(Dr. Harvey Lipkin, advisor)
"Calibration-Free Visual-Servo Robot Contol Using An Eye-in-Hand Camera Arrangement"
A previous method of vision-guided robot control has been developed which demonstrates convergent tracking of a moving target with static cameras in a model independent system. This method is extended to the problem of visual servoing for the moving camera case. This thesis presents a scheme for calibration-free visual servoing for an eye-in-hand camera tracking a moving target. A system that is model independent eliminates the need for calibration, which is inherently difficult in some applications, such as toxic areas, underwater, or in outer space. Robustness to system disturbances, which can be common in industrial settings, is also achieved. Using an eye-in-hand camera is beneficial in that it results in a closer view of the target, giving a clearer, higher resolution image of the target resulting in less error.
This thesis develops and demonstrates a recursive dynamic Gauss-Newton method for model independent, vision-guided robotic tracking control. This method does not require calibration of kinematic or camera models, nor does it require system parameters. Control is accomplished by a scheme which minimizes a nonlinear objective function by estimating the composite system Jacobian and target velocity components in quasi-Newton steps. A recursive method of joint update with an estimated error velocity correction technique demonstrates robust and convergent control for moving target, moving camera visual servoing for a variety of paths investigated.