The main work and specific contributions of this research are as follows:Firstly, a novel Monte Carlo method for robot self-localization (MCL) based on monocular vision is presented. To solve the problem that traditional MCL algorithm cannot achieve good results only using information of landmarks, which is obtained from the robot vision system, in this method MCL algorithm and the UKF is combined together. At the same time, the information of landmarks, the position and azimuth of the robot obtained from vision system, the encoder of the driving motors, the electronic compass of the robot, are all used to improve the precision of the self-localization of the robot.Secondly, a Monte Carlo method based approach for multi-robot localization is proposed. this approach applies grid cells to describe the whole particle set which is used in traditional MCL method to estimate the pose of robot. Then, the sizes of the grid cells are adjusted to capture the characteristic particles which represent the properties of all particles. The characteristic particles can be used to estimate the robot’s position in its operation space. Because the number of the characteristic particles is much less than that of the total particles, this approach can save the computing time greatly. Simulation results show that this approach can obtain good localization performance in multi-robot system. Thirdly, a new path planning method based on local map and global map is presented. When mobile robot runs in part-known environment, we use global topologic map to describe the known environment information. When the robot knows its position, it can extract local environment information from global map, then use the directed line segment to describe known obstacle, and these line segments are applied to build up local map. Robot can rebuild and adjust local map from sensor readings when it running. This method can get good performance in complexity environment by fusing the local and global map information, historical track information and sensor readings.Fourthly, by means of analyzing the traditional control architecture of autonomous robots, we present a new control architecture based on a new type of mobile robot AIM, which has distributed master/slave architecture. Then each modules of this kind of robot is described in details, and a localization experiment based on monocular vision is conducted to verify this control architecture. Finally, the obtained research results are summarized and future work is addressed.
修改评论