Guide: at present, the popularity of automatic driving comes partly from the broad market prospect and consumption expectation, and partly from the media and enterprises’ false praise of the whole market, including the AI technology used in automatic driving.

Since its birth, the concept of autonomous driving has always been one of the areas favored by capital and technology entrepreneurs. With the gradual maturity of 5g business, the popularity of the field of automatic driving is also high. According to the data in 2018, the total financing in the field of automatic driving at home and abroad is more than 7 billion US dollars, about 1.15 billion US dollars at home and more than 5.85 billion US dollars abroad.

At present, the popularity of automatic driving comes partly from the broad market prospect and consumption expectation, and partly from the media and enterprises’ false praise of the whole market, including the AI technology used in automatic driving.

It can be seen from the above related financing information that, in the field of autonomous driving, except for Baidu, a company with big money does not need financing. For start-ups, each financing amount is quite large.

Especially in the United States, Aurora obtained US $530 million in financing in February 2019, setting a new single round financing record for independent autonomous driving companies in the United States. Before long, nuro obtained US $940 million in single round financing from Softbank’s vision fund, which directly pushed the financing level to the peak.

Back home, the epidemic has not affected the flow of capital in the field of automatic driving at all, but increased the use scenarios of automatic driving, such as the contactless transportation of materials in epidemic areas. In February this year, (Xiaoma Zhixing) obtained $400 million of round B financing from Toyota, with a total financing amount of nearly $800 million. Another automatic driving company, momenta, has also obtained more than $200 million of financing since its establishment. The investors include Tencent, dailyme group, Zhenge fund and other high-quality capital. After obtaining the capital, these enterprises are more in-depth research and development in the field of automatic driving, that is, “burning money”, without large-scale chemical reaction.

Not only that, the industry infighting is also very lively. In the last month, the so-called “North American first case of automatic driving” and “China first case of automatic driving” have caused a lot of noise in the industry, which is actually a patent and technology dispute. As the author mentioned above, there is no essential breakthrough in automatic driving, and the core technology and talents are the most powerful tools for enterprises.

So, what are the mainstream players of autopilot? What are the differences in core technologies? It should be noted that this paper introduces the list of companies from the California Vehicle Administration (DMV) of the United States, which released the “report on autonomous driving takeover in 2019”. This report mainly evaluates the technical strength by the number of vehicles, mileage and the number of manual takeover. Although the standard is highly questioned, DMV is still an objective and open quantitative report on autonomous driving in the United States.

“Autonomous driving takeover report 2019” released by California Vehicle Administration (DMV)

Top five players


Baidu’s automatic driving business can be said to be developing very rapidly, from No. 7 in 2018 to No. 1. Baidu’s Apollo platform is mainly used as a software platform for automakers and autopilot to complete the automatic driving system by combining its software with the vehicle and hardware of the depot. It is worth mentioning that Apollo won the bid for a new infrastructure project on March 24 this year, and will participate in the construction of a two-way about 10 km road collaborative test demonstration area in Yangquan economic development zone.


Waymo has become famous in the field of automatic driving. It was first hatched as an automatic driving project of Google in 2009, and became a subsidiary of alphabet in 2016. Waymo maintains the current record of financing in the industry. In March this year, waymo announced that it had obtained $2.25 billion in financing. Waymo has a large overlap with Baidu in business, providing auto driving kit support for car companies, as well as driverless passenger travel service and unmanned freight service. It should be pointed out that in fact, waymo belonged to the purchaser in the early development direction, that is, to purchase vehicles. Starting from 2018, waymo purchased 20000 Jaguar i-pace and 62000 Pacifica from FCA. After obtaining US $2.25 billion in financing, waymo began to act as Party B, cooperating with car companies to provide automatic driving solutions. Not long ago, waymo also demonstrated its fifth generation automatic driving platform, in which self-developed hardware such as lidar and camera are the core technologies of waymo.


Gmcruise (hereinafter referred to as cruise) is GM’s autopilot company. Its predecessor was cruise automation, a software and hardware R & D company focusing on autopilot. In 2016, it was acquired by GM with us $1 billion. Meanwhile, in 2018, Softbank’s vision fund invested US $2.25 billion in gmcruise. On January 21 of this year, after three years of cooperation among GM, Honda and cruise, cruise released its first driverless car named “origin”. Meanwhile, cruise is the first company in the world to release L4 level automatic driving technology. The author believes that cruise’s R & D strength in automatic driving software and hardware, together with the position of general engine factory, and combined with Softbank’s layout in the field of artificial intelligence, can be expected in the future.


Autox was founded in Silicon Valley in 2016, mainly focusing on L4 level automatic driving solutions. It is understood that the company takes civilian as its research and development standard. Specifically, it only uses cameras to realize automatic driving without radar, laser, ultrasonic and other sensors, so as to reduce costs. Compared with the previous ones, AutoX positioning is simple. It is just an autopilot system provider, who claims to build Windows in the car. In other words, Tesla is the same as using the algorithm to solve the problem. But for now, autox has changed its mind and started to use lidar and millimeter wave radar.

Almost at the same time as autox, In 2019, he also obtained the automatic driving license from the local government of California and was founded by two former Baidu executives. And pony, like cruise, has the same investment background as a car factory. As mentioned above, Toyota has taken a stake in pony. Pony has developed software and hardware, including optical radar and infrared sensor, positioning and high-precision map, simulation platform, data storage and processing platform. In November 2019, pony launched the first robotaxi service for the general public in California in the United States.

In the form, we can see that zoox is still in the list, but compared with last year, it dropped from the third place to the seventh place this year, which is not very good-looking. In addition, it was revealed that due to the impact of the local epidemic in the United States, zoom has laid off 120 employees. The financial crisis and operational challenges it faces are very big.

It is worth mentioning that the domestic didi company has followed zoox in the takeover rate. This is the first time didi has participated in the DMV test, and it appears in the list. To sum up, there are five Chinese autonomous driving enterprises (Baidu, autox , Didi, plusai) appear in the top ten list, accounting for half of the country, which also reflects some achievements in the field of automatic driving in China. However, it remains to be seen how large-scale chemical reactions can be generated in the future.

So how far is automatic driving?

According to the standard of DMV, the less vehicles drivers need to take over, the higher the ranking of enterprises, indicating that the more close the automatic driving technology of enterprises is to the real automatic driving. In fact, the above five enterprises are basically L4 level automatic driving, so what are L3 and L4?

Previously, the National Highway Safety Administration (NHTSA) and saeinternational (International Association of Automotive Engineers) both divided the automatic driving standard into l0-l5, but there were some differences between the two standards on L4. Until September 2016, the US Department of transportation clearly established saeinternational j3016 as the standard to define automatic / automatic driving vehicles The global industry reference standard is used to evaluate the automatic driving technology. So far, many automotive industry related enterprises in the world have also adopted saej3016 to define their own related products.

On the division of automatic driving standards

Under the standard of saei international, the author makes a more specific term explanation table for the standard (as shown in the figure above). From the table, l0-l2 is not difficult to understand. L0 is the most common pure manual driving car, L5 is the ideal state of unmanned driving, the system can adapt to any state of vehicle driving, completely replace human. The problem is that L4 can’t adapt to all working conditions on L4 and L5, so another design concept of automatic driving is introduced here, that is, odd (design scope of application, including road type, geographical range, speed range, environmental climate and other factors). If a L4 level vehicle is allowed to drive in its designed odd, it can be called “L5” in a specific state. The real sense of L5 is that the range of odd is infinite, covering all working conditions. To realize L5, we need machine learning, so strictly speaking, the intelligence level of L5 is 99.9999 %(you can understand the amount of data required for this level.).

Therefore, in the above-mentioned autonomous driving enterprises, although their established goal is L4, we don’t know how much data there are in terms of testing and machine learning. But we can be sure that the complexity of working conditions on earth will cause great trouble to the research and development of L4.

It should be noted that according to this list and the introduction, the public’s vision is easily biased, and they will think that automatic driving has entered the research and development stage of L4. The real situation is that most of the dozens of automatic driving enterprises in the world are actually doing L2 level automatic driving. Not all enterprises are born with the golden key like waymo, nor do they have the financial resources that Baidu can research L3 and L4 levels at the same time, so more enterprises can only start from L2 when their financial and technical background strength is not so strong One side of the hematopoiesis, one blood transfusion upgrade.

So, how far is L5 fully automatic driving? The answer is, it’s too early to ask. In 2019, the penetration rate of L2 is less than 30%. Even if L4 and L5 come out tomorrow, how many consumer groups are popularized? Even if they are popularized, how many people dare to drive? There are enough cases of Tesla “hands off the steering wheel, head on the waist.”.

Editor in charge: yyx

Leave a Reply

Your email address will not be published. Required fields are marked *