same ethernet port used for Sunrise Workbench. Also, FRI allows commands at 1ms because it is linked to the RTOS, with that change we lose that too.

At the moment, we use the KONI port for commanding the robot using KUKA's ServoMotion and for a dynamical configuration of the system (most of that is on our development branch). Our plan in the near future is to do all this using the Sunrise port, reenable the KONI port to be used with FRI and so allow a user to choose between different command modes.

As a matter of principle, of course we are open to any collaboration and any help or suggestion is welcome. But in the specific case, it is hard for us to find a submodule that could be neatly shared through a common interface between our two different approaches. We plan to continue using ROS topics and services, and in future actions, for communication because their expressiveness. Apart from that, the rest is actually just a thin wrapper around the Sunrise APIs, so providing an interface on top of it would be a huge overhead with little benefits.

On MoveIt! and Gazebo, if I remember correctly you use a V-REP model of the robot, while those two modules need an URDF model to work. If you plan to use MoveIt! it might make sense to switch to that representation, I also don't have knowledge of the accuracy of the V-REP model to be honest (our URDF model leads to a forward kinematics that coincides to that from Sunrise up to a micron).

Any idea about what we could share?

from iiwa_stack.

@SalvoVirga Based on our conversation via video conference I wanted to make some notes:

• You can at least update the license file for BSD (hopefully today if possible?) and just provide written comment here saying it is ok.
• grl can use the ros configuration, URDF, etc from this repository
• I'll aim to provide the same ros APIs (topics, etc) you are using in C++
• I'll try to integrate your functionality where appropriate
• We will discuss further integration between our tools after your March 1 deadline

grl main kuka driver code and locations:

• Kuka C++ driver classes
  • Independent i.e. no V-REP
  • smart servo
  • FRI
  • configuration
• ROS driver
  • in progress
  • depends on kuka drivers
• V-REP drivers
  • Kuka V-REP class depends kuka drivers
  • Includes hand eye calibration from camodocal

All that sound good? Thanks for the presentation it was extremely informative!

from iiwa_stack.

The code there isn't quite the latest, I'm in the midst of integration and I push changes out as it stabilizes and makes sense. I've got a larger change with support for full configuration of many system components coming down the pipe.

from iiwa_stack.

@ahundt Changed the main LICENSE file to BSD in the master branch, I will modify the single modules later on. You can consider the whole project under BSD from now on.

from iiwa_stack.

Thanks!

from iiwa_stack.

@SalvoVirga Now that some key deadlines are past I was hoping to follow up on this if you're available. Feel free to send an email to [email protected] and we can set something up.

from iiwa_stack.

Since such a topic is better suited to a private conversation, I am closing this issue.

from iiwa_stack.

@marcoesposito1988 @SalvoVirga I've released grl 3.0.0 which includes partial integration with iiwa_stack! If you are interested in adding FRI support to iiwa_stack I provide classes designed specifically for use by other library authors. I'm also using an approach for communication with the controller that avoids #7.

from iiwa_stack.