Running V-Rep headless on CPU cluster

[tokyoghoul]

Hi,

I am trying to implement V-Rep in headless mode with ROS on a GPU cluster via ssh login. The server does not allow GUI.

Does someone have any recommendations?

I also noticed that the ROS nodes from a customization script do not start if you launch V-Rep from headless mode, has anyone had any experience with that?

Thank you for help in advance.

Best,
Francis

[coppelia]

Edit: make sure to first read this topic about headless mode, this thread is old.

Hello,

V-REP does currently not have a true headless mode. This means that if you start V-REP in headless mode, you will still need all the GUI and graphic card dependencies. You can/could however recompile V-REP in headless mode. In that case, the dependencies are not there anymore, but you will not be able to use vision sensors anymore unfortunately.

Which ROS interface are you using? The ROS Interface, or the old ROS plugin? are the customization script routines run?

Cheers

[tokyoghoul]

Hi,

Thank you for your feedback.

So firstly, I am using the ROS Interface. To test running V-Rep in headless mode with ROS. I create a scene (myscene.ttt) with just the ROSInterfaceHelper tool and start the simulation with ROSCORE running (RosInterface loads correctly), all the relevant topics show with "rostopic list". I then run "./vrep.sh -h myscene.ttt", the rosnode 'vrepinterface' shows but none of the topics are displayed? The customization script should run even if the simulation does not start right? However, I also run "./vrep,sh -h s10000 -q myscene.ttt" and there are still no rostopics.

Secondly, Unfortunately I am trying to implement a navigation learning task with visual sensor feedback. Do you have any advice for running V-Rep headless with a virtual screen such as Xpra or something? has this be known to work?

Thanks!

[fferri]

So firstly, I am using the ROS Interface. To test running V-Rep in headless mode with ROS. I create a scene (myscene.ttt) with just the ROSInterfaceHelper tool and start the simulation with ROSCORE running (RosInterface loads correctly), all the relevant topics show with "rostopic list". I then run "./vrep.sh -h myscene.ttt", the rosnode 'vrepinterface' shows but none of the topics are displayed? The customization script should run even if the simulation does not start right? However, I also run "./vrep,sh -h s10000 -q myscene.ttt" and there are still no rostopics.

Not sure if starting the simulation is required or not, but are you actually using "s10000" option without a leading "-" symbol? It should be -s10000. Please refer to V-REP Command line options.

[tokyoghoul]

Not sure if starting the simulation is required or not, but are you actually using "s10000" option without a leading "-" symbol? It should be -s10000. Please refer to V-REP Command line options.

Sorry that was a typing mistake, I used the correct syntax when executing the command.

[coppelia]

Can you make sure that the customization script routines are actually ran, by printing some text from them?

To your second question: we haven't tried this unfortunately. Have maybe also a look at this related topic.

Cheers

[tokyoghoul]

Thanks,

What is the command I should add to the ROSInterfaceHelper tool to print to the bash console to test that the routines run? (I am not super familiar with Lua in V-Rep)

Cheerz

[coppelia]

You can use:

• print("Hello"), which prints to the statusbar
• printToConsole("Hello"), which prints to the console

Cheers