Robot Mapping Based On The Uncertain Information Process Algorithm

This thesis is based on the key technique project of"Intelligent Information Processing and Navigation of the Robot Exploring Moon"in National 863 Project, and puts emphasis on mapping on autonomous mobile robot in uncertain environment.The relative technology and method of mapping is described, the mapping method based on the uncertain information dealing technology is proposed. The information of laser, visual and sonar are combined in order to make the grid map and geometry map, expressing the character of the whole environment.Because of the uncertainty of the environment and information of the sensor, the previous classical, logistic, accurate arithmetic are invalid for the objective model of the environment. D-S evidence theory and the gray system theory are used to fuse the sensor data in order to acquire the uncertain description of the environment using for the obstacle avoiding and environment understanding. According to the D-S evidence theory, the height of the obstacle is abstracted by the laser data, and the gray distributing change rate is abstracted by the CCD camera. The rule of the reasoning is acquired by a lot of experiments. Uncertain process model is used to fuse the sensor data, the obstacle probability of the environment is concluded, which help the robot mapping and avoiding obstacle in the environment. Sonar error model is established to describe the gray cell information, The cell is updated by using the gray system theory according to the sonar data, as it is, all the cell information of the environment can be obtained for the grid map. Simulated experiment has proved the feasibility of the algorithm.A simultaneous localization and mapping algorism (SLAM) has been described based on a Kalman filter and character matching algorithm. The system model and the observation model are established by using laser and millimeter. The robot position and the environment character are prognosticated and updated by Kalmanfilter. In another method of SLAM algorithm, the environment characters are abstracted by laser data, and matched by the consistency of the relative position and the consistency of the length between the same characters. This method has been implemented on the mobile robot mapping; its validity has been improved.