Visual posture estimation and control for redundant manipulator

The paper presents an approach to the posture control of redundant manipulator by using visual feedback. The redundant degrees-of-freedom enables several dexterous motion according to environmental information such as obstacle avoidance. In the paper, the hybrid motion controller including both the posture controller by visual feedback and the end-effector motion controller by using encoder signal is proposed. In the posture controller, the manipulator pose is estimated by particle filter from visual information. That means the posture control is completely realized only by vision sensor signal in our approach. The control model for obstacle avoidance in null space is also proposed by using the optical flow field which is detected by vision. The validity is evaluated by several experimental results.