Mobile Robot Localization Methods Based On Multi-source Information Fusion

Mobile robot localization is the basis and prerequisite for robot path planning and navigation.With the increasing application demands of mobile robots,the localization problem has become a hot research field in robotics nowadays.The background of this work is the ICRA DJI Robo Master Artificial Intelligence Challenge.The thesis focuses on the localization of mobile robots in a known environment.The main work of this thesis is as follows:First,the definition of the coordinate system and the transformation relationship between the coordinate systems are given.The kinematic model of the Mecanum wheel robot and environment map model are established.The localization problem is divided into local localization and global localization.The problem of localization failure is considered and the mathematical description of the localization problem and the problem of localization failure is given.Secondly,the local localization methods of mobile robot is studied.Two single-sensor localization algorithms using odometer and lidar are given respectively.Aiming at the problem of odometer error accumulation,combining the high frequency of odometer and the accuracy of lidar,an EKF-based odometer-lidar fusion localization algorithm is proposed which matchs the lidar point cloud and point cloud map by ICP algorithm to eliminate accumulated errors.Thirdly,the global localization method of mobile robots is studied.The odometer probabilistic motion model and environmental measurement model are established.The setup method of the model parameters is given.A complete AMCL localization algorithm combining augmented Monte Carlo localization and KLD sampling is given.Then,the mobile robot localization correction method is studied.The local localization algorithm's validity to deal with the robot kidnapping problem is analyzed and a local localization correction algorithm based on UWB sensor is proposed.The processing ability of the global localization algorithm for robot kidnapping and symmetric localization is analyzed,and two global localization correction algorithms based on UWB sensors are proposed.Finally,the simulation platform is built and all the proposed algorithms are implented.The effectiveness and feasibility of the proposed localization algorithms and localization correction algorithms are verified.