Reinforcement Learning is a data-driven approach to learn intelligent behaviors through trial and error interaction with the environment. Related to robotics, it offers new chances for learning robot control under uncertainties for challenging robotic tasks.