No description available.

No description available.

No description available.

The performance of robotic manipulators is critical to their widespread use in industry. As manipulators become faster, their potential productivity can rise thus improving the return on the investment required to purchase them. Improving accuracy, on the other hand, increases the range of tasks for which the manipulator is suitable. The speed and accuracy of a manipulator is partly determined by the capability of the algorithm used to control it. Whilst being a highly non-linear multiple input, multiple output device, however, most industrial controllers are derived on the basis that the robot is a series of independent, linear actuator+ link subsystems. The resulting independent joint controller is simple to design and implement but is limited in its performance as link interactions and the non-linear effects of centrifugal and Coriolis forces degrade the accuracy at high manipulator velocities. Improvements in the control of manipulators may be made by incorporating a mathematical model of the manipulator in the control algorithm. Control schemes such as `computed torque' incorporate an inverse model of the manipulator to calculate the input torques required to force the end-effector to follow a desired trajectory. The equations of motion required to implement these controllers are large and complex even for relatively simple manipulators. This thesis explores how bond graph representations of robotic manipulators may be used to automate the implementation of model based controllers. To provide a practical basis for this research the bond graph derived controllers are tested on an experimental rigid, planar, direct drive two-link manipulator. It is shown how the bondgraph for this manipulator, including d.c. motor actuators, can be constructed and used to derive the equations of motion of the manipulator automatically. The bond graph model is then validated by comparing simulations obtained using these equations of motion with experimental data.

No description available.

Η παρούσα εργασία μελετά την εξερεύνηση χώρου από κινούμενα ρομπότ. Πιο συγκεκριμένα, περιοριζόμαστε σε ένα δισδιάστατο χώρο γνωστού μεγέθους, στον οποίο υπάρχουν τυχαία τοποθετημένα εμπόδια. Τα όρια του χώρου είναι εκ των προτέρων γνωστά, ενώ τα εμπόδια είναι ανιχνεύσιμα εφόσον το ρομπότ είναι πλησίον τους. Η μελέτη αφορά σημειακά ρομπότ, τα οποία ξεκινούν από a priori γνωστές συντεταγμένες. Ο χώρος διαμερίζεται σε κελιά και το ρομπότ μπορεί να μεταβεί σε ένα από τα τέσσερα γειτονικά του κελιά. Παράλληλα τα ρομπότ ανταλλάσσουν πληροφορίες μεταξύ τους, αναφορικά με την περιοχή που έχουν ήδη καλύψει. Τελικός στόχος της εργασίας είναι τα ρομπότ, σε συνεργασία μεταξύ τους, να καλύψουν όλα τα κελιά στα οποία είναι διαμοιρασμένος ο χώρος στον ταχύτερο δυνατό χρόνο. / This thesis studies the surveillance problem using mobile robot. We consider an a priori known two-dimensional space, which has randomly placed obstacles. The boundary of the area is known in advance, while the obstacles are detectable, if they are inside the robot's sensing range. The study concerns point-robots, that start from a priori known coordinates. The space is partitioned into cells and the robot can move to one of the four neighboring cells. Furthermore, robots exchange information among themselves regarding the area that is already covered. The ultimate goal of this work is a collaborative scheme for mobile robots in order to survey the area at the smallest possible time.

No description available.

No description available.

No description available.

No description available.