Adaptive 2D visual servoïng using variable structure neural networks

Abstract

We are interested in this paper in the 2D visual servoing for a mobile robot using a Radial Basis Function (RBF) Neural Network (NN). In fact, the interaction matrix, expressing the relationship between the camera motion and the consequent changes on the visual features, contains some parameters to be estimated (depth) and requires a camera's calibration phase. Moreover, the model of the robot can contain uncertainties engendered by the sliding movement. An online identification, using the NN has been proposed to overcome these problems. The RBF NN is used to estimate the block formed by the interaction matrix and the model inverts of the robot. Actually, the variables number of the estimated function is important, which can cause a problem in the use of an excessive number of RBFs. As a remedy, we have proposed a new approach based on proving that a single point of the object is sufficient to solve the 2D visual servoing mobile robot problem. In order to keep only the useful neurons, we have used a variable structure NN whose the number of neurons can increase or decrease during the visual servoing.