Realization Of Mobile Robot Simultaneous Localization And Mapping In Large-scale Environments

The mobile robot is very intelligent, autonomous and maneuverable, and it can help the humanity complete some tasks such as exploration, survey or operation in the complex or dangerous environments. The autonomous moblie robot has to possess the autonomous navigation ability because reliable navigation is the basis of realization of mobile robot automation. Navigation in unknown environments, has attracted more and more attention in mobile robotics field. In unknown environments robot does not exist the priori map environmental information that may provide the accurate navigation information, therefore its'autonomous navigation under the unknown environments becomes a leading problem in the motion robot field. There are two aspects needed to be considered in solving this problem: one is how to realize the localization in the unknown environments, and the other is how to discribe the environments and map the enviroments. Simultaneous Localization and Mapping(SLAM) is presented to deal with these two difficulties. SLAM has attracted significant attention by the researchers of the mobile robotics research communities since it was introduced, and its'important theory and value of application are counted as a key question to the realization of the true autonomous mobile robot by many researchers. So SLAM has become a focus in the autonomous navigarion field. During the past decade, folks have made amazing progress in solving the SLAM problem of the domains from indoors, outdoors and underwater. However, these methods still have many unsolved and unperfect problems when they are implemented in unknown large-scale environments.This dissertation focuses on SLAM in unknown large-scale environments. In order to solve the localization and the map construction precision problem caused by the robot pose and map's uncertainty and the computational complexity which increases along with the number of the environment features in large-scale SLAM problem, submap-based SLAM is a more effective method. A novel algorithm is presented based on submap method. This new algorithm combines the strengths of two popular mapping strategies: FastSLAM and SEIFSLAM, to improve the precision and efficiency. The FastSLAM is used to produce local map, which is periodically fused into an SEIFSLAM. FastSLAM can avoid the linearization of the robot model during operation and provides a robust data association, and submap joining based on SEIFSLAM can improve the whole computational speed and avoid FastSLAM's tendency to be over-confident. This novel algorithm has a better consistency in large-scale SLAM problem. In order to improve the computational speed effectively in the real-time environments, a binary tree-based decision-making strategy is introduced to reduce the computational complexity. And this strategy can make the computational complexity less than O ( nlogn). In addition, compared with other algorithms, simulation experiment and the fomous Victoria Park dataset are used to validate the presented algorithm. And the results show that the new combined SLAM algorithm indeed improves consistency and precision of the estimation of the localization and mapping of the robot which works in large-scale environments, as well as the reducing of computationacl complexity.