A probabilistic roadmap based path planning for visual servo of robotic manipulators

• Main Author: Arvani, Farid.
• Other Authors: Memorial University of Newfoundland. Faculty of Engineering and Applied Science
• Format: Thesis
• Language: English
• Published: 2008
• Subjects: Robot vision; Robots > Control systems; Robots > Motion; Newfoundland studies; University of Newfoundland
• Online Access: http://collections.mun.ca/cdm/ref/collection/theses4/id/32401

Thesis (M.Eng.)--Memorial University of Newfoundland, 2009. Engineering and Applied Science Includes bibliographical references (leaves 108-120) Vision feedback is a competent control technique for a large class of applications but they suffer from several imperfections. The well-known image-based visual servo (IBVS) methods regulate error in the image space i.e. the controller compares the current view of the target against the reference view and generates an error signal at the sampling rate of the vision system. -- Contrary to position-based visual servo (PBVS), which regulates error in Cartesian space, IBVS ensures a local stability and convergence in the presence of modeling error and noise perturbations since the control loop is directly closed in the image space. However, sometimes (and specifically) when the initial and desired configurations are distant, the camera trajectory induced by IBVS is neither physically valid nor optimal due to the nonlinearity and singularities in the relation from image space to the workspace which can cause the target to leave the field of view. Furthermore, introducing constraints such that the target remains in the camera field of view and/or such that the robot avoids its joint limits during servoing is not trivial in classical PBVS and IBVS control techniques. When the displacement to realize is large, this incapability leads to the failure of servoing process. -- This research presents a method to resolve the problems associated with classical servo control. Visual servoing control solutions are local feedback control schemes and thus require the definition of intermediate subgoals at the task planning level. This work introduces and details a trajectory planning scheme in order to achieve more robust visual servoing through the introduction of subgoal images. This ensures that the error signal is kept small since the current measurement always remains close to the desired value so that one can exploit the local stability of the IBVS control solution. The proposed method is based on Probabilistic Roadmaps (PRM) and its flexible platform is used to introduce desired constraints such as visibility constraint, joint limit constraint, obstacle avoidance constraint, and occlusion avoidance constraint to the generated path at the task planning level. It is noteworthy that visibility constraint is intended to keep the target in the camera field of view (FOV). Joint limit constraint restricts the manipulator to avoid its joint limits. Obstacle avoidance and occlusion avoidance constraints ensure that the generated path is collision- and occlusion-free. One of the advantages of the proposed method is that targets are not required to have 3D models. However the method requires a 3D model of the obstacles to avoid obstacle collision and occlusion. -- The proposed method plans the camera trajectory using PRM and then deduces the corresponding trajectories in the image plane which is a discrete geometric trajectory of the target in the image plane. A continuous and differentiable cubic spline presentation of the feature trajectories in the image plane is computed to be used as a time-varying reference to pure IBVS loop. Off-line path planning is performed using the kinematics of a 5-DOF robot arm to confirm the validity of the approach. Simulation of different IBVS scenarios is provided to demonstrate the performance of the proposed method.