[EN] The aim of this master thesis is to study the state of art of reinforment learning, particularly those based on policy search methods and to apply such techniques to a 3DOFs inverted pendulum mechanism. The controller ...