Adaptive Depth Estimation of Homography-Based Visual Servo with Finite-time Convergence

Although the homography-based visual servo (HBVS) has been studied for many decades, how to obtain the depth information with fast and accurate convergence performance has not been fully studied. In this paper, a finite-time (FT) adaptive depth estimation method is suggested for the manipulator regulation of the eye-in-hand (EIH) visual servo system. To realize this objective, a linearized camera model including depth information is first established. To obtain the depth, a set of auxiliary variables are constructed to extract the depth estimation error with a series of simple algebraic calculation. Then, inspired by the sliding mode control, a FT adaptive depth estimation algorithm is developed, in which the predictor and/or observer error are not required, but the depth estimation error is employed to drive the estimation toward to its true value within finite time. With the estimated depth, the HBVS controller of the EIH system can regulate the robot arm to the desired position with fast response. Finally, comparative simulation results are given to show the efficiency of the proposed method.