基于RAC/DR的无人车辆导航定位研究毕业论文
2021-10-27 21:57:59
摘 要
无人驾驶汽车是近几年非常火热的发展趋势,各国都在推进相关方面的研究。但是受限于法律或者成本控制的问题,无人驾驶还难以在大面积领域推广使用,并且近些年屡现无人驾驶汽车发生安全事故。虽然在高速开放场景下的无人驾驶技术遇到瓶颈,但是最近几年,园区无人车由于绕开了道德、法律的约束,正成为无人驾驶技术的突破。作为无人驾驶技术的核心,园区无人车的定位导航系统还是沿用的运用于高速开阔环境下的设备,其高昂的硬件成本阻碍了园区无人车的进一步发展。目前常用的定位技术有:差分技术、航迹推算、惯性导航、地图匹配等,这些技术一定程度上弥补了全球导航卫星定位系统的缺点,但有些问题比如硬件的成本、环境的适应性、运算速度不足等依然有待解决。本文采用的RAC是一种新型GPS接收机,经过一种创新式的天线阵列方式和相应的软件算法,能让水平定位误差控制在10分米级别,不仅提高了定位的精度还提高了定位的稳定性,而且其成本只有几百元,解决了阻碍园区无人车发展的最大痛点。
本文主要研究内容:
(1)研究RAC定位的原理和数据预处理:RAC是实时阵列校准技术的简称,这是一种不使用差分基站且不使用L2或B3精码就可以获得亚米级别精度的技术,其数据格式经过解析后可以获得经度、纬度、高度坐标信息;
(2)研究DR的工作原理和数据预处理:DR是航迹推算的缩写,是一种独立定位方式,其定位依赖于最初的定位精度,其数据格式经过解析后可以得到沿三个方向的加速度、沿三个方向的角速度和里程;
(3)研究将RAC数据和惯导数据融合的方法:首先研究了数据融合的原理,不同传感器的组合方式不同会带来性能上的差异以及不同传感器需要进行同步处理,融合是需要一定的算法,主要介绍了三种卡尔曼滤波流程,最后选择了扩展卡尔曼滤波(EKF);
(4)在计算机上进行软件仿真,分析结果并总结:设计出三段路程,设计了个路程的加速度以及方向,但不涉及转向的过程,然后建立合理的数学模型,依据该模型编写相应的仿真程序,最后分析仿真得到的结果;
(5)进行下一阶段的实验安排设计:现阶段也只是理论上的分析,缺少实验数据的支撑,所以设计了无人小车上传感器安装设计,也设计了实验的大致思路和方法,最后需要将下一阶段的实验结果进行分析总结,获得最佳定位设计(包括传感器安装设计、传感去数量的选择、算法的设计)。
关键词:组合定位;航迹推算;信息融合。
Abstract
Self-driving cars are a very hot development trend in recent years, and all countries are advancing related research. However, due to legal or cost control issues, unmanned driving is still difficult to promote and use in large areas, and in recent years there have been frequent safety accidents in unmanned vehicles. Although the driverless technology in the high-speed opening scene has encountered bottlenecks, in recent years, the park's unmanned vehicles have become a breakthrough in driverless technology because they have bypassed the moral and legal constraints. As the core of unmanned driving technology, the positioning and navigation system of the park's unmanned vehicles is still the equipment used in high-speed and open environments, and its high hardware cost hinders the further development of the park's unmanned vehicles. At present, commonly used positioning technologies are: differential technology, track estimation, inertial navigation, map matching, etc. These technologies make up for the shortcomings of the global navigation satellite positioning system to a certain extent, but some problems such as hardware cost, environmental adaptability, computing Insufficient speed is still to be solved. The RAC used in this article is a new type of GPS receiver. Through an innovative antenna array method and corresponding software algorithm, the horizontal positioning error can be controlled at 10 decimetres, which not only improves the positioning accuracy but also improves the positioning accuracy. Stability, and its cost is only a few hundred yuan, solves the biggest pain point that hinders the development of unmanned vehicles in the park.
The main research content of this article:
(1) Research on the principles of RAC positioning and data preprocessing: RAC is short for real-time array calibration technology. This is a technology that can obtain sub-meter-level accuracy without using differential base stations and without using L2 or B3 fine codes. Its data After the format is parsed, longitude, latitude, and altitude coordinate information can be obtained;
(2) Study the working principle and data preprocessing of DR: DR is the abbreviation of track estimation, it is an independent positioning method, its positioning depends on the initial positioning accuracy, and its data format can be obtained in three directions after analysis Acceleration, angular velocity and mileage in three directions;
(3) Research on the method of fusing RAC data and inertial navigation data: First, the principle of data fusion is studied. Different combinations of different sensors will bring different performance and different sensors need to be synchronized. Fusion requires a certain algorithm , Mainly introduces three Kalman filter processes, and finally chooses Extended Kalman Filter (EKF);
(4) Carry out software simulation on the computer, analyze the results and summarize: design three sections, design the acceleration and direction of the section, but do not involve the process of turning, then establish a reasonable mathematical model, and write the corresponding Simulation program, and finally analyze the results obtained by simulation;
(5) The next stage of experimental arrangement design: At this stage, it is only theoretical analysis, lacking the support of experimental data, so the sensor installation design on the unmanned vehicle is designed, and the general idea and method of the experiment are also designed. Analyze and summarize the experimental results of the next stage to obtain the best positioning design (including sensor installation design, selection of sensor quantity, algorithm design).
Keywords:Combined positioning;Dead reckoning;Information fusion.
第1章 绪论
1.1研究背景与意义
无人驾驶汽车兴起于上世纪五十年代,到了近十几年发展很快并且取得了很多进展,国内外有很多厂商甚至是一些初创科技公司投身于无人驾驶领域,也取得了很多成果。国内的既有如百度、京东、华为等IT公司参与到无人驾驶汽车的研发领域,又有奇瑞、上汽等汽车制造商加入了无人驾驶汽车的研究。国外的也是既有谷歌这样的IT公司早早布局无人驾驶,也有特斯拉这样的后起之秀。毫无疑问的是,在未来,无人驾驶会进入到我们的生活,会极大的影响甚至改变我们的生活方式,为人类提供更大的便利。
然而,近几年来,无人驾驶汽车发生也发生了一些严重的交通事故。在2016年的1月,京港澳高速上,有一辆处于自动驾驶的特斯拉轿车直接撞上了一辆处于作业状态的道路清洁车,造成驾驶员当场死亡,原因是特斯拉的无人驾驶系统未能识别清洁车,以110km/h的车速撞上了清洁车。在2018年的3月,美国亚利桑那州坦贝市一辆优步(Uber)自动驾驶汽车撞死一名骑行女子。同样在2018年,在美国加州山景城的高速公路上,一辆处于自动驾驶状态的特斯拉Model X突然撞上中间的隔离护栏上。以上发生的事故都由于自动驾驶技术仍处于不成熟阶段所造成的,而且带来的破坏也是极大的,一旦发生事故造成的就不只是财产的损失,更是危害到人的生命安全。显然,自动驾驶技术仍然处于发展的阶段,虽然市场给予的期望很高,各国政府都给予了相应的政策,但是一旦遇到这类的安全事故,都无法绕开道德和法律上带来的问题。福特的CEO哈克特在谈到自动驾驶时甚至表示,2021年就推出全自动驾驶“不太实际”。总的来说,受制于城市基建改造和传感器成本,以及自动驾驶的法律法规等,在未来数年内,用于室内服务和物流配送的低速载物无人车,最有可能会率先规模化应用,而更高级别的应用则还需时间。
相比于载客的无人驾驶私家车,低速无人车正风风火火高速发展并且得到实际应用。根据中国汽车技术研究中心数据资源中心管理部项目经理李川鹏透露,低速无人车中的“低速”指的是时速为5~10km,这个对现有传感器、算法都比较适用[1],并且不会产生严重的交通事故。国内做低速无人车的公司很多,诸如百度Apollo、新石器、京东物流等等,而且应用场景非常广泛,包括物流运输、港口矿山、公共交通、环卫作业等,但无一例外都是在园区低速、路线比较固定的场景。仅在2018年,就有智行者自主研发的为首钢园区北京冬奥组委会提供道路清扫和物品运输服务的无人驾驶清洁车“蜗小白”和无人驾驶物流配送车“蜗必达”,中国重汽L4级重型无人驾驶卡车在天津港试运营,百度联合新石器发布的世界上第一台L4级量产无人驾驶物流车在雄安、常州等地试运营,酷哇与中联环境共同合作的无人扫地车在芜湖、合肥、长沙、上海等四所城市试运行。国外的公司同样也在低速无人车领域做出贡献,如在2020年Nuro.Ai在拿到地方政府的许可证后,开始派送他们的无人驾驶送货机器人为得克萨斯州CVS客户运送药品,而他们早在2018年就已经试运营全自动L4级无人配送车。