| Summary: | 碩士 === 國立成功大學 === 機械工程學系碩博士班 === 90 === Robot manipulators are mainly used to execute fixed and repetitive procedures or dangerous tasks in factories with static environments. In recent years, sensor-based robotic systems have been developed, to react appropriately to sudden environmental changes and adapt themselves quickly to new tasks. An active vision system is an effective sensory system, because it mimics human vision and allows for non-contact measurement of the environment using a servomechanism and one or two cameras.
This thesis proposes a coordinated control structure and a set of control strategies for an integrated robotic system, composed of a robotic binocular head and a robot manipulator, to recognize a target object and perform a grasping task. The proposed hand-eye coordination control structure for the integrated robotic system is a hybrid image-based/position-based look-and-move structure. Two image-based look-and-move control strategies are presented to enable the robotic head to saccade and fixate a target object. Three position-based look-and-move control strategies are presented for the manipulator to drive the end-effector into the field of view so that an endpoint closed-loop (ECL) system is formed, to enable the manipulator to approach the target object, and grasp it. Two off-line calibrations are required. One is the calibration of the camera’s intrinsic parameters, and the other is the proposed head/manipulator calibration, required by the control strategy for driving the end-effector into the field of view. Image processing algorithms are also proposed to increase the perception capability of rapidly detecting the location of a target object and the end-effector of the manipulator, such that the computation time for image processing is reduced.
Finally, three sets of experiments are conducted to verify the theoretical derivations and the performance of the system. Three objects, including a cube, a right-triangle pillar and a ball, are placed on a table, and one is selected as the target object in each set of experiments. The robotic head and manipulator coordinate to locate and grasp the target object, which is in the same position but at a different orientation in each experiment.
|
|---|
