Dr Petar Kormushev and Dr Sylvain Calinon from the Italian Institute of Technology (IIT) have developed a robot learning to flip pancakes. The skill is first demonstrated through kinesthetic teaching, where the user grasps and moves the robot to provide an initial example of flipping the pancake. During demonstration, the robot is gravity-compensated to move it easily as if it had no weight. The pancake flipping skill is then refined by reinforcement learning, which allows the robot to reproduce the task in different configurations and positions by extracting automatically the important characteristics of the skill. After 50 trials, the robot learns that the first part of the task requires a stiff movement to throw the pancake in the air and make it flip, while the second part requires the hand to be compliant in order to catch the pancake without having it bounce off the pan.