3D object tracking on active stereo vision robot

The paper describes a control method for visual tracking on the active stereo vision robot. The robot in unique structure has three joints; one is for base rotation and the others for each camera rotation, i.e. neck and eyes. As the system has a redundancy in the meaning of degrees-of-freedom, the novel control approach is introduced and the high performance of the tracking is achieved. The control law is based on both of the image-base control and mechanical impedance control, referring the motion of the target object, and the time delay by image processing is compensated with the prediction by Kalman filter. Consequently, the mooth motion of the robot and the precise position estimation of the moving object in three-dimensional workspace is realized in real time. The validity of the proposed method is confirmed in the physical experimental results.