A Study Of SLAM System Algorithm Based On OcTree

With the intelligent mobile robots are widely used in society, it is important to complete autonomous navigation tasks. In order to achieve navigation by themselves, the technology for Simultaneous Localization And Mapping becomes the priority research.Currently, SLAM system are able to achieve the real-time acquisition and the building of surrounding space environment which shown by point cloud data sets. However, It is obvious that the number of point cloud data set is large, and it is difficult to process by limited robots. Therefore, it is urgent to find a simple and easy data model to store, transmit and process for intelligent robots.Based on OcTree of 3D map, OctoMap is an open source project. It provides a kind of grid technology as a precondition and is implemented by OcTree data structure. OctoMap finds the best way to improve the ray intersection test from mathods of Uniform Grid, Grid Based, BVH and Kd-Tree. OctoMap is a method to facilitate autonomous navigation for intelligent robots.In this paper, a study of SLAM system algorithm for 3D map based on OcTree. The paper implements by the following three parts, which are the processing optimization for point cloud data sets, the reconstruction algorithm for ray traversing voxels and the improvement for node updating algorithm based on OcTree. In part of processing optimization for point cloud data sets, the point cloud data sets are downsampled and removed nosie outlier points. It is optimized by noise point exclusion and large quantity. In part of the reconstruction algorithm for ray traversing voxels, the new method for three-dimensional ray traversing voxels is applied to improve its efficiency. In the meantime, the idle space of inflection point is removed to save spatial redundancy. In part of improvement for node updating algorithm, the depletion issues for system resource are taken into account. The origianl recursion call to update is replaced by loop method.The paper uses standard data sets, indoor halls and outdoor vehicle to comprehensively analize and evaluate for all stages of these algorithms. Analysis and evaluation are processed from the processing optimization for point cloud data sets, the reconstruction algorithm for ray traversing voxels and the improvement for node updating algorithm based on Oc Tree. The paper are shown by chart comparison of time efficiency.