【作者】 李波；

【导师】 马天飞；

【作者基本信息】 吉林大学 ， 车辆工程， 2023， 硕士

【摘要】 自动泊车系统可以控制车辆自动寻找车位并完成泊车,能够有效减小驾驶人的驾驶压力。然而自动泊车系统具有测试过程复杂度高、测试场景难以穷尽等特点,导致传统的场地测试难以满足自动泊车系统的测试需求,因此一种便捷高效的自动泊车系统测试方法亟待研究。本文依托国家重点研发计划“智能汽车复杂场景数字-物理融合模拟测试技术研究”(编号:2022YFB2503402),搭建超声波雷达在环测试平台,并以此为基础构建虚拟测试场景,对基于超声波雷达执行环境感知任务的自动泊车系统进行硬件在环测试。首先建立了考虑大气条件影响的超声波雷达数字模型,并构建了超声波雷达在环测试平台,然后提出了基于贝叶斯优化原理的加速测试场景生成方法,最后基于超声波雷达在环测试平台进行了自动泊车系统的自动化测试。本文具体研究内容如下:(1)超声波雷达建模根据超声波雷达的工作机理对超声波雷达模型进行需求分析,根据建模需求将超声波雷达模型分为检测范围模型和检测距离模型两个部分进行建模。在给定的场景中对超声波传播过程中的各种能量损失进行量化,建立大气条件与超声波雷达探测范围的函数关系,并对函数中的未知参数进行标定,进而推导超声波雷达的探测范围;提出虚拟仿真环境下目标物距离真值的获取方法,并求解给定大气条件下的真实声速,将距离真值去理想化修正为超声波雷达实际探测距离值;最后对建立的超声波雷达模型进行实车验证。(2)超声波雷达在环测试平台构建设计超声波雷达在环测试平台总体架构,确定超声波雷达模型、超声波雷达硬件、超声波回波模拟器、虚拟仿真场景、车辆动力学模型、被测自动泊车算法等测试组件的闭环交互逻辑;基于Car Sim、MATLAB/Simulink、Pre Scan等软件构建测试平台的软件系统;利用Park Pilot URF6型超声波雷达、PC机、超声波回波模拟器、PXIe-8880型嵌入式控制器等硬件组件搭建测试平台的硬件系统,基于用户数据报协议(User Datagram Protocol,UDP)以及控制器局域网络(Controller Area Network,CAN)完成各硬件组件间的数据通信;最后为提高测试效率和节省人力成本,提出测试平台的自动化测试方法。(3)面向自动泊车系统的测试场景生成方法研究基于PEGASUS项目提出的层级模型以及自动泊车系统的测试需求定义功能场景;提取功能场景关键字并映射为相应的场景参数,根据各场景参数的内在联系确定逻辑场景的组成要素,依据国内外相关测试法规确定各场景要素的参数空间,组成逻辑场景;为提高测试效率,将具体场景生成问题建模为以泊车过程中的危险程度为目标函数的高维度非线性优化问题,并利用贝叶斯优化方法进行求解,实现自动泊车测试场景的加速生成。(4)测试试验与分析基于超声波雷达在环测试平台对自动泊车系统以及测试场景生成方法进行测试与分析。利用随机生成的测试场景分别对自动泊车系统进行硬件在环测试和纯软件仿真测试,对比两者的测试结果说明超声波雷达硬件在环测试平台的优势;基于硬件在环测试平台分别对自动泊车系统进行优化测试与组合测试,得到了自动泊车系统的危险场景集合,测试结果表明,相比于组合测试,优化测试在生成的危险场景覆盖度高于85%的前提下分别将平行泊车工况以及垂直泊车工况的危险场景生成效率提高到组合测试的2.2倍、4.9倍。更多还原

【Abstract】 The automatic parking system can control the vehicle to automatically find the parking space and complete the parking,which can effectively reduce the driving pressure of the driver.However,the automatic parking system has the characteristics of high complexity of the test process and inexhaustible test scenarios,which makes the traditional site test difficult to meet the test requirements of the automatic parking system.Therefore,a convenient and efficient automatic parking system test method needs to be studied urgently.Supported on the National Key R&D Program of China(2022YFB2503402),this paper has built an ultrasonic-radar-in-the-loop test platform,and based on this,a virtual test scene is built to conduct hardware-in-the-loop test on the automatic parking system based on the ultrasonic radar to perform environmental awareness tasks.Firstly,the digital model of ultrasonic radar considering the influence of atmospheric conditions is established and the ultrasonic-radar-in-the-loop test platform is constructed,then an accelerated test scenario generation method based on Bayesian optimization principle is proposed.Finally,based on the ultrasonic-radar-in-the-loop test platform,the automatic test of the automatic parking system is carried out.The specific research contents of this paper are as follows:(1)Ultrasonic radar modelingQuantify various energy losses in the process of ultrasonic propagation in a given scenario,establish the functional relationship between atmospheric conditions and the detection range of ultrasonic radar,calibrate the unknown parameters in the function,and then deduce the detection range of ultrasonic radar;A method to obtain the true distance value of the target in the virtual simulation environment is proposed,and the true sound velocity under the given atmospheric conditions is solved.The true distance value is de-idealized and corrected to the actual detection distance value of the ultrasonic radar.Finally,the ultrasonic radar model is verified by the real vehicle.(2)Construction of ultrasonic-radar-in-the-loop test platformDesign the overall architecture of the ultrasonic-radar-in-the-loop test platform,and determine the closed-loop interaction logic of the ultrasonic radar model,the ultrasonic radar hardware,the ultrasonic echo simulator,the virtual simulation scene,the vehicle dynamics model,the tested automatic parking algorithm and other test components;The software system of the test platform is built by Car Sim,MATLAB/Simulink,Pre Scan and other software;The hardware system of the test platform is built by using hardware components such as Park Pilot URF6 ultrasonic radar,PC,ultrasonic echo simulator,PXIe-8880 embedded controller,and the data communication between hardware components is completed based on UDP(User Datagram Protocol)and CAN(Controller Area Network);Finally,in order to improve test efficiency and save labor costs,an automated test method for the test platform is proposed.(3)Research on test scenario generation method for automatic parking systemDefine functional scenarios based on the hierarchical model proposed by PEGASUS project and the test requirements of the automatic parking system;Extract the functional scenario keywords and map them to the corresponding scenario parameters,determine the components of the logical scenario according to the internal relationship of each scenario parameter,determine the parameter space of each scenario element according to the relevant domestic and foreign test laws and regulations,constitute the logical scenario;In order to improve the test efficiency,the specific scene generation problem is modeled as a high-dimensional nonlinear optimization problem with the degree of danger in the parking process as the objective function,and solved by Bayesian optimization method to realize the accelerated generation of automatic parking test scenarios(4)Test and verificationThe automatic parking system and test scenario generation method are tested and analyzed based on the ultrasonic-radar-in-the-loop test platform.The hardware-in-theloop test and software-only simulation test of the automatic parking system are carried out by using the randomly generated test scenarios.The comparison of the test results of the two shows the advantages of the hardware-in-the-loop test platform of the ultrasonic radar;Based on the hardware-in-the-loop test platform,the optimization test and combination test of the automatic parking system are carried out respectively,and the set of dangerous scenes of the automatic parking system was obtained.The test results showed that compared with the combined test,the optimization test improved the generation efficiency of dangerous scenes in parallel parking conditions and vertical parking conditions to 2.2 times and 4.9 times respectively,on the premise that the coverage of the generated dangerous scenes was higher than 85%.更多还原