multivariable non-linear constrained optimization

Hi, I have a function of two variables and I know the bounds for each variable but these bounds aren't tight enough. For example, let x be between 0 and 6 and y between 10 to 20.

function z = fun1( x , y )
if (x-3)^2+(y-15)^2 < 2 
  z=x+y-xy;
else
  z=NaN;
end
end

In this example, we can see that within my known bounds there is an area that the function isn't defined and (in the real problem.not this example) I don't know where this contour is laying.

I have the same problem with single variable function and there I use "fminbnd" which requires only bounds but in all the optimization functions I saw for multivariable optimization I need to input an initial point which I don't know in advance.

Is there a multivariable optimization function who doesn't require initial point?

Thanks, Y.