Dynamically Estimating Mobile

Complex robotic tasks such as autonomous exploration and grasping demand the co-operation of sensors and actuators. In order to integrate sensor measurements and actuator control schemes we must determine the rigid body transformations that relate the native co-ordinate frames of these devices. Equivalently, we need to estimate the relative pose of sensors and actuators in the system. We examine the problem of determining the pose of a robotmounted rangending camera, and present a class of solutions motivated by the idea that mobile camera calibration is best addressed by an ongoing dynamic estimation process. We use range measurements and known robot kinematics to provide the estimate of camera pose which is maximally consistent with the available data. Our scheme uses scene features that are often present in typical workcell scenes and that are easily and reliably extracted. We develop several formulations of the principle, and present experimental results for both simulated and real data sets. R esum e. L'exploration autonome et la pr ehension, comme exemples de tâches robotiques complexes, requirent la coop eration de plusieurs capteurs et de plusieurs moteurs. L'int egration de donn ees visuelles dans le sch ema de contrôle demande que l'on connaisse les transformations qui relient les coordonn ees de base de chaque sous-syst eme. Autrement dit, il faut estimer la pose relative des capteurs et manipulateurs composant le syst eme. Nous examinons l'estimation de la pose d'une cam era t el em etrique manipul ee par robot, et pr esentons une classe de solutions qui d ecoulent de l'id ee que la calibration est un proc ed e dynamique et continuel. Les mesures prises par la cam era et la connaissance de la cin ematique du manipulateur permettent d'extraire estimation la plus consistante de la transformation cam era-robot. La m ethode exploite des particularit es g eom etriques qui sont pr esents dans la plupart des sc enes typiques, et qui sont facilement identi ables. Nous d eveloppons plusieurs formulations du principe, et nous pr esentons les r esultats d'essais avec des donn ees simul ees et r eelles. Acknowledgements The author would like to acknowledge the valuable assistance provided by several individuals in the preparation of this work. Gilbert Soucy and Pierre Tremblay created many of the tools used to control the robot and rangender in our lab, and provided speedy technical assistance on numerous occasions. John Lloyd and Peter Whaite consistently provided keen insight and advice at critical moments, and their assistance has been invaluable. Prof. Vincent Hayward was patient and accommodating beyond all reasonable call as this thesis neared completion. Prof. Frank Ferrie o ered thoughtful guidance and supervision, as well as timely encouragement when prospects looked grim. Lisa Ramsaran smoothed out a great deal of bad punctuation, and provided much-needed moral support. Thank you. dlb 94/04/19 TABLE OF CONTENTS TABLE OF CONTENTS LIST OF FIGURES iii LIST OF TABLES iv CHAPTER