Re: CERTI-Devel Digest, Vol 23, Issue 1 and Issue 2

[Christian Stenzel]

Hello,

Eric wrote:

> I would say that the tutorial should primiraly
> "demonstrate usage of HLA RTI API"
> and specify clearly if the API used is:
>    HLA 1.3
>    and/or
>    IEEE-1516
>    and/or
>    CERTI specific
>   
You are right. Only for linguistic clearness: I
will write a tutorial demonstrating the usage of
HLA RTI API. The tutorial deals with the HLA 1.3
standard. If 1516 will be supported by CERTI
middleware I will extend the tutorial.

So it is of course an HLA tutorial with CERTI. I think,
that this is the aim of task 6911. Please correct me,
if I am wrong.

> I'm OK with the refined aims and find the idea of showing
> source code a good idea. Regarding this feature I think
> you may think of an optional "easily parsable output"
> --parsable-ouput command line options which generate
> outputs that may be captured and sent to a wrapping GUI.
> This is not a _really_ important feature but it could be nice
> with almost no more added work.
> I'll propose an output format and I may implement the wrapping GUI
> you may just think that I (or any other) should easilly be able
> to change the way your tutorial send output.
>
> Doxygen usage for documentation is good, I'll post a note
> on how to easily generate doxygen documentation using CMake
> (as it is [almost] done for CERTI).
>
> I'm Ok with the example application.
> One may add one or several
> "observer" federates which either store (to file)
> or plot the state var of your closed loop controller.
>
> Such "observer" federate may be made relatively generic.
I plan to have an implementation of both federates at the end
of the next week. At first I write the code without any
extra functionality. After that I add the proposed features step by step.

So everyone can examine the development of the code and give hints
in each iteration.

The proposal of Eric to implement an easily parseable output sounds
good. We will discuss it in detail in a few weeks, if the main part of
the code is accepted.

Now I will answer Pierres comments regarding Matlab-CERTI-bindings:
We developed a Matlab-RTI-Interface for DMSO RTI-1.3NG and Windows.
Last month I tried to port this binding to Linux. It's possible but
on modern Linux distributions it is tricky, because you need standard c++
libraries compatible to the DMSO binaries which are compiled with g++3.0
on RedHat7.2.

That's why I looked for an alternative RTI and found CERTI. Last week
I compiled the Matlab-Toolbox with CERTI using CMake.

I plan to put it under the LGPL, if anyone is interested. It would be
nice, if it can be combined with the CERTI tree in a subdirectory or 
something
similar. Yesterday, I spoke with the other developer of the toolbox. He
has no objections against my plans, but added, that the origin of the 
toolbox
should be visible. I think, a link from our university homepage to your
project homepage should be enough.

> > Matlab or Simulink ?
> > Matlab variables and HLA classes and objects linking ?
> > Relation between Simulink time and HLA time ?)

The toolbox realizes an RTI-Interface within Matlab. I suppose, that it is
easily extendable to simulink with the help of s-functions. We flated the
object orientation to a procedural mex-Interface. That's why one MAtlab 
instance can
only be one federate at time, this is the limitation. I suppose your 
third question deals with some
kinds of ODE-solver in simulink. I suppose that it is difficult to 
realize numerical integration methods
with adaptive step sizes (e.g. Runge Kutta 4/5). Fixed steps should be 
no problem.

I believe RTI bindings to Scilab and Octave can be relized in the same 
manner.

Another project I done for a company deals with HLA and Fortran90. You 
can use
the RTI-API in Fortran90. This is often required for companies which 
have large
simulation code written in Fortran. Reimplementation is often impossible.
I've done it for pitchRTI (1516). If there are some people interested in 
such a work
I could prepare Fortran90-CERTI-bindings and put it under the LGPL.

Perhaps these two interfaces (Matlab,F90) can increase the CERTI user 
community

So long
Christian