A colleague just told me it would be sensible to comment the diagram I
posted, as it could be confusing otherwise. It's illustrating two ways of
interfacing the control subsystem (one or more SOMANET Nodes) with ROS:

1. The SOMANET User App written using rosxmos can communicate directly with
other nodes anywhere in the system using ROS communication techniques,
whether through Ethernet or Ethernet over EtherCAT (EoE).

2. The SOMANET User App written without rosxmos needs a conventional ROS
driver node to bridge communication, e.g. to EtherCAT CoE or some VoE based
protocol (similar to pr2_controllers). 

Both are displayed in the same diagram to show that real-time control
commands, if required, still need to be sent in a different way, as ROS
messaging isn't really rt capable. Anyhow the idea is to get rid of rt
communication between the PC running ROS and the control subsystem, by
putting everything that requires real-time behavior down to the
sensor/actuator level and only have intelligence apps left on the PC (what
ROS is good for) that send higher level, non-timing-critical commands to the
distributed control nodes (like a path to move the end effector along).

Nik

-----Ursprüngliche Nachricht-----
Von: ros-users-bounces@code.ros.org [mailto:ros-users-bounces@code.ros.org]
Im Auftrag von Nikolai Ensslen
Gesendet: Dienstag, 12. Juni 2012 13:46
An: ros-users@code.ros.org
Betreff: Re: [ros-users] is a native ROS device possible?

See this diagram for illustration:
http://www.synapticon.com/wp-content/uploads/2012/06/ROS-SOMANET-2.png