Since we're on the topic, you can find some clips of some work I did way back in 1995 on robotic catching at http://www.mit.edu/nsl/www/. The arm can only catch underhand tossed objects. The trajectory is planned to match position and velocity with the object and then decelerate along a smooth path. This allows for a greater window of time for closing the hand. Removing the matching constraint allows the arm to catch faster objects, but then the limiting factor becomes the vision system (60fps) and the timing of hand closure. We also added in some aerodynamic predictors that let us catch paper airplanes, but those are much harder to throw back.
Slashdot Mirror
Since we're on the topic, you can find some clips of some work I did way back in 1995 on robotic catching at http://www.mit.edu/nsl/www/. The arm can only catch underhand tossed objects. The trajectory is planned to match position and velocity with the object and then decelerate along a smooth path. This allows for a greater window of time for closing the hand. Removing the matching constraint allows the arm to catch faster objects, but then the limiting factor becomes the vision system (60fps) and the timing of hand closure. We also added in some aerodynamic predictors that let us catch paper airplanes, but those are much harder to throw back.