Error Handling WITHOUT using try-catch statments

Hello,

I am trying to convert some of my MATLAB functions in to a form that is compatible with Embedded MATLAB.

I have a try-catch statement that attempts to run a simulation (try) and if the simulation fails (catch) returns a zero. By doing this the function can deal with failed simulations without crashing.

Unfortunately embedded MATLAB does not allow try-catch statements.

Does anyone know of a way I can handles this error without using try-catch?

Thanks.

[Kaustubha Govind]

Do you mean that you are simulating another model from within an Embedded MATLAB Block in a Simulink model? Is there a reason you chose to use the Embedded MATLAB block as opposed to the Model reference block?

One potential solution would be to perform the try-catch in a separate MATLAB function, which is called from the Embedded MATLAB block after declaring it using eml.extrinsic (coder.extrinsic starting in R2011a).

[Will]

No I am writing an m-file which conforms to the Embedded MATLAB rules - then I am going to convert it to C-code. It will not be put inside a Simulink model.

I also know you cannot use the MATLAB 'sim' command so I am having to write my own C interface to replace the aspects of the m-file which use the sim command.

Answer 1

[Arnaud Miege]

I wouldn't recommend trying to run a Simulink model within an Embedded MATLAB Function, which itself resides in another Simulink model. It means that at each time step of the "caller" model, you're running a simulation of the "called" model. This is extremely inefficient as you could potentially have thousands of time steps. As suggested, a better solution would be to use a Model Reference to include the "called" model by reference into the "caller" model.

[Will]

I am not using a Simulink model within an Embedded MATLAB Function - I am simple writing an m-file which I intend to convert to C code.

The m-file calls the model and the model returns results which are then processed in the m-file.

I probably provided too much information: all I wanted to determine is if there is another way to handle errors (other than a try-catch statment)?

[Arnaud Miege]

That won't work. You can't generate C code from the sim command. These are the functions supported for code generation:

http://www.mathworks.com/help/releases/R2011a/toolbox/eml/ug/bq1h2z7-11.html

Using Simulink Coder, you can generate C code from the model itself.

[Will]

Hi Arnaud, thanks for your feedback.

I have already created C code for the Simulink model, imported into Visual Studio and verified it is running correctly. I know I can't use the 'sim' hence I am writing an interface for the model code in C.

Regarding the error handling, perhaps another question might lead us to a suitable answer:

If I provide a model in Simulink with a set of model parameters that cause the simulation to reach NaN at time x, the simualtion will stop and MATLAB will generate an error. Now, if I generate code for this model, provide it with the same parameters and run it in the generated C - what will happen? Surely somewhere - a flag is raised or a return statement is null or something is returned from the model_step function that I can use to determine that the simulation has failed?

Thanks.

[Arnaud Miege]

I'm not sure, sorry. One thing is with the generated C code, you only work one time step at a time, there's no notion of simulation time, so I'm not sure it makes sense to think of a simulation having failed.

[Will]

Ok. Well I will have a look through the generated code to see how/if it deals with NaNs at a given time step - if I can catch the NaNs in the generated code then essentially I can determine whether the simulation has failed.

Answer 2

[Walter Roberson]

The only error control mechanism supported by C is signals. I don't know if error conditions produce signals; I do not recall having heard anyone refer to that as a possibility.