VisionBased Objective Selection for Robust Ballistic Manipulation
2009
The use of visual measurement for high-speed and high-precision robotic tasks requires careful analysis of the fundamental properties of the sensing paradigms chosen. In this work we analyze the information generated by a state estimator for a target moving object that is to be intercepted by a robotic manipulator. Selection of potential interaction objectives is carried out through analysis of the data stream from the visual system. The underlying sensing problem is designed for a principally unstructured environment under fixed-camera visual measurement. The test domain is chosen to be ballistic manipulation of a sliding object on a low-friction planar surface: the air hockey problem.