universalrobots / universal_robots_ros2_driver Goto Github PK

Currently we use Denis' fork:

https://github.com/destogl/ur_msgs.git

We should get this merged back into https://github.com/ros-industrial/ur_msgs

Implemented by FZI

After setting the Speed Slider to 0 using the io controller service, the robot goes into protective stop giving the following warning:

As we work toward a minimal demonstration, it would help us to know if the RobotStateHelper is necessary. It looks like there are hardly any references to it in the package.

The reason it's somewhat difficult to port is a simple action client doesn't exist for ROS2 yet.

@fmauch @gavanderhoorn do you have any input?

Currently both the ROS1 and ROS2 driver repos will have the same URDF and mesh files. Investigate a 3rd repo which will house these and get pulled in by the driver packages.

Most of the pins seem to work except i cannot set the analog_out[1] current. Another problem i encountered is that after setting an input using the service call i wasn't able to change it in the teach panel. I think as the safety device the teach panel should always be able to override values we set using service calls.

Originally posted by @MGBla in #66 (comment)

I commented out the dependency on ur_client_library because it can not be found by rosdep. We are not exporting it properly. Opening follow-up issue for this.

Originally posted by @destogl in #71 (comment)

I am trying to build the driver per the instructions, but the vcs step fails because .repos.yaml does not exist in the src directory.

vcs import --skip-existing --input src/Universal_Robots_ROS2_Driver/.repos.yaml src
usage: vcs import [-h] [--input FILE_OR_URL] [--force] [--shallow]
[--recursive] [--retry N] [--skip-existing] [--debug] [-w N]
[--repos]
[path]
vcs import: error: argument --input: can't open 'src/Universal_Robots_ROS2_Driver/.repos.yaml': [Errno 2] No such file or directory: 'src/Universal_Robots_ROS2_Driver/.repos.yaml'

Thanks,

bb

This is just a quick fix for preventing unwanted behavior of the real robot at startup. The real fix will come in a later pull request.

Originally posted by @MGBla in #63 (comment)

FZI will provide input on milestone.

Joint position/velocity/acceleration will already be published on the joint_states topic, so we may not need this

#44

We should add a warning for the demo script to specify how the robot will. This way the user is aware of what will happen and can remove cables etc. that might be damaged by the motion

Currently the speed slider cannot be set to 100%. The controller seems to accept the input but on the call to the hardware interface something seems to go wrong. Thus the request is ignored

Need to investigate how to handle speed scaling.

Requires extension of JTC.

FZI

change hard-coded IP (make it a launch file parameter)

Without speed scaling

In ROS1, the speed scaling slider used a fraction (0.01-1.0, I believe)

On develop branch I get a bunch of build errors related to rt_active_goal_. It looks like it is not declared anywhere.

Some of the errors are:

scaled_joint_trajectory_controller.cpp:153:11: error: could not convert ‘((ur_controllers::ScaledJointTrajectoryController*)this)->ur_controllers::ScaledJointTrajectoryController::<anonymous>.joint_trajectory_controller::JointTrajectoryController::rt_active_goal_’ from ‘joint_trajectory_controller::JointTrajectoryController::RealtimeGoalHandleBuffer’ {aka ‘realtime_tools::RealtimeBuffer<std::shared_ptr<realtime_tools::RealtimeServerGoalHandle<control_msgs::action::FollowJointTrajectory> > >’} to ‘bool’
  153 |       if (rt_active_goal_)

scaled_joint_trajectory_controller.cpp:181:27: error: ‘using RealtimeGoalHandleBuffer = class realtime_tools::RealtimeBuffer<std::shared_ptr<realtime_tools::RealtimeServerGoalHandle<control_msgs::action::FollowJointTrajectory> > >’ {aka ‘class realtime_tools::RealtimeBuffer<std::shared_ptr<realtime_tools::RealtimeServerGoalHandle<control_msgs::action::FollowJointTrajectory> > >’} has no member named ‘reset’
  181 |           rt_active_goal_.reset();

Implement this: https://github.com/UniversalRobots/Universal_Robots_ROS_Driver/tree/master/ur_calibration

Per the conversation here.

Let's change the member function names to be snake_case rather than camelCase. I would do it in a new PR

The tests are currently not used in the CMakeLists.txt file.

should include speed scaling

Test tolerance failing of trajectory completion with small movements

Currently we are working on a fork of the Client Library repo with additional ROS2 changes. However, since the Client Library is plain c++ it should support both the ROS1 and ROS2 driver without additional branches.

Currently the driver only works using the external control ur_caps but not the remote control

• This is an issue in hardware interface
• Logic to the controller has to be checked

It should only update tool I/O when they are changed. Use doubles for beta release.

Decide how to publish additional info to ROS side, tool state, digital I/O, robot mode, safety mode, and tcp pose. Do we create controllers or to spin another thread to publish this stuff

We need to make this easy for other people to use, so let's add installation instructions to the README. It should include a link to MoveIt2 installation instructions, at least.

Hi,

since this is the repository with OSS GH-Actions can be used for free. They are somewhat more responsive and industrial_ci and ros-tooling WG provide configurations for it. The benefits are additional checks, e.g., linters, uncrustify, etc.

I have have also a "standard" setup used across ros-control WG. Should I add this here?

Depends on #16

We should discuss how to handle the goal_time_tolerance when the robot is scaled down.
Right now the scaling factor is not applied to the time tolerance, which will be a problem if a rather small tolerance is chosen and the robot scales its movement down. However from my point of view this is the desired behavior since the action fails if the user specified time tolerance is violated. But on the other hand the user might not be aware that the speed scaling will kick in which would lead to confusions why the action will return unsuccessful.

Improve capability to specify i/o at specific times in traj

These may be things you already know, but as someone trying this out, here are somethings that I encountered... Good news: I was able to successfully get things running. I was building within a docker container based on ros:foxy

1. I had to do a git clone https://... instead of the command in the doc (artifact of being in a docker environment?)
2. I needed to do an apt update before the rosdep
3. I need to build the client library. Initially, I cloned the official client library from UR, but that did not work at runtime (I can send a log file, but I suspect you know this already). I then tried the version of the client library on your site and everything worked.
4. In hardware setup, I think you mean to say that you can either switch to remote mode, or use an external node. As it is you say to switch to remote mode and then describe how to use the external node.
5. I manually had to start the controllers as was indicated in the doc. but then it all worked.
   bb

Document how to run with hardware

This is just a reminder to delete one or the other, because it's not clear which one is used at the moment