Predicting a moving object position for visual servoing: theory and experiments

In order to perform visual servoing tasks in a robotic system, one is confr onted with the low sampling rate of standard cameras and the time delay int roduced by image processing. One way to circumvent the time-delay problem i s to estimate future positions of the moving object of interest employing p rediction techniques. In this work, three prediction techniques, namely Kal man filtering and two adaptive techniques employing least squares with forg etting factor and the projection algorithm, respectively, are evaluated in terms of their prediction error. Experimental results show that the adaptiv e techniques give best results and the Kalman filter-based predictor shows a high sensitivity to velocity changes of the moving object. Copyright (C) 2001 John Wiley & Sons, Ltd.