An error occurred while establishing communication with the robot

[LeonRobot]

hi, i use the the OrebroUniversity/industrial_core, but when i run the roslaunch robot_interface.launch
it show the [ERROR] [1504762335.231789044]: Get unload pointer failed, buffer size: 0, smaller than byte size: 4
[ERROR] [1504762335.231843198]: Unload pointer returned NULL
[ERROR] [1504762335.231877206]: Failed to unload message joint: 8 from data[0]
[ERROR] [1504762335.231913226]: Failed to unload joint data
[ERROR] [1504762335.231945259]: Failed to initialize joint message

BUT if change from “ static const industrial::shared_types::shared_int MAX_NUM_JOINTS =20” to “static const industrial::shared_types::shared_int MAX_NUM_JOINTS =10” in simple_message/include/simple_message/joint_data.h , It works.

I want to know why the ERROR occur, and if I need to change other Settings.
Much appreciated!

[lz89]

I have the same issue.
I though the MAX_NUM_JOINTS should change to 20 for Yumi due to the discussion here
What I did was to clone the original industrial_core repo and check out the kinetic stable branch, then change the MAX_NUM_JOINTS from 10 to 20.
Then I get the error, but if I change back to 10, the robot_interface.launch under yumi_support folder works. It is very confusing for me to setup the RAPID and ROS side, the documentations provided by other forked repo are either outdated or has different user scenario.

[tstoyanov]

What branch of the yumi repo are you using? It may be that you have a different value for MAX_NUM_JOINTS on the ros side and the RAPID side. For a good overview on setting things up, I would also suggest checking through the wiki on the kth fork of this repo (https://github.com/kth-ros-pkg/yumi/wiki).

[gavanderhoorn]

@tstoyanov: the instructions on those pages contain references to the EGM interface driver. Afaik that is (still) not publicly available.

[lz89]

@tstoyanov I'm using the master branch for ROS side, but feature-egm branch for the RAPID side.
The reason I am doing this is that the RAPID code under yumi_hw (master branch) only have:
ROS_motion_left.mod and ROS_motion_right.mod for left/right arm. I don't know how should I load these to the tasks: T_ROB_L, T_ROB_R, ROS_MotionServer_L, ROS_MotionServer_R (or only one motion server as you mentioned in another issue that the yumi_hw uses only one motion server for both arm). I then follow the wiki in kth fork repo to load the modules to the server (link). BTW, I feel there are mistakes in the kth instruction. For example, they require the task set to Normal mode while keep the trust level to be SysFail which is not doable.

I did read through the wiki in the kth fork which to some extent is helpful but also it introduces some confusion as the instruction is not completed and aims for EGM.

I have also read through all the opened/closed issues in this repo as well as the ethz fork wiki (https://github.com/ethz-asl/yumi/wiki). I am still confused. So would you please give me some advises about:

1. If I don't have EGM and want to use the direct control (NOT trajectory download), shall I use the yumi_hw under master branch of this repo.
2. If I use yumi_hw master branch, how to link the mod files to the tasks, and what tasks shall I create? I know how to create task and load the modules in the controller etc.
3. If everything is setup at RAPID side (robot), what shall I run at ROS side (PC) in order to establish the communication? I tried to run roslaunch yumi_hw yumi_hw.launch, but I got an error:

ERROR: cannot launch node of type [yumi_hw/yumi_hw_ifce_simple]: can't locate node [yumi_hw_ifce_simple] in package [yumi_hw]

4. I assume I need to use the ros_control package (wiki.ros.org/ros_control) to control and communicate with the robot? (Sry that I am not very familiar with this package yet)
5. How do I get end-effector Cartesian control using ROS?

Thank you very much!

[lz89]

Regarding the MAX_NUM_JOINTS issue, I think I found the answer from this pull request: #19

[tstoyanov]

Hi @lz89 ,

1. If you want to do direct control and don't have EGM, you should look at the feature-ros-control branch. This does not offer MoveIt support: you would have to use ros_control controller instead (though it is possible to translate to action server, just needs some work)
2. In general, you should load the the ROS_common, ROS_messages and ROS_socket in all tasks (static). You should make a new task under either T_ROB_L or T_ROB_R that loads the ROS_StateServer. You should then load ROS_motion_left to the main task in ROB_L and conversely ROS_motion_right to ROB_R. Loading the gripper tasks should be identical to the instructions on the KTH wiki.
3. If everything is running smoothly, you should be able to roslaunch yumi_launch yumi_pos_control.launch. This should set up a position streaming interface. I have not tested this extensively. The yumi_vel_control.launch at the moment loads a custom controller that you probably do not have installed. If you modify it to load the joint_trajectory_vel_controller instead, this should give you a nice velocity control interface.
4. Yes. You would need to get some familiarity there.
5. That should be done by writing a cartesian ros_control controller. There might already be some examples of that available.

[LeonRobot]

@tstoyanov @lz89 @gavanderhoorn Thank you for your answers. I think this problem has been solved.