Intelligent visual servo control of an air hockey playing robot

The primary focus of this dissertation is the use of intelligent control for ballistic manipulation of a sliding, planar object: the air hockey problem. Intelligent control of robotic manipulators is viewed as an integrated sensing, planning and control system based on the principles of hybrid or switched control. Multiple closed-loop control laws are designed for various system operating points and control objectives. It is demonstrated that sensing, planning and control should be carefully integrated to achieve a high level of performance for robotic systems.; The use of control theory and end-effector motion planning in visual feedback control (fundamental to the air hockey problem) is considered in detail. A provably stable camera-robot calibration scheme is derived using the tools of adaptive and nonlinear control, redundant manipulator control, and the principles of logic-based control switching. It is shown that an estimate of the accuracy of the camera-robot calibration leads to an estimation-space-based hybrid control scheme under which each controller inherits a set of achievable objectives for end-effector control, effectively constraining the motion planning problem.; The use of a planar robot manipulator for the air hockey problem allows for a task hierarchy under which end-effector trajectory tracking is the primary task and redundancy resolution is utilized to achieve secondary tasks. The set of secondary objective controllers consists of criteria optimizing controllers and constraint satisfaction controllers. The system framework is seen to include both stability-based switching laws and heuristic switching rules, forming a solid foundation for intelligent control. Experimental results are shown for a simplified version of the system framework.