Modeling and Control of Sampled-Data Image-Based Visual Servoing With Three-Dimensional Features

Image-based visual servoing (IBVS) with 3-D features refers to the use of 3-D feature vectors in the visual control. The availability of low-cost and lightweight RGB-D cameras makes it natural to use 3-D point coordinates of the acquired image to construct the feature vector. In this article, by using 3-D features, we propose, first, a novel sampled-data model of the feature dynamics, which, in contrast to the usual forward Euler approximation, retains the rigid motion constraint. The need to introduce the sampled-data model arises from the fact that, due to the limited camera frame rate and actuation delays, the effects of a finite sampling time of the visual control system cannot be neglected. The stability analysis of the resulting discrete-time control system is carried out, showing that the desired equilibrium point of the visual control system is almost globally asymptotically stable. Finally, a new IBVS control algorithm is designed by resorting to the Lyapunov direct method. It explicitly takes into account the camera velocity limits while ensuring stability at the same time. Furthermore, despite large sampling times, it guarantees the absence of hidden oscillations and a smooth approach of the camera to the prescribed target configuration. The experiments are carried out in an emulated in-store logistic scenario by performing pick&place and object handover tasks, testifying to the effectiveness of the approach.