Lyapunov-stable eye-in-hand kinematic visual servoing with unstructured static feature points

In this paper, we address the visual servoing problem of robot manipulators with eye-in-hand cameras. To servo-control the image position of a feature point, traditional image-based controllers require the computation of the point's position vector with respect to the camera's frame. However, when the point's location is uncertain, the stability of traditional visual servoing controllers can not be rigorously guaranteed. To contribute to this problem, in this paper we present two new kinematic image-based controllers that do not require the exact location of static features. The first controller is a depth-free method that uses the camera's calibration matrix and visual feedback to compute a quasi-position vector of the feature point. The second controller uses adaptive control techniques to iteratively estimate the calibration matrix and the point's position vector. We prove the stability of both servo-controllers using Lyapunov theory, and present experimental results to evaluate its performance.