2 p.m. Location: ECSS 3.503
In this talk, we will discuss the esign and control of an air hockey-playing robot that we built at the University of Illinois as a testbed to investigate the integration of a real-time computer vision, nonlinear control theory and machine learning. We will present some results on teh control of rigid objects through impacts using the so-called Routh two-dimensional impact model and derive the reachable set of puck velocities achievable through a single impact between the mallet controlled by the human player and the air hockey puck. These results have implications for both motion planning and control, not only for air hockey, but also in applications such as robot locomotion, non-prehensile manipulation, and robotic manufacturing.