Research On Vision Navigation Technology For Service Robot

With the development of robot and information technology, intelligent service robot will play a key role in future digital home. Intelligent service robot roams in the dynamic indoor environment with sensors which perceive the surrounding environment and guide the robot. The character of vision system is that it can capture a great deal of information of the environment, and the inexpensive price, but the uncertainty of the camera has brought enormous challenge to the robot real-time localization and navigation. So to explore the stereo vision technology of service robot employed in the dynamic home environment is very important to solve the problem of the environment map building and real-time self-localization of the mobile robot under the non-structural environment.The robot platform which is used in our research is a fully independent intellectual property service robot which is designed by Shenyang SIASUN, equipped with a stereo camera system. The research is aiming at studying the relational technology of the stereo vision system. In theory and practice, several key technologies have been explored, such as stereo camera calibration, artificial landmarks based mobile robot self-localization, vision and odometer based simultaneous localization and mapping etc. The main works are as follows:The camera pinhole model is introduced, and the relationship between the coordinate system in the camera model is established. To the real situation, Zhang's calibration method is used to calibrate the stereo camera, and the internal and external parameters of the camera have been defined.In the mobile robot self-localization based on artificial landmarks, a new artificial landmark model is designed, and the corresponding algorithm for landmark detection and recognition is proposed. Experiment results show that the algorithm is robust to illumination, rotation and scales. The self-localization formula based on artificial landmarks and stereo vision is deduced, experiment results show that the accuracy of the system could fully meet the requirements for mobile robot self-localization. In vision based simultaneous localization and mapping, the EKF algorithm widely used in SLAM system has been studied, and the principle of EKF is expounded. Vision based EKF-SLAM algorithm implement process is proposed. The simulation experiment results show the reliability and realizability of the system, and the experiment which mobile robot run in the laboratory shows the feasibility of the algorithm.