Revision history [back]

Yes, your concern is warranted: for each publisher-subscriber pair, a new TCP-connection is established. So the amount of bandwidth is proportional to the number of subscribers.

I can't think of any elegant solutions right now, my only ideas:

• Republish the data as you suggested. In fact, there already is a republish node for image transport. However, you'd probably have to rename the topics in your code.
• Try compressing your image topics by installing the required codecs on all machines and using the compressed streams (see known transport packages). Mayby this already saves enough bandwidth for doing the connection twice (which I doubt and doesn't scale further)?
• Use a powerful pc that can both control the robot and record data. Have one or both of the "operators" work remotely on this machine (ssh -X).