基于二维激光雷达的无人运动平台环境感知方法研究

发布时间：2018-06-01 06:17

本文选题：二维激光雷达 + 聚类算法　；参考：《北京理工大学》2016年硕士论文

【摘要】：随着现代化进程的加速推进,无人运动平台扮演着日益重要的角色,其中环境感知技术是无人运动平台中关键技术之一。为此,本文以旅行家二号试验车、二维激光雷达以及惯性传感集成模块为实验平台,针对无人运动平台的障碍物的检测以及无人运动平台的定位问题展开了深入讨论。在障碍物检测方面,本文以二维激光雷达数据为基础,首先利用阈值滤波和改进的中值滤波算法对数据进行预处理,然后借助激光雷达数据的上下文丰富障碍物的数据量,最后通过聚类算法对障碍物进行了识别。同时本文提出利用凸包算法解决了聚类过程中存在的冗余聚类信息,这种方法大大提高了系统的鲁棒性。针对无人运动平台的定位问题。本文首先对无人运动平台的运动模型进行了分析,并确定了影响无人运动平台定位准确度的因素,然后通过IF-THEN模糊预测模型对无人运动平台的定位误差进行了估计。最后通过二维激光雷达前后帧特征数据对无人运动平台的定位估计进行了最优处理。数据结果显示基于模糊预测和二维激光雷达特征数据的定位算法要优于单纯依靠惯性器件的定位信息。在无人运动平台的硬件及软件系统方面,本文以旅行家二号试验车为移动平台,通过SICK二维激光获取环境感知信息、ASV940惯性传感器集成模块获取运动信息、移动电脑进行算法运行。在软件方面,本文基于VS2010平台,实现了激光雷达数据实时采集功能界面、旅行家二号试验车的运动控制界面、惯性传感器集成模块信息显示界面以及算法功能选择界面。在实验过程中,本文分别在动态、静态环境中对障碍物识别算法进行了验证。实验表明,两类环境中CBFD聚类算法(基于峰值点的聚类算法)在聚类合理性上均优于K均值聚类算法和ISODATA聚类算法(迭代自组织数据分析算法)。
[Abstract]:With the acceleration of the modernization process, unmanned sports platform plays an increasingly important role, among which environmental awareness technology is one of the key technologies in unmanned sports platform. Therefore, this paper takes Traveler No. 2 test vehicle, two-dimensional lidar and inertial sensor integration module as the experimental platform, and discusses the obstacle detection of unmanned motion platform and the positioning of unmanned motion platform in depth. In the aspect of obstacle detection, based on the two-dimensional lidar data, the threshold filter and the improved median filtering algorithm are first used to preprocess the data, and then the context of the lidar data is used to enrich the amount of obstacle data. Finally, the obstacles are identified by clustering algorithm. At the same time, a convex hull algorithm is proposed to solve the redundant clustering information in the clustering process, which greatly improves the robustness of the system. This paper aims at the localization of unmanned platform. In this paper, the motion model of unmanned motion platform is analyzed, and the factors that affect the accuracy of unmanned motion platform are determined, and then the positioning error of unmanned platform is estimated by IF-THEN fuzzy prediction model. Finally, the location estimation of the unmanned moving platform is optimized by using the feature data of the front and rear frames of the two dimensional lidar. The results show that the localization algorithm based on fuzzy prediction and two-dimensional lidar feature data is superior to the location information based on inertial devices alone. In the aspect of hardware and software system of unmanned motion platform, this paper takes Traveler No. 2 as the mobile platform, acquires the environment sensing information by SICK two-dimensional laser and ASV940 inertial sensor integration module to obtain motion information. The mobile computer runs the algorithm. In terms of software, based on VS2010 platform, this paper realizes the real-time data acquisition interface of lidar, the motion control interface of traveller No. 2 test vehicle, the information display interface of inertial sensor integration module and the algorithm function selection interface. In the process of experiment, the algorithm of obstacle recognition is verified in dynamic and static environment. Experimental results show that CBFD clustering algorithm (peak point based clustering algorithm) is superior to K-means clustering algorithm and ISODATA clustering algorithm (iterative self-organizing data analysis algorithm) in clustering rationality.
【学位授予单位】：北京理工大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TN958.98

【相似文献】