Robust And Accurate Pose Estimation In Complex Environment

Pose estimation technology has always been a research hotspot in the field of intelligent autonomous mobile robots,and it is also the basis of robot autonomous localization technology.In recent years,the development of pose estimation technology has proposed many excellent solutions for mobile robot localization,mapping,and navigation.However,most pose estimation methods are currently limited to "conventional scenarios",which is difficult to deal with complex scenarios,such as global localization problems in GPS failure scenarios,positioning accuracy problems in complex factory environments,and localization problems in complex outdoor environments.Great question.In view of the above problems,the research work mainly includes the following three parts:(1)Aiming at the global localization problem in GPS failure scenarios,a robust global localization scheme is proposed.The algorithm mainly consists of three modules,namely the global point cloud map preprocessing module,the position recognition module,and the pose estimation module.The map pre-processing module divides the global map into local maps,and builds the local maps after the robot is powered on.The location recognition module calculates the matching score of the local map and the global submap based on the ESF global descriptor to realize location recognition.The pose estimation module adopts the process from coarse matching to fine matching.The rough matching is realized based on the SHOT descriptor.The result is used as the initial value of the ICP or NDT algorithm,and finally the global localization is achieved.The algorithm of this paper has been tested in high-speed environment,urban environment and mine environment.The experimental results show the robustness of the algorithm.(2)In view of the current situation that two-dimensional localization depends on other auxiliary facilities such as magnetic nails,two-dimensional codes,reflective columns,etc.to achieve the millimeter level,this paper combines adaptive Monte Carlo localization with scanning matching to make positioning accuracy not limited Map resolution does not depend on other auxiliary facilities to achieve accurate positioning.The algorithm has been tested in an actual environment.The static environment can reach 1cm and the dynamic accuracy is about 2cm.(3)For outdoor complex environments,most pose estimation methods are applied to scenarios in which motion does not occur suddenly and the geometric environment is good.In this paper,Unscented Kalman Filter(UKF)is used to establish the IMU velocity and motion model and the NDT scan matching observation model to fuse the laser information with the IMU information.The experimental results show that the method can effectively deal with strenuous motion and can achieve robust pose estimation in complex scenes such as mines.In the end,the full text is summarized and the future research content is prospected.