Mixed-model based simultaneous localization and mapping approach for mobile robot

A new simultaneous localization and mapping (SLAM) approach based on a mixed map model using sonar data and odometry information is presented. The mixed model composed of occupancy grids and line maps is utilized to represent the environment map. Firstly, three region models and Bayes' rules are used to construct an occupancy grid map. The map precision is enhanced through fusing the information of several sonar sensors at different times. Then, the Hough transform is introduced to extract line features and the line feature maps are created. The local map and the global map are matched by comparing orientation, collinearity and overlap of the straight-line segment in the maps. Finally, the simultaneous localization and mapping are accomplished with the line features and extended Kalman filter through state prediction, observation prediction and estimation phase, which can estimate the robot pose and correct the map model. The simulation results and the real experimental results indicate the feasibility and validity of this approach.