Problem with the use of “eig” in a Simulink Matlab function

Hi,

we are implementing a Simulink model to retrieve the components of Quaternions from attitude determination block. While using the command "eig" in a matlab function, in order to retrieve the eigenvector corresponding to the maximum eigenvalue, with this procedure:

function q_eigen = fcn(K)
  [q_e,lambda] = eig(K);
  [lambda_max, col] = max(max(lambda));
  q_eigen = q_e(:,col);
end

Simulink dispalys the error "Inferred complexity ("on") for data "q_eigen" does not match back propagated complexity ("off") from Simulink."

We have tried to put "real" command before q_e, but we obtain the same result.

[Walter Roberson]

Is it possible that any eigenvalue could be complex?

[Christine Tobler]

I don't know much about this, but the following older question seems related

https://www.mathworks.com/matlabcentral/answers/9838-simulink-complexity-propogation

and has some answers that to me seem promising to try out.

Using

q_eigen = real(q_eigen);

at the end of the function might remove this error, but unless you are very certain these eigenvectors are actually guaranteed to be real, it will mean returning a wrong result, since the imaginary component shouldn't just be discarded if it exists.

Answer 1

[Sulaymon Eshkabilov]

Maybe you can try max(max(real(lambda))) instead of real(q_e) to see your Simulink model simulation even though your eigenvalues are complex-valued.