Optimizers avoiding local peaks

I have red this example.

Local peaks should be avoided and I know I can avoid
that by optimizing 2 parameters and show a 3d graph.

However, when there are much more parameters non-exhaustive
optimizers work quite good especially CMAE is very fast.

To avoid local peaks this is mentioned in the manual:

stop if adding 0.1 standard deviation vector in any principal axis direction does not change the value of objective value

Is this something that is integrated in the optimizer or should I check that manually
after an optimization?

Are there some parameters that can avoid local peaks that are not documented?
Which optimizer would be the best to avoid local peaks?

All logic is integrated in the CMA-ES, SPSO and Tribes engines. These algorithms are constructed so they do everything to avoid local peaks “by design”. By adding extra “runs” as explained in the manual, you can do few iterations that may get better results if your search space is very difficult.