The Lidar-Based Mapping And Relocalization In Large-Scale Environment

3D environment reconstruction and relocation have always been the important parts of the field of mobile robot research,and also the premise and guarantee for the practical application of robots.Because the measurement precision of a 3D laser is high and it is not affected easily by environmental factors such as illumination,it makes laser more practical than visual sensor.Therefore,the paper mainly studies constructing a sparse map for large-scale environment and relocation problems based on 3D laser radar data.For the issue of mapping based on point cloud,the paper proposes a sparse 3D point cloud map generation method,which use the curvature of the trajectory point and other constraints to sparse the trajectory of the laser,and then build a sparse 3D point cloud map by using the time synchronization relationship between the trajectory point and the point cloud.For the issue of relocalization,the paper proposes a 3D point cloud relocalization method for large-scale environment,which first utilizes 2D projection to perform fast similarity metric to obtain several candidate scenes in the prior map similar to the current scene.Then it utilizes a 3D similarity measurement method to verify,so that the scene with the highest similarity with the current scene is found in the candidate scene,and finally the relocation is completed in the a priori map by the registration algorithm.This paper validate the proposed algorithms on the campus dataset of DLUT,and then it's compared with the existing methods.The result shows that the sparse method proposed in this paper overcomes the problem of precision degradation caused by raster downsampling by traditional methods,and overcomes the shortcomings of non-nodes in topological maps.The relocalization algorithm proposed in this paper solves the problem of time comsuming and the memory overhead caused by feature extraction or filtering algorithm,guarantees the real-time performance and memory efficiency of the algorithm,and can also solve the problem of robot starting at any point in the existing map.