I use fmincon to solve an optimization problem. However, as fig.1 shows, at first, the objective value increases and the first-order optimality increases, which indicates that fmincon does not progress. After around 500 iterations, fmincon begins to progress (decreased objective value, decreased first-order optimality and remaining feasible), but the converge rate is rather low, just refer to the figure 2, it does not converge after 2000 iterations. I wonder if anyone has come up with these situations, and how do you solve it?
Thanks a lot!