Research On The Simultaneous Localization And Mapping Methods For Mobile Robot Based On Hierarchical Optimization

As mobile robots become more autonomous and intelligent,their autonomous navigation capabilities will become increasingly important.Navigation needs a map of the environment,while simultaneous localization and mapping(SLAM)give mobile robots the ability to build maps by themselves.SLAM means that the mobile robots need to estimate the pose of themselves and build the map of surrounding environment concurrently based on the sensors equipped by the robots in the case of no useful pose information and priori maps existing.This makes the SLAM problem extremely challenging due to the interdependence of robot localization and mapping.Therefore,researches on mobile robots SLAM have important theoretical and practical values.Traditional SLAM algorithms such as extended Kalman filter based SLAM algorithm(EKF-SLAM),particle filter based SLAM algorithm(PF-SLAM)and graph-based SLAM algorithm(Graph-SLAM)have drawbacks in map consistency,loop closure and real-time map construction of large-scale environment.To cure the above problems,this paper has conducted an in-depth research,the main works and innovations are described as follows:(1)To solve the consistency problem in SLAM,a hierarchical optimization based simultaneous localization and mapping algorithm is proposed.The proposed SLAM algorithm uses a submap as a node of the graph under the Graph-SLAM framework.A tree structure is adopted to organize the incremental parameterized node pose,and a global level optimization to the node tree is carried out to distribute the loop closure error to the restricted loop.At last,an optimization to the submaps is applied to eliminate inconsistencies in submaps and between submaps.(2)Because the EKF-SLAM which relies on the predefined landmark model have the drawbacks that it is limited to the specific environment and many useful environmental information is discarded in the generated feature maps,to solve these drawbacks,an EKFSLAM which uses raw laser data as landmarks is proposed to build the local submap,this EKF-SLAM not require a specific geometric landmark model.Combining the advantages of scan matching based SLAM and PF-SLAM,a particle filter based hill climbing scan matching SLAM is proposed for the construction of submap.(3)A predictive forward submap segmentation method is proposed.After analyzing the shortcomings of the traditional submap segmentation methods,the idea that segmenting the submap based on the certainty of what degree the submap of the past T time localizes the mobile robot pose is proposed,and a mathematical description of the idea is carried out.(4)To solve the loop closure problem in SLAM,a quadtree based submap correlation matching loop closure detection method is proposed.Compared with the high-precision map,the low-precision map overestimates the probability of obstacles.Based on this foundation,a quadtree is adopted to orignize the loop closure search space and a quadtree based submap correlation matching loop closure detection method is proposed.The proposed SLAM algorithm is tested by simulation,open source data and physical environment experiments and the consistency and the real-time performance of the proposed SLAM algorithm are analyzed.Simulation and experimental results show that the proposed SLAM algorithm can satisfy the real-time requirement and the consistency of the proposed SLAM is better than other similar SLAM algorithms.