Randomized Path Planning for Visual Servoing

Visual servoing has been introduced as a promising solution for sensor-based robotic applications. The basic visual servoing task is to guide the motion of a robot with respect to a target object based on the feedback obtained through a vision system. Despite their popularity, Image-Based Visual Servoing (IBVS) schemes suffer from stability and convergence issues. Moreover, in IBVS techniques, there is no direct control over the image/camera/robot trajectories induced by the servoing loop in the image and physical spaces. Therefore, these trajectories might violate the image and/or physical constraints usually encountered in visual servoing tasks. Incorporating path planning strategies into the visual servo loop is a promising effort towards accounting for a variety of constraints. In this thesis, we propose a general and global path planning framework for image-based control built on the efficiency and success of randomized sampling-based path planning techniques. The proposed planner explores the camera planning space for permissible camera trajectories satisfying image constraints (e.g., camera field of view and occlusions) and simultaneously tracks these trajectories in the robot configuration space to check for robot kinematic constraints and collision with obstacles. The exploration in camera planning space follows a tree-based randomized planning scheme and a local controller is used to track camera trajectories in the robot configuration space. The proposed framework yields global trajectories for the whole robotic system. The solution trajectory is projected into the image space to obtain the corresponding feature trajectories pertinent to a target object. An image-based visual servoing scheme is then adopted to execute the solution feature trajectories. We implemented the proposed framework on a 6 degrees of freedom (DOF) robotic arm and a 9-DOF wheeled mobile manipulator. The effectiveness of the proposed planning scheme in accounting for a variety of image and physical constraints is shown through a number of real world experiments. We also provide an empirical study on the performance of the image-based trajectory tracking scheme under modeling and calibration uncertainties.