Observability of Depth Estimation for a Hand-eye Robot System

This paper deals with the depth observability problem of a hand-eye robot system. In contrast to earlier works, this paper presents a complete study of this observability problem. The velocity of the active camera in the handeye robot system is considered as the input. The observability of depth estimation is then related to the velocity of the camera. A necessary and sufficient condition for the types of camera velocities necessary to ensure observability is found. This compensates for the results of earlier works, in which the velocity of camera was estimated. The theory is also verified by both simulations and experiments in this paper. Furthermore, a modified LQ visual servo control law is proposed to vary the weighting matrices so that depth estimation is improved while the level of control performance is still retained.