How can I use fmincon for maximum likelihood optimization?

I am trying to use fmincon to obtain mle's for my model. Using the 'mle' function it gives me a solution, but the solution is not unique. I have therefore applied constraints to some of the parameters and tried to use 'fmincon' instead. But every time I run 'fmincon', with different starting values, it keeps giving me

                         Max     Line search  Directional  1st-order 
Iter F-count  f(x)   c'straint   steplength   derivative   optimality Procedure
  0      5  12059.9       0                                         
  1     10    NaN    9.095e-13          1     -1.34e+04     Inf   
  2     15    NaN        NaN            1          NaN      Inf     Hessian not

I'm not sure what's going wrong. I've included my code for the objective function, log-likelihood in this case, and the command lines I am using to run the optimization problem.

 function ln_like = Skellam_Log_likelihood(parameters)

D = importdata('E:\Matlab Data Files\Emprical Data\GoalDifference.dat','D');

     mu = parameters(1);
     h = parameters(2);
     a1 = parameters(3);
     a2 = parameters(4);

 for i = 1:1:length(D)

    pdf(i)  = Skellam_pdf(D(i),mu,h,a1,a2);

 end

    pdf = log(pdf);
    ln_like = -1*sum(pdf);

 end

 x0 = [1 1 .2 -.2];
 opts = optimset('Display','iter','Algorithm','active-set');
 [x fval] = fmincon(@Skellam_Log_likelihood,x0,[],[],[0 0 1 1],0,[],[],...        [],opts)