Visual Control of Robotic Manipulators
Autore
Paulo Sequeira Gonçalves - Technical University of Lisbon - Instituto Superior Tecnico - [2005]
Documenti
Abstract
The work in thesis aims at the visual control of robotic manipulators, i.e. visual servoing. It is presented the state-of-the-art on the subject and the computer vision tools needed to its implementation. In this thesis are presented six contributions to visual servoing, namely the development of an
experimental apparatus, two dynamic visual servoing controllers, the application of fuzzy filters to kinematic visual servoing, the fuzzy modeling of the robot-camera system and the fuzzy control based on the inverse model.
The experimental apparatus has three different components, namely a planar robotic manipulator with two degrees of freedom, a 50 Hz vision system and the developed software to control and inter-connect the two previous components. The developed experimental apparatus allowed the real-time experimental validation of the controllers proposed in this thesis.
The robot joint actuators are directly driven by dynamic visual
servoing, in opposition to kinematic visual servoing that generates the joint velocities needed to drive the robot, by means of an inner velocity control loop. The first contribution to dynamic visual servoing is an image based control law specially developed to the robot of the experimental apparatus, with the eye-in-hand. The second contribution is a position based control law to the eye-in-hand configuration, applicable to robots with more than two degrees of freedom. For both the controllers the asymptotic stability is demonstrated.
The application of fuzzy logic to image based kinematic visual
servoing, revealed three contributions. With the application of
fuzzy filters to path planning and to regulator control, the
overall performance of visual servoing is improved. The robot
joint velocities diminish at the initial control steps and its
oscillatory behavior is also diminished when the vision sample
time is high. The inverse model of the robot-camera system is
obtained by means of fuzzy modeling. A practical methodology for obtaining the model is also presented. The fuzzy inverse model is directly used as the controller of the robot-camera system, in order to deliver the joint velocities, needed to drive the robot to the desired position. It was also used a fuzzy compensator to compensate possible mismatches between the obtained model and the robot-camera system.
experimental apparatus, two dynamic visual servoing controllers, the application of fuzzy filters to kinematic visual servoing, the fuzzy modeling of the robot-camera system and the fuzzy control based on the inverse model.
The experimental apparatus has three different components, namely a planar robotic manipulator with two degrees of freedom, a 50 Hz vision system and the developed software to control and inter-connect the two previous components. The developed experimental apparatus allowed the real-time experimental validation of the controllers proposed in this thesis.
The robot joint actuators are directly driven by dynamic visual
servoing, in opposition to kinematic visual servoing that generates the joint velocities needed to drive the robot, by means of an inner velocity control loop. The first contribution to dynamic visual servoing is an image based control law specially developed to the robot of the experimental apparatus, with the eye-in-hand. The second contribution is a position based control law to the eye-in-hand configuration, applicable to robots with more than two degrees of freedom. For both the controllers the asymptotic stability is demonstrated.
The application of fuzzy logic to image based kinematic visual
servoing, revealed three contributions. With the application of
fuzzy filters to path planning and to regulator control, the
overall performance of visual servoing is improved. The robot
joint velocities diminish at the initial control steps and its
oscillatory behavior is also diminished when the vision sample
time is high. The inverse model of the robot-camera system is
obtained by means of fuzzy modeling. A practical methodology for obtaining the model is also presented. The fuzzy inverse model is directly used as the controller of the robot-camera system, in order to deliver the joint velocities, needed to drive the robot to the desired position. It was also used a fuzzy compensator to compensate possible mismatches between the obtained model and the robot-camera system.
Questa tesi è correlata alle categorie