基于激光雷达的室内同步建图与定位系统研究

发布时间：2018-05-15 02:10

本文选题：同步建图与定位 + 激光雷达　；参考：《哈尔滨工业大学》2017年硕士论文

【摘要】：同步建图与定位是当今在未知环境中实现定位过程中必须解决的关键问题,采用激光雷达进行二维地图的建立和定位是常用的实现手段之一。其中涉及到的主要问题为递增建图与定位,即通过激光雷达对未知环境进行扫描,获取当前时刻的观测与原有创建地图之间的变换关系。本文围绕激光雷达的数据采集,数据预处理以及数据之间的匹配等问题,研究如何通过激光雷达扫描的数据建立地图并实时更新地图与定位,主要的研究内容包括:激光雷达的数据获取。在未知环境中,数据采集和处理是进行后续工作的前提。研究激光雷达的数据采集和传输过程,对栅格地图的建立进行研究。在分析了原有的栅格地图建立方法后,提出简化的栅格地图创建方法,将二维平面均分成等面积的栅格,计算每一个扫描点所落入的栅格的位置,并保存该栅格,将所有保存的栅格输出显示得到栅格地图。基于分支定界的搜索匹配算法,获取当前时刻的扫描点集与原有创建的地图之间的变换关系。在不求取对应特征点的情况下,提出基于分支定界的搜索匹配策略,对其可行性进行证明,并与当前所存在的基于遍历搜索的匹配算法在时间复杂度上进行分析比较。在此基础上,提出基于松弛因子的优化搜索算法,通过松弛因子自适应的改变搜索步长,以减小分支定界搜索过程中步长对搜索结果的影响。基于变基准点集的多帧匹配算法。提出两两匹配的策略,将所采集到的点集都转换至第一帧点集所在的坐标系下,完成匹配并输出定位结果。为减小多帧匹配过程中两两匹配的累计误差,提出倒匹配的匹配策略,并对采集到的点集进行回环检测,完成多帧匹配优化。随着配帧数的增加,累计误差的出现会使地图的精度降低,通过提取地图中的直线来改善显示效果。对传统的分割和合并算法进行分析和改进,并进行仿真验证。实时仿真系统实现。在Visual Studio平台下进行实时仿真。首先,对所需要的PCL库和激光雷达的驱动文件进行配置;其次,介绍并设计实时仿真过程中的总体实现方案以及总体流程;最后实现对未知环境下的同步建图和定位。
[Abstract]:Synchronous mapping and location is a key problem to be solved in the process of realizing localization in unknown environment. The establishment and localization of two-dimensional maps using lidar is one of the commonly used methods. The main problems involved are incremental mapping and location, that is, scanning the unknown environment through lidar to obtain the transformation relationship between the observation of the current moment and the original map. Based on the problems of data acquisition, data preprocessing and data matching between lidar, this paper studies how to set up maps and update maps and location in real time through the data scanned by lidar. The main research contents include: data acquisition of lidar. In unknown environment, data acquisition and processing is the premise of follow-up work. The data acquisition and transmission process of lidar is studied, and the establishment of raster map is studied. After analyzing the original raster map building method, a simplified raster map creation method is proposed. The two-dimensional plane is divided into grids of equal area, the location of the grid falling into each scanning point is calculated, and the grid is saved. Display all saved grid output to get grid map. The search matching algorithm based on branch and bound is used to obtain the transformation relationship between the scanning point set at the current moment and the original map. Without finding the corresponding feature points, a search matching strategy based on branch and bound is proposed, its feasibility is proved, and the time complexity of the existing matching algorithm based on traversal search is analyzed and compared. On this basis, an optimal search algorithm based on relaxation factor is proposed. The relaxation factor adaptively changes the search step size to reduce the influence of step size on the search results in the process of branch and bound search. Multi-frame matching algorithm based on variable reference set. A pairwise matching strategy is proposed in which the collected point sets are converted to the coordinate system in which the first frame points set is located. The matching is completed and the location results are outputted. In order to reduce the cumulative error of pairwise matching in the process of multi-frame matching, a matching strategy of inverted matching is proposed, and the collected point sets are detected in return loop to achieve the optimization of multi-frame matching. With the increase of frame allocation, the accuracy of map will be reduced with the emergence of cumulative error, and the display effect can be improved by extracting straight lines from the map. The traditional segmentation and merging algorithms are analyzed and improved, and the simulation results are verified. Real-time simulation system is implemented. Real-time simulation is carried out on Visual Studio platform. Firstly, the required PCL library and the driver file of lidar are configured; secondly, the overall implementation scheme and overall flow in the real-time simulation process are introduced and designed; finally, the synchronous mapping and location in unknown environment are realized.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN958.98

【参考文献】