Research On Active Visual Servoing Of Mobile Robot

With the rapid development of various technologies involving image process-ing, filtering, control theory and computer vision, visual sensor has been intro-duced into the robot manipulators or mobile robot to increase the flexibility, intelligence and precision. The hot topics of vision-based control include viusal navigation, visual tracking, visual servoing and visual SLAM. Visual servoing is a very challenging issue, which has been proposed to drive a robot to reach the desired position/orientation through visual feedback. However, current visual ser-voing system still faces several key drawbacks such as oversensitivity to camera calibration error and image noise, difficulty to keep the target in the field of view (FOV problem), low servoing efficiency due to the large time cost of image pro-cessing, failure of path planning in complex environment, and so on.For the aforedmentioned troubles with visual servoing control, this disserta-tion constructs an active visual servoing system of the mobile robot and makes a further study on the several key technologies. In general, the work in this dissertation can be summarized as follows:(1)The experiment platform of active visual servoing system is constructed to test the designed Algorithms conveniently. Concretely, the hardware part of the experiment tested is composed of Pioneer3-DX mobile robot, PTU D46-17 pan-tilt platform and DH-SV400FC camera. Furthermore, an object oriented software is programmed by VC++ language, which can improve the efficiency of visual servoing algorithm implementation. Meanwhile, the designed software has the strong ability of the reusability, maintainability, expansibility.(2)A global exponential observer is designed to estimate the camera intrin-sic matrix on-line in the active visual servoing system. Specially, The camera is mounted on the pan-tilt platform and can be rotated in pan and tilt directions, while feature image points are extracted from the real-time image captured by the camera. Furthermore, the relationship between the speed of the camera rotation and the change of image feature points is firstly analyzed based on camera model and the kinematic restriction of the pan-tilt platform. Subsequently, no less than four image points are employed to construct an observer to estimate the cam-era intrinsic parameters. Finally, the stability theorem of the linear time-varying systems is utilized to prove that the proposed observer yields global exponen-tial convergence of the camera intrinsic parameters. Simulation results validate the proposed nonlinear observer, which has such advantages of correct and fast convergent rate.(3)An adaptive active visual servoing system of mobile robot is presented to solve the FOV problem and to improve the servoing efficiency. Specially, in the lower level, an adaptive visual tracking controller is designed to keep the target points lying around the center of the image plane by rotating the pan platform of the camera. On the higher level, a switched controller is utilized to drive the mobile robot to reach the desired configuration through straight-line path. The designed active visual servoing system presents such advantages as follows:ⅰ. a satisfactory solution for the field of view (FOV) problem;ⅱ. global high servoing efficiency;ⅲ. free of any complex pose estimation algorithms usually required for visual servoing systems. Both simulation and experiment results are provided to validate the effectiveness of the proposed active visual servoing method.(4)An active visual servoing system without desired image is proposed for some special application in certain circumstances. Compared with the traditional visual servoing system, the designed scheme can extract the state information from the initial and current image. Specially, the reference coordinate is firstly es-tablished by the target points. Subsequently, the angles between the mobile robot and the target plane can be calculated by real-time image feedback. Meanwhile, the distance information is estimated on-line by a nonlinear observer. Based on these, a switched controller is utilized to drive the mobile robot to the desired position/orientation denoted in reference coordinate. Experiment results validate the proposed algorithm.