自动驾驶技术概论 | Introduction to Autonomous Driving Technology

第 1 章：自动驾驶技术概论 | Introduction to Autonomous Driving Technology

1.1 引言：从科幻到现实的交通革命

自动驾驶技术的发展历程堪称现代工程史上最激动人心的叙事之一。当1886年卡尔·本茨发明第一辆真正意义的汽车时，恐怕没有人能想象到一个多世纪后，汽车将不再需要人类驾驶员。从威廉·赫伯特·史蒂文森在1921年发明的遥控汽车，到如今装配着数十个传感器、运行着数百万行代码的智能车辆，自动驾驶技术走过了一条漫长而曲折的道路。这项技术不仅代表着汽车工业的颠覆性变革，更将成为重塑城市形态、改变人类出行方式的根本性力量。

The development of autonomous driving technology represents one of the most compelling narratives in modern engineering history. When Karl Benz invented the first truly functional automobile in 1886, few could have imagined that more than a century later, cars would no longer require human drivers. From the radio-controlled car invented by William Herbert Stevens in 1921 to today's intelligent vehicles equipped with dozens of sensors and running millions of lines of code, autonomous driving technology has traversed a long and winding road. This technology not only represents a transformative revolution in the automotive industry but will fundamentally reshape urban forms and transform human mobility patterns.

理解自动驾驶技术，不能仅停留在表面应用层面。从传感器的物理原理到人工智能算法的设计逻辑，从功能安全标准的制定到法规框架的构建，这是一门融合了计算机科学、机械工程、电子工程、数学、心理学乃至伦理学的交叉学科。正如现代汽车已经不再是简单的机械产品，而是“轮子上的计算机”，自动驾驶系统更是将这一趋势推向了极致。本书旨在为读者提供对这一复杂技术体系的系统性理解，从基础理论到前沿实践，从硬件架构到软件算法，从技术挑战到商业落地，逐一展开深入讨论。

Understanding autonomous driving technology cannot stop at the surface level of applications. From the physical principles of sensors to the design logic of artificial intelligence algorithms, from the formulation of functional safety standards to the construction of regulatory frameworks, this is an interdisciplinary field that integrates computer science, mechanical engineering, electronic engineering, mathematics, psychology, and even ethics. Just as modern automobiles are no longer simple mechanical products but "computers on wheels," autonomous driving systems have pushed this trend to its extreme. This book aims to provide readers with a systematic understanding of this complex technological ecosystem, discussing everything from basic theory to cutting-edge practice, from hardware architecture to software algorithms, from technical challenges to commercial implementation.

本章作为全书的开篇，将为读者搭建起整个自动驾驶技术体系的认知框架。我们将首先回顾自动驾驶技术的发展历程，从中理解技术演进的内在逻辑；随后深入剖析SAE（Society of Automotive Engineers）自动化等级标准，这是理解和讨论自动驾驶技术的基础语言；接着系统介绍自动驾驶系统的典型架构范式，包括感知、定位、规划、决策、控制等核心模块的功能划分与协同机制；最后，我们将对比分析传统汽车电子电气架构与新型域控制器架构的差异，帮助读者理解这场正在进行的架构革命。完成本章学习后，读者将具备足够的背景知识来深入探索后续章节中的各项具体技术。

As the opening chapter of this book, this chapter will build a cognitive framework for readers to understand the entire autonomous driving technology ecosystem. We will first review the development history of autonomous driving technology to understand the internal logic of technological evolution; then deeply analyze the SAE (Society of Automotive Engineers) automation level standards, which form the foundational language for understanding and discussing autonomous driving technology; next, systematically introduce typical architecture paradigms for autonomous driving systems, including the functional division and collaboration mechanisms of core modules such as perception, localization, planning, decision-making, and control; finally, we will compare and analyze the differences between traditional automotive electrical/electronic (E/E) architectures and new domain controller architectures, helping readers understand this ongoing architectural revolution. After completing this chapter, readers will have sufficient background knowledge to explore the specific technologies discussed in subsequent chapters in depth.

1.2 自动驾驶技术的发展历程 | The Evolution of Autonomous Driving Technology

1.2.1 早期探索：梦想的萌芽（1920s-1960s）

人类对自动驾驶车辆的想象可以追溯到近百年前。1921年，美国俄亥俄州的两位发明家奥古斯特·库什莫尔和拉尔夫·蒂托成功测试了一辆无线电遥控汽车，这被视为自动驾驶车辆的先驱实验。然而，真正具有里程碑意义的探索来自20世纪50年代。1953年，英国道路研究实验室（Road Research Laboratory）进行了一项引人注目的实验：他们在一辆汽车前方安装了20个线圈组成的检测阵列，通过地面电缆产生的磁场来引导车辆沿着预定路径行驶。这种基于电磁感应的自动驾驶方案虽然简陋，却展示了通过外部基础设施实现车辆导航的可行性和基本原理。

Human imagination about autonomous vehicles can be traced back nearly a century. In 1921, two inventors from Ohio, USA—August Kucsma and Ralph Tieto—successfully tested a radio-controlled car, which is considered a pioneering experiment in autonomous vehicles. However, truly significant exploration came in the 1950s. In 1953, the British Road Research Laboratory conducted a remarkable experiment: they installed an array of 20 coils in front of a vehicle, using magnetic fields generated by ground cables to guide the car along a predetermined path. Although this electromagnetic induction-based autonomous driving scheme was primitive, it demonstrated the feasibility and basic principles of vehicle navigation through external infrastructure.

与此同时，美国通用汽车公司也在积极探索未来交通的愿景。1956年，通用汽车在Motorama展览会上推出了具有未来感的Firebird系列概念车，其中Firebird III概念车畅想了无人驾驶高速公路的场景——驾驶员可以通过嵌入道路的电子向导系统来控制车辆。这种V2I（Vehicle-to-Infrastructure）协同理念在七十多年后的今天依然是智能交通系统研究的重要方向。1960年代，丰田、日产等日本汽车制造商也开始在自动驾驶领域进行初步尝试，尽管这些尝试大多停留在辅助驾驶层面，但其探索为后续技术发展奠定了基础。

Meanwhile, General Motors in the United States was actively exploring visions of future transportation. In 1956, GM unveiled the futuristic Firebird series concept cars at the Motorama exhibition, with the Firebird III concept car envisioning scenes of driverless highways—drivers could control vehicles through an electronic guide system embedded in the road. This V2I (Vehicle-to-Infrastructure) collaboration concept, more than seventy years later, remains an important research direction in intelligent transportation systems. In the 1960s, Japanese automakers such as Toyota and Nissan also began preliminary attempts in autonomous driving. Although these attempts mostly remained at the driver assistance level, their explorations laid the foundation for subsequent technological development.

1.2.2 学术研究的黄金时代（1970s-1990s）

自动驾驶技术从概念验证走向系统化研究，关键转折点发生在1970年代。美国斯坦福大学人工智能实验室的汉斯·莫拉维克（Hans Moravec）被视为现代计算机视觉和自动驾驶研究的先驱。1979年，莫拉维克主持开发了Stanford Cart项目，这是一辆装备了摄像头的移动机器人平台。通过分析摄像头采集的单帧图像，系统需要花费数小时才能完成对单个物体的识别和定位。尽管处理速度极其缓慢，但Stanford Cart展示了通过视觉感知实现自主导航的可行性，这在当时是极具开创性的研究。

The critical turning point for autonomous driving technology moving from concept verification to systematic research occurred in the 1970s. Hans Moravec from Stanford University's Artificial Intelligence Laboratory is regarded as a pioneer of modern computer vision and autonomous driving research. In 1979, Moravec led the development of the Stanford Cart project, a mobile robot platform equipped with cameras. By analyzing single frames captured by the camera, the system required several hours to complete the recognition and localization of a single object. Although the processing speed was extremely slow, the Stanford Cart demonstrated the feasibility of autonomous navigation through visual perception, which was groundbreaking research at the time.

1980年代是学术研究加速发展的时期。1984年，卡内基梅隆大学（CMU）的Navlab项目正式启动，这是迄今为止持续时间最长的自动驾驶研究项目之一。Navlab项目由美国国防高级研究计划局（DARPA）资助，目标是开发能够自主越野行驶的无人车辆。早期的Navlab车辆装备了立体视觉系统、激光测距仪和惯性导航系统，能够在复杂地形中实现自主导航。1986年，Navlab 1在良好路况下实现了每小时31公里的自主行驶速度，这在上世纪八十年代是相当令人瞩目的成就。

The 1980s was a period of accelerated development in academic research. In 1984, Carnegie Mellon University's (CMU) Navlab project officially launched, one of the longest-running autonomous driving research projects to date. The Navlab project, funded by the U.S. Defense Advanced Research Projects Agency (DARPA), aimed to develop unmanned vehicles capable of autonomous off-road driving. Early Navlab vehicles were equipped with stereo vision systems, laser rangefinders, and inertial navigation systems, enabling autonomous navigation in complex terrains. In 1986, Navlab 1 achieved an autonomous driving speed of 31 kilometers per hour under good road conditions, which was quite remarkable in the 1980s.

同一时期，欧洲的自动驾驶研究也在蓬勃发展。1987年，德国慕尼黑联邦国防军大学（Universität der Bundeswehr München）启动了EUREKA PROMETHEUS项目，这是迄今为止欧洲最大的自动驾驶研究计划。该项目汇集了来自多个欧洲国家的科研力量，投入超过7亿欧元资金，持续了长达八年之久。EUREKA PROMETHEUS项目在视觉导航、路径规划、驾驶员辅助系统等领域取得了大量重要成果，其中许多技术已成为现代自动驾驶系统的标准配置。

During the same period, autonomous driving research in Europe