Affine visual servoing: a framework for relative positioning with a robot

Abstract

In this paper, a framework for vision-based relative positioning called affine visual servoing (AVS) is presented, in which the changes in shape of image contours are used in order to control the positioning process. A feedforward control strategy complements the feedback loop, thus enhancing both speed and the overall performance of the system. The full linearization of the visual servoing problem results in a dramatical reduction of the computational burden for both the control and vision subsystems. Good simulation results with stable behavior show that the approach may be successfully applied for the design of complex positioning tasks, such as navigating by means of natural landmarks, or mimicking human gestures.