Did Tesla make a decision against "ancestors"?

你听说了吗，特斯拉要改变她的祖先，登上激光雷达！

Did you hear that Tesla is going to change her ancestry and get on the lidar!

众所周知，特斯拉一直是纯视觉自动驾驶仪的坚定支持者。马斯克本人就是一个非常激进的"黑激光雷达"，嗅到了利用这条路线发展无人技术的可能性。在2019年特斯拉的"自主日"产品发布会上，他甚至坦言"任何依赖激光雷达（自动驾驶的发展）的人都是注定要失败的。这些昂贵的传感器是不必要的。它们就像一堆昂贵的附件。你会看到[1]

As we all know, Tesla has always been the staunchest supporter of pure visual autopilot. Musk himself is a very radical "lidar black", sniffing at the development of unmanned technology using this route. At Tesla's "autonomy day" product launch in 2019, he even said frankly that "anyone who relies on lidar (development of automatic driving) is doomed. These expensive sensors are unnecessary. They're like a bunch of expensive appendixes. You'll see[ 1]

然而，在全球科技界，众说纷纭。

However, in the global science and technology circle, there are many different opinions.

来自佛罗里达州的汽车驾驶顾问格雷森·布鲁尔特（Grayson brulte）最近在推特上发布了一张特斯拉携带"疑似"激光雷达设备的照片。

Grayson brulte, an auto driving consultant from Florida, recently tweeted photos of a Tesla carrying a "suspected" lidar device.

作者陈木雷

Chen Mulei, author

总编

General editor

团队规划

Planning of the headlamp team

你说的就是你说的，马斯克

What you say is what you say, musk

好吧，整件事其实是一个大武隆，满脑子都是补脑药，暗无天日。

Well, the whole thing is actually a big Wulong, full of brain tonic, shadowy and groundless.

事后，当地车辆管理局证实，这辆车的车牌确实属于特斯拉，但这纯粹是市场猜测雷达来自哪家制造商。没有官方信息表明特斯拉和卢米纳有合同。唯一能追踪到的是几位"匿名人士"向通讯社提供的"私人信息"，称卢米纳为特斯拉提供了雷达用于"测试和开发"[2]。

Afterwards, the local vehicle authority confirmed that the vehicle license plate really belonged to Tesla, but it was purely market speculation about which manufacturer the radar came from. There is no official information indicating that Tesla and luminar have a contract. The only thing that can be traced is the "private information" provided by several "anonymous persons" to the news agency, saying that luminar provided Tesla with radar for "test and development" [2].

但这些信息既不可核实，也没有问题。

But the information is neither verifiable nor problematic.

首先，这已经不是特斯拉第一次在车顶发现"疑似雷达"装置了。2020年6月，在加州特斯拉总部附近也有类似的目击记录[3]。

First of all, it's not the first time Tesla has been found to have a "suspected radar" device on its roof. In June 2020, there was a similar witness record near Tesla headquarters in California [3].

其次，特斯拉是否必须遵循雷达技术路线来安装雷达？不一定。guidehouse insights的一位分析师表示，特斯拉走激光雷达路线"非常不现实"。

Secondly, does Tesla have to follow the radar technology route to install radar? Not necessarily. An analyst at guidehouse insights said it was "very unrealistic" for Tesla to take the lidar route.

"特斯拉更有可能利用激光雷达技术来测试他们基于摄像头（视觉识别）的自动驾驶系统。如果你改变路线，将导致已经交付的车辆严重贬值——特斯拉无法改装已经售出的100万辆汽车。"2]

"Tesla is more likely to use laser radar technology to test their automatic driving system based on camera (visual recognition). If you change the route, it will lead to a serious devaluation of the vehicles that have been delivered - Tesla can't retrofit one million vehicles that have been sold. "[ 2]

最后，马斯克选择的自动驾驶仪技术非常非常极端——不仅仅是激光雷达，他不想要任何雷达。

Finally, Musk's choice of autopilot technology is very, very, very extreme - not just lidar, he doesn't want any radar.

5月26日，特斯拉宣布将继续推进特斯拉远景的转型：从5月起，为北美市场生产的model 3和model y车型将不再配备雷达。这将是特斯拉首次完全依靠摄像机视觉和神经网络处理来实现自动驾驶、全自动驾驶和一些主动安全功能[4]。

On May 26, Tesla announced that it would continue to promote the transition of Tesla vision: from May, the model 3 and model y models produced for the North American market will no longer be equipped with radar. This will be the first time that Tesla completely relies on camera vision and neural network processing to realize automatic driving, fully automatic driving and some active safety functions [4].

（1） 摄像机：利用计算机视觉识别周围环境和物体，判断前方车辆的距离；

(1) Camera: use computer vision to distinguish the surrounding environment and objects, and judge the distance of the car in front;

（2） 雷达：分为毫米波雷达、激光雷达和超声波雷达。它利用发射波和反射波的时差和相位差来获得目标的位置和速度。

(2) Radar: it is divided into millimeter wave radar, laser radar and ultrasonic radar. It uses the time difference and phase difference between the transmitted wave and the reflected wave to obtain the position and velocity of the target.

但对于L3以上的自动驾驶来说，这远远不够。

But for automatic driving above L3, this is far from enough.

从L2到L3的升级是从辅助驾驶到（有限）无人驾驶的升级。路况监测将由车辆完成，对环境信息的准确性和实时计算能力的要求成倍提高。前端传感器的性能直接决定了系统能否建立精确的路况模型。

The upgrade from L2 to L3 is an upgrade from assisted driving to (limited) driverless driving. The monitoring of road conditions will be completed by vehicles, and the requirements for the accuracy of environmental information and real-time computing capacity increase exponentially. The performance of front-end sensors directly determines whether the system can build an accurate model of road conditions.

对于高级自动驾驶，市场上有两种解决方案：纯视觉和强感知

For high-level automatic driving, there are two solutions in the market: pure vision and strong perception

（1） 视觉解决方案：以摄像头为主导，配合毫米波雷达等低成本传感器实现纯视觉计算。这条路线的主要厂商是特斯拉和百度阿波罗Lite（中国唯一一款自动驾驶纯视野城市道路的闭环解决方案，搭载了威马W6等车型）。视觉路线成本低，商业化简单。缺点是传感器的精度、稳定性和视野都有局限性。依靠算法来弥补硬件缺陷，需要强大的研发能力。

(1) Vision solution: led by the camera, with millimeter wave radar and other low-cost sensors to achieve pure visual computing. The main manufacturers of this route are Tesla and Baidu Apollo Lite (the only closed-loop solution for automatic driving pure vision urban road in China, which is carried on Weima W6 and other models). The cost of visual route is low and the commercialization is simple. The disadvantage is that there are limitations in the accuracy, stability and field of vision of the sensor. Relying on algorithms to make up for the hardware defects requires strong research and development ability.

（2） 强感应方案：以激光雷达为主导，与摄像机、毫米波雷达等传感器相结合。目前，参与这条技术路线开发的企业很多，典型的代表企业有waymo、Uber、百度阿波罗等科技旅游公司。

(2) Strong sensing scheme: led by lidar, combined with camera, millimeter wave radar and other sensors. At present, there are many companies participating in the development of this technology route, and the typical representative enterprises are waymo, Uber, baidu Apollo and other technology and travel companies.

与视觉方案相比，激光雷达具有距离远、分辨率高、受环境光影响小、无需深入学习等优点，这对于在软件开发方面没有优势的传统汽车企业尤为重要。然而，激光雷达的性能和成本正面临瓶颈。虽然被很多车企寄予厚望，但距离大规模组装还很遥远。

Compared with the vision scheme, lidar has the advantages of long range, high resolution, little influence by the environment light, and no need of deep learning, which is particularly important for traditional car enterprises that have no advantages in software development. However, the performance and cost of lidar are facing bottlenecks. Although it is highly expected by many car companies, it is still far from large-scale assembly.

在一些城市，你已经可以体验无人驾驶出租车，这可能是你最接近激光雷达的机会。我们经历过百度的无人驾驶出租车，说真的，有点晕。

In some cities, you can already experience driverless taxi, which may be your closest opportunity to lidar. We have experienced Baidu's driverless taxi, seriously, a little dizzy.

激光雷达由发射系统、接收系统、信号处理和控制系统组成，英文称为激光探测与测距。其基本原理是发射端发射一束激光，光束与物体相遇后，通过漫反射返回接收端，信号处理模块计算出与物体的距离和方位信息。同时，系统通过采集的环境信息绘制三维环境图，精度可达厘米级。

Lidar, which is called laser detecting and ranging in English, is composed of transmitting system, receiving system, signal processing and control system. The basic principle is that the transmitter emits a beam of laser, after the beam meets the object, it returns to the receiver through diffuse reflection, and the signal processing module calculates the distance and orientation information with the object. At the same time, the system draws three-dimensional environment map through the collected environment information, and the accuracy can reach centimeter level.

几种激光雷达路线的比较：

Comparison of several lidar routes:

1.机械旋转激光雷达是目前最成熟的技术方案，已广泛应用于机器人塔西/机器人车和实验车。然而，这种路线雷达的体积是巨大的，而且极其昂贵。高性能产品的平均价格在3000美元以上，只能用于技术验证，不可能商业化。

1. The mechanical rotating lidar is the most mature technical solution, which has been widely used in robotaxi / robobus and experimental vehicles. However, the volume of this route radar is huge and extremely expensive. The average price of high-performance products is more than 3000 US dollars, which can only be used for technical verification and is unlikely to be commercialized.

2.微电子机械系统（MEMS）半固态激光雷达，简称MEMS雷达，是一种成熟的半固态激光雷达。该路线可大大降低生产成本和传感器尺寸，是替代机械雷达的首选路线。

2. Micro electro mechanical system (MEMS) semi-solid lidar, or MEMS radar, is a mature semi-solid lidar. This route can greatly reduce the production cost and sensor size, and is the preferred route to replace mechanical radar.

3.固态激光雷达是一种没有任何机械运动部件的产品。这是最不成熟的技术类型，但它有着光明的前景。目前，有两条清晰的路线：光学相控阵（OPA）和快速大面积扫描硬件（flash）。

3. Solid state lidar is a kind of product without any mechanical moving parts. It is the most immature type of technology, but it has a bright future. At present, there are two clear routes: optical phase array (OPA) and fast large area scan hardware (flash).

其中，OPA雷达具有扫描速度快、精度高、体积小、可控性强、抗振动能力强等一系列优点，具有相对清晰的路径，降低了成本。它被认为是未来最有希望成为主流激光雷达的产品形态。

Among them, OPA radar has a series of advantages, such as fast scanning, high precision, small size, strong controllability and strong anti vibration, and has a relatively clear route to reduce the cost. It is considered as the most promising product form to become the mainstream laser radar in the future.

闪光雷达是目前唯一没有扫描系统的技术路线。其工作原理接近数码相机，能够满足最高级别的车辆规格要求。但是，它存在着距离短、分辨率不够的缺点。去年年底，长城汽车宣布旗下旗舰SUV将搭载"全球首款全固态激光雷达"IBEO4D，其探测范围为130米。

Flash radar is the only technical route without scanning system at present. Its working principle is close to that of digital camera and can meet the requirements of the highest level of vehicle specifications. However, it has the defects of short range and insufficient resolution. At the end of last year, Great Wall Motor announced that its flagship SUV would be equipped with ibeo 4D, the "world's first all solid-state lidar", which has a detection range of 130 meters.

激光雷达，一种"现金能力"

Lidar, a "cash capability"

如果仅从技术角度看，激光雷达确实是一项非常好的技术，甚至可能是目前传感硬件的最佳解决方案。

If only from the perspective of technology, lidar is indeed a very good technology, and may even be the optimal solution of the current sensing hardware.

特斯拉坚持纯视觉路线与马斯克本人有关，但不一定是理性的选择。

Tesla's insistence on a purely visual route is related to musk himself, but not necessarily a rational choice.

特斯拉的技术人员曾经说过，"人们可以用纯视觉驾驶汽车。为什么汽车不能呢？这种观点乍一看是合理的，但却是诡辩的。人们之所以依赖视觉，是因为他们没有其他复杂的感觉器官，但汽车没有生理上的限制——当汽车配备了设备时，它们就完蛋了。

Tesla's technicians once said, "people can drive cars with pure vision. Why can't cars?" This view is reasonable at first sight, but it is a sophistry. The reason why people rely on vision is that they don't have other complex sensory organs, but cars don't have physiological limitations -- they are finished when they are equipped with equipment.

配备各种传感器的汽车是目前业界的主流，比如特斯拉的"一路走黑"却只有这一款。使用不同的传感器设备来应对不同的驾驶场景，这无疑在灵活性和安全性方面更加安全，这些都是自动驾驶必须解决的问题。

A car equipped with a variety of sensors is the current mainstream of the industry, such as Tesla "one way to the black" but only this one. Using different sensor devices to deal with different driving scenarios, which is undoubtedly more secure in terms of flexibility and safety, these are what automatic driving must solve.

那么，阻碍激光雷达上车的核心障碍是什么呢？

So what is the core obstacle that prevents lidar from getting on the bus?

成本。

Cost.

激光雷达太贵了，早期甚至要几万元。据说百度自动驾驶汽车使用的Velodyne机械式激光雷达价格高达70万元/[7]——一套设备比车贵，这个价格绝对不可能商业化。

Lidar is too expensive, even tens of thousands of dollars in the early days. It is said that the price of Velodyne mechanical lidar used by Baidu autopilot car is as high as 700 thousand yuan / [7] - a set of equipment is more expensive than the vehicle, and this price is absolutely impossible to commercialize.

因此，长期以来，激光雷达路由的最大方向就是降低成本。以色列雷达制造商innoviz首席执行官奥马尔·凯拉夫（Omer keilaf）在2020年底接受媒体采访时表示，如今的激光雷达组件有1000美元的选择，显然有一条技术路线可以将价格降至500美元[8]。

Therefore, for a long time, the biggest direction of lidar route is to reduce the cost. Omer keilaf, chief executive officer of innoviz, an Israeli radar manufacturer, said in an interview with the media at the end of 2020 that today's lidar components have a $1000 option, and it is clear that there is a technical route to reduce the price to $500 [8].

制造更便宜的雷达也是国内制造商的方向。90后汽车评测团联合创始人王若兰告诉《聚光灯》记者："目前，一台雷达的硬件成本可以达到4000元左右，明年将量产。但软件包肯定不止这些。最终价格取决于议价能力。"

Building cheaper radars is also the direction of domestic manufacturers. Wang Ruoran, co-founder of the post-90s car evaluation group, told the spotlight: "at present, the hardware cost of a radar can reach about 4000 yuan, and it will be mass produced next year. But the software package is definitely more than that. The final price depends on the bargaining power. "

但从市场角度看，价格还是不便宜。如果一辆车配备三个激光雷达和一个匹配算法，整个方案的价格仍能达到2万元左右。这对于均价在10万元至20万元之间、利润微乎其微、终端买家对价格极为敏感的国内车企来说，恐怕是无法接受的。

But from a market point of view, the price is still not cheap. If a car is equipped with three lidars and a matching algorithm, the price of the whole scheme can still reach about 20000 yuan. This may not be acceptable to domestic car companies whose average price is between 100000 yuan and 200000 yuan, whose profit is negligible, and whose terminal buyers are extremely sensitive to the price.

因此，最先安装激光雷达的不是新车制造商，而是奥迪。早在三年前，奥迪就为旗下高端轿车安装了4系机械式激光雷达，并号称实现"全球首款量产L3轿车"；2021年，一批国产新车将配备各种新型的、成本更可控的激光雷达，这已成为车企展示肌肉、让自己看起来比别人更聪明的"财富密码"。

Therefore, it was not the new car makers who first installed lidar, but Audi. As early as three years ago, Audi installed 4-line mechanical lidar for its high-end cars, and claimed to realize "the world's first mass production L3 car"; In 2021, a batch of new domestic cars will be equipped with various new and more controllable cost lidars, which has become the "wealth code" for car companies to show their muscles and make themselves look more intelligent than others.

答案是否定的。

The answer is No.

自动驾驶作为一项技术，无疑具有巨大的潜在商业价值。但有时技术落地不仅是一个商业问题，更是一个复杂的社会问题——特别是对于自动驾驶来说，它具有巨大的社会影响，甚至对人与技术的关系产生颠覆性的影响。

As a technology, automatic driving undoubtedly has great potential commercial value. But sometimes technology landing is not only a commercial problem, but also a complex social problem - especially for autonomous driving, which has a huge social impact, and even has a subversive impact on the relationship between people and technology.

技术不能孤立地存在于我们的社会中。它需要与人类互动，以实现其作为工具的价值。但技术嵌入社会的过程本身就是一个复杂的系统工程，超越了技术：如何运用新技术伦理规范新技术的使用？如何建立人们对科技的信任？如何将受限场景中的新技术植入到常见场景中？配套的法律法规能否与时俱进？重大事故发生后，如何修复人们对技术的不信任？

Technology cannot exist in our society in isolation. It needs to interact with human beings to realize its value as a tool. But the process of technology embedded in society itself is a complex systematic project beyond technology: how to use new technology ethics to regulate the use of new technology? How to build people's trust in technology? How to implant new technologies from restricted scenarios into common scenarios? Can the supporting laws and regulations keep up with the times? After a major accident, how to repair people's distrust of technology?

毫无疑问，这些大问题不是单个企业能解决的，特斯拉和百度也不能。

There is no doubt that these grand issues can not be solved by a single enterprise, neither can Tesla nor Baidu.

人脸识别在技术层面上并不是特别复杂，在技术滥用和隐私侵权方面引起了足够的讨论。更有甚者，可以重塑人机关系的自动驾驶技术：当自动驾驶仪掌舵时，如何认定交通事故责任？光是这一点，恐怕就足以让一代律师献出自己的头发。

Face recognition, which is not particularly complicated at the technical level, has caused enough discussion on technology abuse and privacy infringement. What's more, autopilot technology, which can reshape the human-computer relationship: how to identify the responsibility for traffic accidents when autopilot is at the helm? This alone is probably enough for a generation of lawyers to give their hair.

高水平的自动驾驶技术不仅需要自身与相关产业的协调发展，更需要从官方到民间全面系统的制度建设。

What high-level automatic driving technology needs is not only the coordinated development of itself and related industries, but also the comprehensive and systematic system construction from the official to the private.

在此之前，谈论任何大规模的商业化还为时过早。

Before that, it's too early to talk about any large-scale commercialization.

参考文献：

References：

[1] 安德鲁J。霍金斯：这是埃隆·马斯克和其他所有人争夺全无人驾驶汽车的比赛。边缘。2019.04.24.

[1] Andrew J. Hawkins：IT'S ELON MUSK VS. EVERYONE ELSE IN THE RACE FOR FULLY DRIVERLESS CARS. The Verge. 2019.04.24.

https://www.theverge.com/2019/4/24/18512580/elon-musk-tesla-driverless-cars-lidar-simulation-waymo

https://www.theverge.com/2019/4/24/18512580/elon-musk-tesla-driverless-cars-lidar-simulation-waymo

（2）Edward Ludlow、Gabrielle Coppola和Dana Hull：特斯拉测试发光传感器麝香被嘲笑为"傻瓜的差事"。彭博社。2021.05.24.

[2] Edward Ludlow, Gabrielle Coppola & Dana Hull: Tesla Tests Luminar Sensor Musk Scorned as 'Fool's Errand'. Bloomberg. 2021.05.24.

https://www.bloomberg.com/news/articles/2021-05-24/tesla-testing-luminar-laser-sensor-musk-called-fool-s-errand

https://www.bloomberg.com/news/articles/2021-05-24/tesla-testing-luminar-laser-sensor-musk-called-fool-s-errand

[3] 阿德利亚·切利尼·莱因克尔：在激光雷达测试中，特斯拉加快了自动驾驶仪的改变。《投资者商报》。2021.05.25.

[3] ADELIA CELLINI LINECKER: Tesla Steps Up Change To Autopilot Amid Lidar Test. Investor`s Business Daily. 2021.05.25.

https://www.investors.com/news/elon-musk-tesla-ceo-changing-his-mind-on-lidar-sensors/

https://www.investors.com/news/elon-musk-tesla-ceo-changing-his-mind-on-lidar-sensors/

[4] 过渡到特斯拉愿景。特斯拉。

[4] Transitioning to Tesla Vision. Tesla.

https://www.tesla.com/support/transitioning-tesla-vision

https://www.tesla.com/support/transitioning-tesla-vision

[5] 倪正阳：激光雷达的成本有望降低，配套车辆的运量也即将到来。德邦证券。2021.04.29

[5] Ni Zhengyang: the cost of lidar is expected to be reduced, and the volume of supporting vehicles is coming soon. Deppon securities. 2021.04.29

[6] 何俊义、程思琦、陶义军：《激光雷达：价值、空间、模式与趋势》。中信建投。2021.02.24

[6] He Junyi, Cheng Siqi & Tao Yijun: lidar: value, space, pattern and trend. China CITIC construction investment. 2021.02.24

[7] 中国电子报：关于苹果的汽车制造，这篇文章就足够了。2020年12月28日

[7] China Electronic News: about Apple's car making, this article is enough. December 28, 2020

https://new.qq.com/rain/a/20201228A0N8VL00

https://new.qq.com/rain/a/20201228A0N8VL00

[8] 约翰·科齐尔：激光雷达首席执行官：埃隆·马斯克5年前做出了正确的选择。。。今天不行。福布斯2021.02.27。

[8] John Koetsier: Lidar CEO: Elon Musk Made The Right Choice 5 Years Ago ... Not Today.Forbes.2021.02.27.

https://www.forbes.com/sites/johnkoetsier/2021/02/27/lidar-ceo-elon-musk-made-the-right-choice-5-years-ago--not-today/?sh=2a4dcfa1141b

https://www.forbes.com/sites/johnkoetsier/2021/02/27/lidar-ceo-elon-musk-made-the-right-choice-5-years-ago--not-today/?sh=2a4dcfa1141b

[9] 高登、于海宁、高一楠：华为"赋能"汽车系列之二：华为激光雷达——龙入海，天眼开启。长江证券。2021.05.19

[9] Gaodeng, Yu Haining & Gao Yinan: Huawei's "enabling" Car Series II: Huawei lidar - Dragon into the sea, the eye of the sky is opening. Changjiang Securities. 2021.05.19

Link：https://new.qq.com/omn/20210603/20210603A046B200.html

update time：2021-06-03 18:49:16