Kernel-based sliding mode control for visual servoing system

In this study, a new approach to design a controller for a visual servoing (VS) system is proposed. Kernel-measurement is used to track the motion of a featureless object which is defined as sum of weighted-image value through smooth kernel functions. This approach was used in kernel-based VS (KBVS). To improve the tracking error and expand the stability region, sliding mode control is integrated with kernel measurement. Proportional-integral-type sliding surface is chosen as a suitable manifold to produce the required control effort. Moreover, the stability of this algorithm is analysed via Lyapunov theory and its performance is verified experimentally by implementing the proposed method on a five degrees of freedom industrial robot. Through experimental results, it is shown that the performance of tracking error in the proposed method is more suitable than KBVS, for a wider workspace and when the object is placed near the boundary of the camera´s field of view.