This paper presents the results of the application of a new process co
ntrol technique, the Learning Control, to a mineral processing operati
on. A hierarchical system of learning automata is used as a model of t
he controller. An empirical simulator capable of reproducing the dynam
ic of the autogenous grinding process is considered as the random envi
ronment in which the hierarchical system of automata operates. A proba
bility distribution is associated to the manipulated variable. This di
stribution is continuously adjusted by the learning system using a rei
nforcement scheme. Numerical results have demonstrated its control pro
perties, transparent tuning and robustness, while requiring minimal co
mputational load.