Study of visual servo-performance of robots based on task function

To study the visual servo-performance of a robot on a common theoretical basis, a generalized visual servosystem model is proposed based on the task function method. The dynamic performance of the position-based visual servoing (PBVS) and the image-based visual servoing (IBVS) in the Cartesian space and the image spaces are investigated respectively based on this model. Simulation results show that both the PBVS method and the IBVS method are robust to the camera calibration errors; and are comparable in stability and convergence. However, their dynamic performances are greatly different. For the PBVS method, the shortest Cartesian trajectory can be achieved, but the camera field of view may be lost because the corresponding image trajectories are not controlled. On the other hand, for the IBVS method, the shortest image trajectory can be achieved, but the Cartesian trajectory will deviate from the shortest one due to the lack of direct control in Cartesian space. Especially, when the servo system performs a large rotation with respect to the camera frame, even the camera retreat may occur.