【作者】 高明；

【导师】 杨力宏； 庞则桂；

【作者基本信息】 西安电子科技大学 ， 电子信息硕士（专业学位）， 2023， 硕士

【摘要】 近年,随着传感器、人工智能与芯片等技术的进步,智能辅助驾驶领域得到了快速发展。用于辅助驾驶领域的车载传感器主要包括毫米波雷达、激光雷达、摄像头、超声波传感器。其中激光雷达具有测距精度高、角辨率高等优点,是用于环境三维成像与目标识别的关键传感器。调频连续波(Frequency Modulated Continuous Wave,FMCW)激光雷达采用相干光电探测体制,可同时对目标距离、速度进行测量,是未来激光雷达阵列化、小型化、芯片化的重要技术发展路线。FMCW相干激光雷达,在发射端采用宽带激光线性调频调制,并在接收端通过光学混频获得外差去斜信号,该去斜信号的频谱信息可计算获得目标距离信息。系统需要对去斜信号进行高速采样,并设计具有距离结算,距离校准,数据打包传输等功能的数字信号处理系统。在辅助驾驶领域要求该系统具有高吞吐低延时,高可靠性的需求。FMCW数据处理的数字系统能够对FMCW激光雷达的输出信号进行数字化处理,进而提取相关信息用于目标探测。本文针对FMCW激光雷达相干探测体制,探究了FMCW激光雷达的运作方式。由于双边带FMCW激光雷达相较于单边带具有更好的抗多径干扰能力,该特性使雷达体制本身具有可靠性,故本文选择双边带线性调频信号作为FMCW信号调制方式,并建立数学模型,推导距离的数据结算方法,设计了数据处理系统的芯片架构,并基于ARM Cortex-M3内核搭建SOC数字系统,确定了以FFT为数据处理核心的FMCW数据处理方法。在FMCW激光雷达数据处理系统中,快速傅里叶变换(Fast Fourier Transform,FFT)将外差信号的时域数据转换为频域数据,通过频率位置计算目标物体的距离信息。结合MCU与丰富外设IP可实现灵活配置参数、数据调度与距离结算等。为此,针对FMCW激光雷达数据的频谱转换,论文研究高吞吐低延时FFT模块的数字电路设计。FFT模块主要包含乘法器、加法器等计算模块与分布式存储等。论文分析数据位宽与量化误差关系,存储资源与效率优化等因素。针对FMCW激光雷达数据的频谱转换,论文研究高吞吐低延时FFT模块的数字电路设计。基于对算法精度、计算效率以及算法可靠性的考量,设计了基-4蝶形计算FFT模块,可获得1024点频谱信息,结合数字控制电路能够结算物体可靠的距离信息。当数字系统工作在极端苛刻工况条件下,如硬件故障,存储器可能会出现访问出错,为增强数字系统的稳定性和对工作环境的适应性,特增加存储器保护电路和看门狗保护电路,通过存储器保护电路硬件冗余降低失效风险,通过看门狗解决程序bug与软件异常处理与复位,以此为数字系统提供错误检测和修复机制,以确保存储器访问和程序运行的正确性。根据验证结果,数据处理单元能够获得1024点频谱信息,高可靠性数字系统能够结算物体距离,对于带宽400MHz,周期400μs的线性调频信号,1004.8米的实际物体距离下,数字处理系统距离测量误差率能够达到1.11%,测量一次距离耗时低于45μs,满足测量需求。算法的对称性和汉明码校验能够提升数据处理的可靠性。看门狗在程序运行过程中提供可靠性保障。对于存储器访问出现的1bit错误能够100%避免,2bit错误能够提示系统做出中断响应,保证了存储器访问的可靠性。综上所述,论文研究高可靠性SOC芯片,集成MCU的控制功能与FMCW激光雷达数字信号处理功能,为芯片级FMCW激光雷达的发展提供支持。更多还原

【Abstract】 In recent years,with the advancement of technologies such as sensors,artificial intelligence and chips,the field of intelligent assisted driving has developed rapidly.The onboard sensors used in the field of assisted driving mainly include millimeter wave radar,LiDAR,cameras,and ultrasonic sensors.Among them,LiDAR has advantages such as high ranging accuracy and high angular resolution,making it a key sensor for environmental 3D imaging and target recognition.Frequency Modulated Continuous Wave(FMCW)LiDAR adopts a coherent photoelectric detection system,which can simultaneously measure target distance and velocity.It is an important technological development path for future LiDAR array,miniaturization,and chip based development.FMCW coherent laser radar adopts broadband laser linear frequency modulation at the transmitter and heterodyne de skewing signal is obtained through optical mixing at the receiver.The spectral information of the de skewing signal can be calculated to obtain target distance information.The system requires high-speed sampling of the de skewed signal and the design of a digital signal processing system with functions such as distance settlement,distance calibration,and data packaging and transmission.In the field of assisted driving,the system is required to have high throughput,low latency,and high reliability.The digital system for FMCW data processing can digitize the output signal of FMCW lidar and extract relevant information for target detection.This article explores the operation mode of FMCW LiDAR based on its coherent detection system.Due to the better anti multipath interference ability of the dual sideband FMCW lidar compared to the single sideband,this characteristic makes the radar system itself reliable.Therefore,this article selects the dual sideband linear frequency modulation signal as the modulation method of the FMCW signal,establishes a mathematical model,deduces the data settlement method for distance,designs the chip architecture of the data processing system,and builds an SOC digital system based on the ARM Cortex-M3 core,The FMCW data processing method with FFT as the core has been determined.In the FMCW lidar data processing system,Fast Fourier transform(FFT)converts the time domain data of heterodyne signals into frequency domain data,and calculates the distance information of target objects through the frequency position.By combining MCU with rich peripheral IP,flexible parameter configuration,data scheduling,and distance settlement can be achieved.For this purpose,the paper studies the digital circuit design of high throughput and low delay FFT modules for frequency spectrum conversion of FMCW lidar data.The FFT module mainly includes calculation modules such as multipliers and adders,as well as distributed storage.The paper analyzes the relationship between data bit width and quantization error,as well as factors such as optimizing storage resources and efficiency.The paper studies the digital circuit design of a high throughput and low delay FFT module for frequency spectrum conversion of FMCW lidar data.Based on considerations of algorithm accuracy,computational efficiency,and algorithm reliability,a B-4 butterfly shaped FFT module was designed,which can obtain 1024 points of spectral information.Combined with a digital system,it can calculate reliable distance information of objects.When the digital system operates under extremely harsh working conditions,such as hardware failures,memory access errors may occur.In order to enhance the stability of the digital system and its adaptability to the working environment,memory protection circuits and watchdog protection circuits are specially added.Hardware redundancy in the memory protection circuit is used to reduce the risk of failure,and watchdog is used to solve program bugs and software exception handling and resetting,This provides error detection and repair mechanisms for digital systems to ensure correct memory access and program operation.According to the verification results,the data processing unit can obtain 1024 points of spectral information,and the highly reliable digital system can calculate the distance between objects.For a bandwidth of 400 MHz and a period of 400 μs The linear frequency modulation signal of s can achieve a distance measurement error rate of 1.11% for the digital processing system at an actual object distance of 1004.8 meters,and the distance measurement time for one measurement is less than 45 μs.Meet measurement needs.The symmetry of the algorithm and Hamming code verification can improve the reliability of data processing.The watchdog provides reliability assurance during program operation.The 1-bit error in memory access can be 100% avoided,and the 2-bit error can prompt the system to make an interrupt response,ensuring the reliability of memory access.In summary,the paper studies a highly reliable SOC chip that integrates the control function of MCU with the digital signal processing function of FMCW LiDAR,providing support for the development of chip level FMCW LiDAR.更多还原