structs as input to simulink s-function blocks

Hi all,

as a Simulink newbie I'm at a loss here, perhaps someone can give me some advice.

---- lengthy background information ---->

I've inherited a tool consisting of a Simulink model and some Matlab functions it is called from.

The tool is for parametrizing a vehicle controller and uses Simulink functions like operspec, findop and linearize to find operating points and linearizations of a vehicle model.

While the Simulink *.mdl look fairly simple, the vehicle model is coded in C and is massively parametrized, parameters being held in nested structs. The model code (shared code with standalone C-simulation) is capsuled in C S-functions.

The whole thing is working well at the present but has one massive drawback: knowledge of the simulation parameters is shared between Simulink S-function blocks by global variables: For example the C-struct actuator_par is used by 2 S-function blocks, StepActuators and StepVehicle. Both S-functions are linked against a shared library which holds a global C-struct actuator_par. This variable is initialized by a mex-function, linked against the same shared library and called from Matlab before initializing the Simulink model.

---- lengthy background information ----<

What I'm trying to do now is to get rid of the global C-variable commonication between blocks.

I have recently capsuled all relevant functions of the C-simulation into mex-Files including the tedious work of translating between C-structs and equivalent Matlab-structs, in order to be able to use the very same C-simulation code for parameter optimization utilizing the optimization toolbox.

Thus I hoped, I could do something along these lines (example):

• add actuator_par as S-function parameter to the block StepActuators
• define the M-struct actuator_par using the appropriate Mex-capsuled initialization routine, thus creating a struct variable in the matlab workspace
• access this variable from within the S-function block and translate it into a C-struct just as I do within the mex capsule, perhaps just substituting ssGetSFcnParam(S, 0) for prhs[0].

I tried this and the S-function compiles without warnings, operspec complains justly when there is no variable actuator_par defined within the Matlab base workspace. However when I provide actuator_par I get a new set of error messages that seem very non-descriptive to me, for example:

Conversion to logical from struct is not possible.

Error in opcond.AbstractOperatingPoint/update (line 56)
    if RunParameterManager

Error in operspec (line 109)

So, before I try to figure out what this message means, can anybody tell me if my approach is even possible? I'm open to other suggestions, the main goal is to have all S-function blocks within the model have access to a set (constant) parameter C-structs and, if possible, a second set of (not constant) state C-structs.