Vision-based control of an air hockey playing robot

We demonstrate an integrated sensing and control design for high-speed ballistic manipulation in two dimensions. We show the effectiveness of a vision-based switched control scheme for precision impulsive manipulation. Vision-based hybrid control is a promising avenue of research for fully autonomous interaction between robots and complicated environments. Much work remains to be done in both computer vision and hybrid systems. In the context of robot air hockey, important future areas to investigate include the problems of modeling and identification of the environment dynamics, in particular, the table friction and puck spin. We show how the puck spin significantly influences its trajectory and subsequently our ability to perform accurate trajectory estimation and prediction.