How to fix Failure in initial objective function evaluation

Hello, I'm trying to evaluate function

f_gamma_r_delta_v = @(x,beta) [...
  Px+Q*cos(phi0+beta)+x(2)*cos(gamma0+x(1));
  Q*sin(phi0+beta)+x(2)*sin(gamma0+x(1));
  S*cos(gamma0+x(1))+U*cos(delta0+x(3))+Wx;
  S*sin(gamma0+x(1))+U*sin(delta0+x(3))+x(4)];

using fsolve

[x,fval]=fsolve(f_gamma_r_delta_v,x0,options,beta);
      gamma(i)=x(1)*180/pi;
      R(i)=x(2);
      delta(i)=x(3)*180/pi;
      V(i)=x(4);
      x0=x;

but I can not fix the Failure in initial objective function evaluation although I've read some answered questions.

Error in fsolve (line 230)

Thank you very much for help.

Marek

Answer 1

[Walter Roberson]

You would get a failure if:

• Failure if any of the variables other than x in your f_gamma_r_delta_v are not scalars
• exception: if Q is a row vector and phi0+beta is a column vector of the same length, their algebraic product would be a scalar, which would work
• Special emphasis: you are going to have a failure if any of those variables are empty. This is a common failure mode when the variables turn out to be globals
• Failure if any of those variables are nan or inf
• Special emphasis: 0 times +/- inf is nan, which is a failure. 0 times inf is not 0!
• Sometimes rotation matrices are just inconsistent and you cannot solve them
• Except for the last couple of releases, failure if complex values result; the newest couple of releases can fsolve() on complex-valued expressions

I notice, by the way, that the code after that uses i as a vector index. Be careful: by default i is sqrt(-1) . To avoid confusion and make sure you do not accidentally index at sqrt(-1) it is better practice to avoid using i or j as variable names.

Answer 2

[Marek Majkut]

Thank you for your answer.

All of my variables are real scalars (I checked it in the Workspace).

First beta equals 0, but this should not be a problem.

Non of the variables is empty, nan or inf (checked in the Workspace).

My script is too long so I don't want to send it whole. I'll try to fix it, but thank you for your answer anyway.

I wish you a nice day.

Marek

[Walter Roberson]

You could run the code up to that point and save the variables in a .mat file and attach that with minimal code to reproduce the problem from there.

By the way, please see

http://www.mathworks.com/help/matlab/math/parameterizing-functions.html

You are using an undocumented feature of fsolve when you pass in beta the way you are.