基于多传感器融合的车辆导航系统研究与实现

发布时间：2018-01-13 14:23

本文关键词：基于多传感器融合的车辆导航系统研究与实现　出处：《北京工业大学》2016年硕士论文　论文类型：学位论文

【摘要】：随着科技的发展与社会的进步,城市交通问题越来越受到人们的关注,无人驾驶智能车技术已经成为了解决城市交通问题的重要手段,其中导航定位技术是实现车辆无人驾驶的关键环节。导航定位系统主要采用相应传感器设备采集车辆位姿信息以及车速信息,并发送给执行机构产生相应策略后进行底层控制。本课题着重研究导航系统多传感器融合技术,同时建立了导航系统软件平台,实时显示车辆位姿信息。为无人驾驶智能车提供精确、可靠的导航信息。本文以北京工业大学智能车为实验验证平台,实现校园路况下在线精确采集车辆位姿。主要内容如下:文章对国内外智能车辆研究现状及无人驾驶车导航定位方法的发展与现状进行了探讨与阐述。随后介绍了本文涉及到的两种常见的导航定位系统:全球定位系统(GPS)与惯性导航系统(INS),以及它们的组成、原理及误差分析。根据导航系统建立了车辆运动学模型,并研究了常用的几种运动学模型,根据车辆行驶状况提出了改进的多模型结构,提高了模型的运算速度与准确性。同时,针对无人驾驶智能车在城市道路环境中,GPS信号以非视距(NLOS)传播而造成伪距估计偏差,导致定位不准确的问题,以视觉信息为筛选条件,融入在线采集车辆行驶周围环境的机器视觉信息,结合GPS提供的车体当前位置坐标、航向角、速度值等,有效筛选出因非视距传播而造成的偏差数据。其次,数据融合技术是导航系统能否成功导航定位的关键词素,因此本文对卡尔曼滤波及相关滤波技术进行了深入的研究,并以容积卡尔曼滤波算法(CKF)为基础解决单一滤波算法与智能车系统模型不匹配、鲁棒性不强、容易受到初始值误差影响难以适应无人驾驶智能车运动特点的问题,在CKF基础上增加衰减因子、自适应滤波理论,引入IMM结构,分别克服了历史数据对滤波的影响、系统线性化误差、以及参数设置难以适应模型不确定性的问题。并优化GPS坐标转换算法,更适应无人驾驶智能车实时性的要求。最后,介绍了智能车辆导航系统总体设计方案,基于Nov Atel差分GPS硬件平台,上位机在Visual Studio 2008环境下对采集到的数据进行分析与融合处理,并实现导航系统的可视化效果,将车辆位姿信息下发给底层执行机构完成相应控制。通过实车实验,证明所研究的导航系统适用于无人驾驶智能车辆,验证了本文提出的方法的有效性。
[Abstract]:With the development of science and technology and the progress of society, people pay more and more attention to urban traffic problems. The technology of driverless intelligent vehicles has become an important means to solve urban traffic problems. The navigation and positioning technology is the key link to realize the vehicle unmanned driving. The navigation and positioning system mainly uses the corresponding sensor equipment to collect the vehicle position and posture information as well as the speed information. And sent to the executive mechanism to generate the corresponding strategy for the underlying control. This topic focuses on the study of navigation system multi-sensor fusion technology, and build the navigation system software platform. It can provide accurate and reliable navigation information for driverless smart vehicle. This paper takes the smart vehicle of Beijing University of Technology as the experimental verification platform. The main contents of this paper are as follows:. In this paper, the current situation of intelligent vehicle research and the development and status quo of driverless vehicle navigation and positioning methods are discussed and expounded. Then, two common navigation and positioning systems are introduced in this paper. Global Positioning system (GPS). GPS) and inertial navigation systems (ins). According to the navigation system, the vehicle kinematics model is established, and several commonly used kinematics models are studied, and an improved multi-model structure is proposed according to the vehicle driving condition. The calculation speed and accuracy of the model are improved. At the same time, the error of pseudo-range estimation is caused by the spread of GPS signal in urban road environment by non-line-of-sight (NLOS). The problem of inaccurate positioning is caused by taking visual information as the screening condition, integrating with the on-line acquisition of the machine vision information of the surrounding environment of the vehicle, combined with the current position coordinates, heading angles and velocity values of the car body provided by GPS. The deviation data caused by the non-line-of-sight propagation are screened out effectively. Secondly, the data fusion technology is the key word of whether the navigation system can successfully navigate and locate. Therefore, the Kalman filter and related filtering techniques are studied in this paper, and based on the volume Kalman filter algorithm (CKF) to solve the single filtering algorithm and smart car system model mismatch. The robustness is not strong, and it is difficult to adapt to the motion characteristics of driverless intelligent vehicle by initial error. Based on CKF, the attenuation factor is added, the adaptive filtering theory is introduced, and the IMM structure is introduced. It overcomes the influence of historical data on filtering system linearization error and parameter setting which is difficult to adapt to the uncertainty of the model and optimizes the GPS coordinate transformation algorithm. Finally, the overall design scheme of intelligent vehicle navigation system, based on Nov Atel differential GPS hardware platform, is introduced. The data collected by the upper computer are analyzed and fused in the environment of Visual Studio 2008, and the visualization effect of the navigation system is realized. The position and pose information of the vehicle is sent to the bottom executive to complete the corresponding control. The experimental results show that the navigation system is suitable for the unmanned intelligent vehicle, and the effectiveness of the proposed method is verified.
【学位授予单位】：北京工业大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：U463.67;TP212

【相似文献】