Thread Subject: Integrating compiled m-file with compiled simulink (RTW?) model

Hi,

I have a series of Matlab GUIs with different
functionalities (database, visualisation, excel data
transfer). All of these compile straightforwardly with the
Matlab Compiler 4.3.

In one of the GUIs I have a button that initiates the
simulation of a simulink model and as
functions 'sim', 'simset' etc, cannot be compiled in
matlab, I have turned to Real Time Workshop.

However, before I begin to understand how to create stand-
alone applications of simulink models via RTW, I would like
to know if there is a way to integrate the command to run
this model into the compiled GUI executable.

i.e. the code [T,X,Y]=sim('mymodel'...)

will not compile using mcc.

If I manage to get a stand-alone application
for 'mymodel'...how do I replace the above code into the m-
file/executable.

Best regards,

Matt

"Matthew Wade" <cosmicspacepig@googlemail.com> schrieb im Newsbeitrag
news:g4vq6l$8lr$1@fred.mathworks.com...
> Hi,
>
> I have a series of Matlab GUIs with different
> functionalities (database, visualisation, excel data
> transfer). All of these compile straightforwardly with the
> Matlab Compiler 4.3.
>
> In one of the GUIs I have a button that initiates the
> simulation of a simulink model and as
> functions 'sim', 'simset' etc, cannot be compiled in
> matlab, I have turned to Real Time Workshop.
>
> However, before I begin to understand how to create stand-
> alone applications of simulink models via RTW, I would like
> to know if there is a way to integrate the command to run
> this model into the compiled GUI executable.
>
> i.e. the code [T,X,Y]=sim('mymodel'...)
>
> will not compile using mcc.
>
> If I manage to get a stand-alone application
> for 'mymodel'...how do I replace the above code into the m-
> file/executable.
>
> Best regards,
>
> Matt
>
>
>

Hi Matt,
you will need to use the Rapid-Simulation target for the Real-Time Workshop
compilation. Take a look at the function rsimgetrtp for collecting the
parameters, and esp. the option "AddTunableParamInfo" . This function itself
is not compilable, but once you have the parameter structure, you can update
it from your code.

The code will be replaced by something like

paramStruct = load('paramfile.mat');
paramStruct.parameters.paramA = 42;
save paramfile paramStruct
system('theCompiledModel.exe -p paramfilein=paramfile');
result = load('output.mat');
plot(result.rt_tout, result.rt_yout)

There are some demos under Real-Time Workshop->Accelerated Simulations which
should give additional help.

Titus

"Titus" <titus.edelhofer@mathworks.de> wrote in message >
Hi Matt,
> you will need to use the Rapid-Simulation target for the
Real-Time Workshop
> compilation. Take a look at the function rsimgetrtp for
collecting the
> parameters, and esp. the option "AddTunableParamInfo" .
This function itself
> is not compilable, but once you have the parameter
structure, you can update
> it from your code.
>
> The code will be replaced by something like
>
> paramStruct = load('paramfile.mat');
> paramStruct.parameters.paramA = 42;
> save paramfile paramStruct
> system('theCompiledModel.exe -p paramfilein=paramfile');
> result = load('output.mat');
> plot(result.rt_tout, result.rt_yout)
>
> There are some demos under Real-Time Workshop-
>Accelerated Simulations which
> should give additional help.
>
> Titus
>
Thanks Titus,

At least I now know it can be done. But I first must learn
RTW before getting to that stage.

Best regards
Matt

Let me describe the problem that I have now:

What I have done:

a). In my m-file code, I have a function that initiates all
the states/parameters in the simulink model. These are
taken from an Excel file, which calculates certain
parameters for the model via a macro.

e.g. for a reactor block I have GENERAL, which is a vector
comprising a number of states (General(1) = initial
(1)...General(n) = initial (n) ), where initial is a column
of calculated initial states from the Excel file.

If the user selects a different option, these states will
change in the excel file and I will get a new General
vector.

b) I called all initial parameters to the workspace, open
the simulink model and selected:

- Inlined Parameters - Configuration (all parameters
selected from the initialising function (e.g. called from
Excel file). I did not change the Storage Class.

- Rapid Simulation System Target and fixed-step solver.
(the model runs on variable-step solver, so this may be a
problem?)

- Build: The Simulink Model is successfully compiled

b). I used the following code on the command line:

rtP = rsimgetrtp('model','AddTunableParamInfo','on');

the rtP.parameters.map is populated with the state vectors
(e.g. GENERAL) etc.

saved myrtp.mat rtP

c). from the command line:

!model.exe - p myrtp.mat -s 0.1

where -s is the simulation time.

Now here are the problems:

1. The simulation time -s, should be 25 (as dictated by the
user), if I use 25, there is a memory error message. I can
only run the model using 0.1 or lower.

2. The output from the model (model.mat) does not
correspond to the output I get when I run the model from
simulink (negative numbers, unstable).

3. in the Parameter file (rtP), I am unsure how to update
the parameters - there is no indication where I replace,
e.g. General(1)_0 with General(1)_1 (or General_0 with
General_1).

I am sure I am missing a few steps here, but as I have not
done this before, I would appreciate any ideas.

Best regards,
Matt

"Matthew Wade" <cosmicspacepig@googlemail.com> schrieb im Newsbeitrag
news:g6kghq$4he$1@fred.mathworks.com...
> Let me describe the problem that I have now:
>
> What I have done:
>
> a). In my m-file code, I have a function that initiates all
> the states/parameters in the simulink model. These are
> taken from an Excel file, which calculates certain
> parameters for the model via a macro.
>
> e.g. for a reactor block I have GENERAL, which is a vector
> comprising a number of states (General(1) = initial
> (1)...General(n) = initial (n) ), where initial is a column
> of calculated initial states from the Excel file.
>
> If the user selects a different option, these states will
> change in the excel file and I will get a new General
> vector.
>
> b) I called all initial parameters to the workspace, open
> the simulink model and selected:
>
> - Inlined Parameters - Configuration (all parameters
> selected from the initialising function (e.g. called from
> Excel file). I did not change the Storage Class.
>
> - Rapid Simulation System Target and fixed-step solver.
> (the model runs on variable-step solver, so this may be a
> problem?)
>
> - Build: The Simulink Model is successfully compiled
>
> b). I used the following code on the command line:
>
> rtP = rsimgetrtp('model','AddTunableParamInfo','on');
>
> the rtP.parameters.map is populated with the state vectors
> (e.g. GENERAL) etc.
>
> saved myrtp.mat rtP
>
> c). from the command line:
>
> !model.exe - p myrtp.mat -s 0.1
>
> where -s is the simulation time.
>
> Now here are the problems:
>
> 1. The simulation time -s, should be 25 (as dictated by the
> user), if I use 25, there is a memory error message. I can
> only run the model using 0.1 or lower.
>
> 2. The output from the model (model.mat) does not
> correspond to the output I get when I run the model from
> simulink (negative numbers, unstable).
>
> 3. in the Parameter file (rtP), I am unsure how to update
> the parameters - there is no indication where I replace,
> e.g. General(1)_0 with General(1)_1 (or General_0 with
> General_1).
>
> I am sure I am missing a few steps here, but as I have not
> done this before, I would appreciate any ideas.
>
> Best regards,
> Matt

Hi Matt,

w.r.t. 2): this might correspond to you using the fixed step solver for the
compiled application and the variable step solver (with e.g. Zero-Crossing
detection) in the Simulink program. Try to use the fixed step solver in
Simulink and see, if you get the same results/problems.

Titus

Hi Titus,

The model will not run with a fixed-step solver from
Simulink, I have to use Variable-step.

I have noticed -S option to allow specification of Solver
type, parameters. Maybe this will work.

Cheers
Matt

> Hi Matt,
>
> w.r.t. 2): this might correspond to you using the fixed
step solver for the
> compiled application and the variable step solver (with
e.g. Zero-Crossing
> detection) in the Simulink program. Try to use the fixed
step solver in
> Simulink and see, if you get the same results/problems.
>
> Titus
>
>

the -S option with a solveroptions.mat file does not seem
to do anything, even when I save solveroptions.mat with
nonsense parameters (e.g. solver 'null'). Running the
executable does not seem to take these options into
consideration - the output is still the same.

"Matthew Wade" <cosmicspacepig@googlemail.com> wrote in
message <g6ki01$mle$1@fred.mathworks.com>...
> Hi Titus,
>
> The model will not run with a fixed-step solver from
> Simulink, I have to use Variable-step.
>
> I have noticed -S option to allow specification of Solver
> type, parameters. Maybe this will work.
>
> Cheers
> Matt
>
> > Hi Matt,
> >
> > w.r.t. 2): this might correspond to you using the fixed
> step solver for the
> > compiled application and the variable step solver (with
> e.g. Zero-Crossing
> > detection) in the Simulink program. Try to use the
fixed
> step solver in
> > Simulink and see, if you get the same results/problems.
> >
> > Titus
> >
> >
>