Automatic calibration of an air hockey robot

In this work, we propose an algorithm which enables an air hockey playing robot to quickly adapt itself to new environments. There are a number of parameters in each environment which affect the puck's motion (e.g. coefficients of restitution). The robot should be able to estimate these parameters accurately enough in order to predict the position of the puck and to reach its goals. Our algorithm does this task well without need of a high speed camera and in a noisy environment. We implemented this algorithm on an air hockey simulator. It successfully tunes the parameters with the average number of 17.12 shots and the average error of 1.39% in its shot planning after the calibration process is done.