In the paper, a vision based assistant technique for underwater manipulations is proposed. Especially, 3D target position estimation issue and inverse kinematics issue are addressed. Regarding target position estimation issue, a technique which can estimate 3D position of the objects utilizing 2D position information on the two images from cameras deployed in stereo is addressed. Via touch screen, the 2D position information is fed by operators for guaranteeing reliability of the estimation results, which is one of the most important points in the underwater interventions. Regarding kinematics issues, a kinematics solution based on the weighted damped least squares is investigated to overcome the limitations: workspace limit and singularity, unintended changes of the robot configurations, and joint position limits. By designing appropriate cost functions for each limitation, the proposed kinematic solution provides a solution optimized in the manner of weighted norm, even when the manipulators meet those limitations. Through the laboratory experiments, the performance of the proposed assistant technique is verified.
展开▼