Re: EMC: comparison of RCS and OMAC/OSACA architectures

[Fred Proctor <frederick.proctor-at-nist.gov>]

Herman Bruyninckx wrote:

> When looking at the EMC web pages, I see mentions to both the OMAC and
> RCS architecturs. As far as I know, the RCS is based on Albus'
> hierarchical architecture, while the OMAC works towards a less
> hierarchical (``holonic'') structure, for example based on CORBA
> components.
> 
> I have a couple of ``strategic'' questions, because I am preparing an
> Open Source robot control software project, and would like to be as
> compatible to EMC as possible :-)

I've forwarded this to our OMAC rep, John Michaloski
(john.michaloski-at-nist.gov) for his answers. Here are my answers:

> 1. Aren't these two architectures (RCS vs. OMAC) mutually incompatible?

They're two different things. RCS is a reference model architecture for
building control systems. Two people who follow the RCS methodology to
build a CNC machine will likely design different systems. The motion
control hierarchy will probably be the same, but the design of an
automatic tool changer would probably have different box-ologies.
Certainly the messages will be similar but not identical, and you
couldn't plug-and-play components from one fellow's implementation with
another.

OMAC already prescribes the components and their interfaces, with the
goal of plug-and-play. The components are not arranged in an RCS
hierarchy, but are more holonic, as you noted. We had as much influence
on the OMAC spec as anyone else, a group that included some holonic
folks.

> 2. What is the current (or desired) evolution of EMC in this respect?

EMC is much more an OMAC-like thing than is RCS. EMC is an
implementation of a CNC, following the RCS methodology. It's way
different from OMAC: the message set is different, the semantics of
similar messages are different, etc. It's not plug-compatible with OMAC.
Also, OMAC is emphasizing COM/DCOM communication, or whatever Microsoft
is touting, while EMC is on the open source bandwagon.

> 3.  Is EMC working towards OMAC compliance or not?

No, not as far as any active work in the developer's group is concerned.
However, at NIST we have taken the motion control portion and wrapped it
in an OMAC tortilla. John Michaloski is the one to talk to about this.

> 4.  Is this OMAC compliance worth achieving in the first place, or not?

I think so, as far as plug-and-play interoperability of commercial CNC
components is concerned. This is important to NIST, and is why we're
working with OMAC. Ultimately I'd expect that the meat of the EMC code
would be built into OMAC-compliant components, depending on how
prevalent OMAC compliance becomes. For example, some open-source EMC
enthusiasts may decide to buy an OMAC-compliant GUI and plug it into
their open source EMC. The GUI would work with commercial controllers,
and also EMC.

--Fred