基于激光雷达的同时定位与室内地图构建算法研究

发布时间：2018-08-30 10:28

【摘要】：随着我国工业自动化程度的提高,越来越多的工厂和仓储车间实现了智能化生产和运输,而不需要人工的参与。在这一过程中,自动导引车(Automated Guided Vehicle,AGV)作为一种新型的智能运输装置,起到了重要作用。本课题以此为背景进行了基于激光雷达的室内移动机器人同时定位与室内地图构建算法的研究。首先定义了机器人系统坐标系,完成了基于里程计的运动模型和激光雷达环境感知模型的搭建,并根据激光雷达传输数据类型研究了激光雷达数据的读取方法。通过对比论证了本课题采用栅格地图作为地图模型的可行性,完成了基于概率模型的栅格地图状态更新算法的研究,并用MATLAB验证了该算法的建图效果。针对基于粒子滤波SLAM定位算法中因粒子退化导致位姿估计误差较大的问题,本文以贝叶斯滤波理论为基础,通过粒子滤波定位算法的理论推导,分析了粒子滤波过程中粒子退化的原因,针对该问题,完成了重采样算法的研究,设计了基于不同重采样算法的粒子滤波器,并完成仿真环境中的定位实验,根据实验效果选择了更适用于本课题的分层重采样算法的粒子滤波器。在此基础上,引入Rao-Blackwellized粒子滤波器(RBPF)的概念,并进行了常规RBPF-SLAM算法的研究。在常规RBPF-SLAM算法框架的基础上研究了混合提议分布的理论,并提出了基于分层重采样的改进RBPF-SLAM算法,通过编写仿真程序利用公开数据集进行仿真实验,验证了算法改进的有效性。以Turtlebot机器人为机器人实验平台,PC为上位机,分别运用原始的和改进后的SLAM算法进行真实环境中定位与建图实验,实验表明运用改进后的算法能够得到更好的定位与建图效果。最后完成算法的嵌入式移植,并结合AGV样机进行给定运行轨迹下的定位与建图,验证了算法在实际应用中的有效性。
[Abstract]:With the improvement of industrial automation in China, more and more factories and warehousing workshops realize intelligent production and transportation without the need of manual participation. In this process, (Automated Guided Vehicle,AGV (automatic guidance vehicle) plays an important role as a new type of intelligent transportation device. In this paper, the algorithm of simultaneous localization and indoor map construction of indoor mobile robot based on lidar is studied. Firstly, the coordinate system of robot system is defined, the motion model based on odometer and the environment sensing model of lidar are built, and the reading method of laser radar data is studied according to the type of data transmitted by laser radar. The feasibility of using the raster map as the map model is compared and the research of the raster map status updating algorithm based on the probability model is completed. The effect of the algorithm is verified by MATLAB. In order to solve the problem of large position and orientation estimation error caused by particle degradation in SLAM location algorithm based on particle filter, this paper is based on Bayesian filtering theory and deduces the theory of particle filter localization algorithm. The reason of particle degradation in particle filter is analyzed. The resampling algorithm is studied, particle filter based on different resampling algorithm is designed, and the localization experiment in simulation environment is completed. According to the experimental results, the particle filter, which is more suitable for the layered resampling algorithm, is selected. On this basis, the concept of Rao-Blackwellized particle filter (RBPF) is introduced, and the conventional RBPF-SLAM algorithm is studied. Based on the conventional RBPF-SLAM algorithm framework, the theory of mixed proposal distribution is studied, and an improved RBPF-SLAM algorithm based on stratified resampling is proposed. The effectiveness of the improved algorithm is verified by writing a simulation program and using the open dataset to carry out simulation experiments. Using Turtlebot robot as the robot experimental platform and PC as the host computer, the original and improved SLAM algorithm are used for the real environment localization and map building experiment respectively. The experiment shows that the improved algorithm can get better localization and mapping effect. Finally, the embedded transplant of the algorithm is completed, and the location and construction of the AGV prototype under the given running trajectory are carried out, which verifies the effectiveness of the algorithm in practical application.
【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】：TP242;TN958.98

【参考文献】