In this Chapter we address the “Sim-to-Real” effect, i.e. the different results that can be achieved during task execution, when learning is carried out in simulation. To reduce this effect, simulation must be as close to real as possible and therefore modeling and learning methods must be tuned to this need. In Sect. 1 we describe the software technology used, and in Sect. 2 we present the learning methods and their evaluation metrics. In Sect. 3 we present the results obtained in simulation, and in Sect. 4 we discuss the results obtained after sim-to-real transfer. Section 5 concludes this Chapter.