What does this other Fmincon error mean?

Hi there

I'm doing a constrained optimization problem where I first generate a few random intial points, I then only keep the feasible ones. I do this by running the objective function to see if it returns a NaN. NaN is returned when the

Then I use the feasible set of initial values and I call fmincon in a FOR loop. I am not using parallel computing toolbox.

However, I sometimes receive an error after fmincon begins from a feasible point.

Here is the output:

    Error using barrier (line 143)
Finite difference derivatives undefined at initial point. Fmincon cannot continue.

Error in fmincon (line 841)
  [X,FVAL,EXITFLAG,OUTPUT,LAMBDA,GRAD,HESSIAN] = barrier(funfcn,X,A,B,Aeq,Beq,l,u,confcn,options.HessFcn, ...

Error in FreeTailObjConstr (line 74)
[x,fval,exitflag] = fmincon(fun,x0,Aineq,Bineq,Aeq,Beq,Lwr,Upr,cfun,options,param);

Error in multiStartFreeTail (line 120)
      [x,fval,exitflag] = FreeTailObjConstr(x0, UB, LB, muVal);
end

What does this mean? Is it happening because even though the first point is feasible, the next point fmincon attempts returns is NaN?

Any help would be appreciated.

Cheers

Amir

[Nayan Rawat]

I have a similar problem. Any solutions?

Answer 1

[Alan Weiss]

fmincon estimates derivatives by taking finite difference steps. Your initial point was feasible, but once fmincon took a tiny step to estimate a derivative, the objective was no longer feasible. fmincon cannot recover from this sort of error at the initial point, though it can recover if the error occurs during subsequent iterations.

Alan Weiss

MATLAB mathematical toolbox documentation

[Amir Patel]

Thanks for clearing that up Alan.

To get around the NaN error, I tried setting a cost to a large value before the Simulink integrator's break.

But now, fmincon is struggling to converge...

Any advice on how to tackle this sort of problem?