Additional approach /solution for ROS master discovery / synchronization

[Alexander Tiderko]

Hi,

I published a possible solution to discover and synchronize the ROS masters. You can find it on:

http://www.ros.org/wiki/multi_master_fkie

For the usage you don't need to set any parameter or change your configuration compared to single Master solution, except for ROS_MASTER_URI.  (theoretically ;) ) 

In the next weeks I will upload an approach for a GUI to show / manage availbale ROS master...

Regards,

Alexander Tiderko

[Daniel Stonier]

Hi Alexander,

Seems like you have a few new ideas in there, including some I have on my todo!

I quickly scanned the code, but no luck running the examples (maybe due to running precise-fuerte). I was wondering if you might give a brief explanation of how it is working (much faster than us working through the code in detail). Some questions I have:

• If I read it right, you seem to be syncing entire ros trees from remote to local masters. Correct?
• Is there a significant benefit to multicasting the heartbeat signals?
• Is multicasting practical in most environments (I'm a bit ignorant about setting up multicasting)?
• You configure the multicast group uri by rosparam for all participants?
• How is your sync different from the original master_sync by willow?

Cheers,
Daniel.

[Alexander Tiderko]

Hallo Daniel,

Am Freitag, 25. Mai 2012 11:02:33 UTC+2 schrieb Daniel Stonier:

On 24 May 2012 14:58, Alexander Tiderko wrote:

Hi,

I published a possible solution to discover and synchronize the ROS masters. You can find it on:

http://www.ros.org/wiki/multi_master_fkie

For the usage you don't need to set any parameter or change your configuration compared to single Master solution, except for ROS_MASTER_URI.  (theoretically ;) ) 

In the next weeks I will upload an approach for a GUI to show / manage availbale ROS master...

Regards,

Alexander Tiderko

Hi Alexander,

Seems like you have a few new ideas in there, including some I have on my todo!

I quickly scanned the code, but no luck running the examples (maybe due to running precise-fuerte). I was wondering if you might give a brief explanation of how it is working (much faster than us working through the code in detail). Some questions I have:

I tested it on (oneiric-fuerte)... You have an error description for me?

• If I read it right, you seem to be syncing entire ros trees from remote to local masters. Correct?

Correct! 

• Is there a significant benefit to multicasting the heartbeat signals?

The alternative to multicast messages are static configuration or a central DNS server. A static configuration file is not really practicable. Futhermore the zeroconf uses multicast to exchange the information.

• Is multicasting practical in most environments (I'm a bit ignorant about setting up multicasting)?
• You configure the multicast group uri by rosparam for all participants?

You can configure the group by rosparam. The dafault value is: 226.0.0.0:11511

The the rosparam are ~mcast_group and ~mcast_port

• How is your sync different from the original master_sync by willow?

 willow's master_sync subscribes topics of local ROS master and register their by a foreign ROS master. 

 The rxgraph shows not fine ;)

-> this master_sync register the topics and services only on local ROS master. There are no subscribtions.

-> All topics and services are determine automatically. Disadvantage: since the ROS master provides no interface such subscribeParam an overhead accrues to determine new topics, services. This is done by master_discovery node.

Regards,

Alexander