Laser-Based Simultaneous Localization And Mapping

Simultaneous localization and mapping (SLAM) problem is a key issue for autonomous mobile robot. SLAM problem not only is very important to accomplish autonomous navigation and other complex intelligent tasks, but also embodies perception ability and intelligence of robots. As a very complicated problem, SLAM problem relates to many aspects of mobile robot research, especially to representation of environment, description and handling of uncertain information, data association, exploration strategy, and so on. This paper focuses on theory and implementation of simultaneous localization and mapping. The main research achievements are as followings.Three methods of environment representation are introduced, including topological maps, occupancy grids, and feature maps. Indoor environment is modeled as feature map since it has plenty of segments, and an agglomerative hierarchical clustering algorithm is used to extract line features and build local maps.The probabilistic form and graphical model of SLAM problem are introduced. Also, the two main solutions to the SLAM problem, including extended Kalman filter method and Rao- Blackwellized particle filter method are introduced.An algorithm for solving SLAM problem for indoor environment is presented in this paper. The considered scenario is that of a mobile robot using a 2D laser range finder providing range scans and two encoders used as odometry. An agglomerative hierarchical clustering algorithm is used to extract line features whose parameters are extracted from range scans, while the corresponding covariance matrices are computed from the statistical properties of the raw data. Simultaneous updates of the robot pose and line features estimate are performed via extended Kalman filtering. All operations required for building and maintaining this map, such as motion model, observation model, prediction and state-updating are described and formulated. Scan matching algorithm is used to correct the building map further.Many experiments are performed based on mobile robot with proposed SLAM method. The experimental results demonstrate that the proposed method can successfully create a consistent map while locate itself using the map.