Definite integral in (embedded) Matlab function, with passing additional parameters to the integrand

Hi,

I have a simulation file to model the dynamic behaviour of system, at each time step , I need to evaluate a time-independent integral. To do so I use (embedded) Matlab function but I got many errors. Any help would be appreciated.

inside my (embedded) Matlab function is:

%%________________ method one

function forces=my_embeded_func(positions)

y1=position(1);

y2=position(2);

forces=integral(@(theta),y1+y2*theta,0,2*pi)

end

%_______________

Notes: 1-my integrand is actually very complicated but I put something simple here.

2-I have been told to use "quadgk" instead of "inregra;", but it did not worked

3- I also tried the following method but it did not helped!

%%________________ method two

function forces=my_embeded_func(positions)

    coder.extrinsic('myFunc')

    y1=position(1);

    y2=position(2);

    forces=myFunc(y1,y2)
end

% where

function a=myFunc(y1,y2)

    a=quadgk(@(theta),y1+y2*theta,0,2*pi)

end

%_______________

Thanks,

Ehsan

Answer 1

[Mike Hosea]

If you don't need to generate code for your model, I think you can use your second approach, but you need to fix a few unrelated errors. Put myFunc into a MATLAB file myFunc.m. There is a syntax error. Remove the comma between @(theta) and y1:

 function a=myFunc(y1,y2)
    a=quadgk(@(theta)y1+y2*theta,0,2*pi)
 end

Now let's consider your block code. I didn't know whether you meant for your variable to be called "positions" or "position". I just picked one. Second, to call an extrinsic function, you have to pre-allocate the output to tell the compiler how to interpret the mxArray that comes back from MATLAB. The block code should look something like this:

function forces=my_embeded_func(position)
    coder.extrinsic('myFunc')
    y1=position(1);
    y2=position(2);
    % Define the output type for myFunc, in this case a scalar double.
    forces=0; 
    % Evaluate myFunc in MATLAB.
    forces=myFunc(y1,y2)
end

If you do need to generate code, the way I tell people to do this is to use persistent variables to store the parameters of the function. The whole thing could go into the block.

function forces=my_embeded_func(position)
  y1=position(1);
  y2=position(2);
  forces=myFunc(y1,y2);
end

function a=myFunc(y1,y2)
  integrand('set positions',y1,y2);
  a=quadgk(@integrand,0,2*pi);
end

function v = integrand(theta,y1set,y2set)
  persistent y1 y2
  if isempty(y1)
    y1 = 0;
    y2 = 0;
  end
  % More than one input means we are setting the persistent
  % parameter values, not evaluating the integrand function 
  % on this particular call. The theta argument is ignored.
  % We make it a string 'set positions' in the calling code
  % solely for readability.
  if nargin > 1
    y1 = y1set;
    y2 = y2set;
    % nargout had better be zero, since we haven't set v.
    return
  end
  v = y1 + y2*theta;
end

[Ehsan Asadi]

Thanks Mike,

It worked perfectly, just a quick question: With respect to simulation speed, which one is faster if I have to do this integral multiple time (in a loop) in each iteration?

I have a relatively complicated integrand which make the computing time very important for me.

Thanks you,

Ehsan

[Mike Hosea]

It could go either way depending on the integrand, but my guess would be the persistent approach rather than the extrinsic.