Shulou(Shulou.com)11/24 Report--

"Intelligence has come to a fork in the road, is it to reduce the cost of autopilot to stay in L2 or L2, or to increase safety redundancy to L3?"

During the recent 2023 China Automobile Forum, Zhu Xichan, a professor at Tongji University and director of the Institute of Automotive Safety Technology, said. At a time when L2 is no longer rare and L4 may be a long way off, the progress of L3 has been closely watched by the industry. After all, only after crossing the "watershed" of L3 can we officially enter the door of autopilot.

A few days ago, Xin Guobin, vice minister of the Ministry of Industry and Information Technology, said: "conditional autopilot (L3 level autopilot) and the commercial application of higher-level autopilot features will be supported, and a new version of the Intelligent Networked Automotive Standard system Guide will be released. We have started this work, and the local authorities are responding positively."

In addition, the Ministry of Industry and Information Technology revealed in mid-May that the Guide to the Intelligent Network Automotive Standard system will be officially released soon, speeding up the formulation of more than a dozen key and urgently needed standards, including supporting the commercial application of self-driving functions above level L3.

The further support of the current policy may inject new vitality into the high-level self-driving industry, which is in a difficult situation, and let the relevant enterprises see the hope of getting out of the predicament. Around this phenomenon, this paper will discuss three questions:

1. What are the twists and turns in the development of L3 self-driving?

2. What is the biggest obstacle to the commercial landing of L3?

3. The L3 standard is coming, does it mean that China's autopilot will officially begin the L3 era?

1. Stuck in the "valley of terror" in 2017, it can be called the "highlight moment" of L3 autopilot.

In July of that year, Audi released the Audi A8, the world's first new car capable of L3 self-driving, which is also defined as the first mass-produced car developed specifically for L3 self-driving.

Subsequently, some domestic car companies began to follow. Geely, SAIC, Changan and Guangzhou Auto have all revealed plans for L3 autopilot mass production models. SAIC Roewe MARVEL X Pro and Guangzhou Auto Ean AION LX released in 2019 already have L3 self-driving functions, and Zhu Huarong, chairman of Changan Automobile in March 2020, also personally endorsed UNI-T equipped with China's first production-level structural road L3 self-driving system.

Almost at the same time as passenger cars, in the commercial vehicle market, some technology companies incubated by heavy truck companies have emerged, targeting the creation of L3 self-driving heavy trucks, and a small number of pre-installed products have been taken offline.

However, it was reported in 2020 that Audi abandoned the world's first L3 self-driving mass production program after nine years, a thousand-person team and 1 billion euros of research and development costs.

This is seen as a landmark event of the L3 failure of the whole car factory.

After the release of the report, although Audi responded that it did not cancel the L3 self-driving technology research and development project, but only promoted it to the Volkswagen group level and integrated into the Car.Software software vehicle development department, the industry controversy over L3 has not stopped.

Car companies marching into the L3 have started to mention less about the L3 one after another since 2021. Some companies even chose to skip L3 and develop L4 / L5 autopilot technology directly from the beginning.

For example, the ideal is to skip the development of L3 autopilot and go straight to the L4 Robotaxi. Foreign car companies, including Ford and Volvo, have publicly abandoned the L3 and chose to leap directly from the L2 to the L4.

Volvo CEO Hakan Samuelsson believes that the L3 is unsafe. Volvo is currently working with Uber to develop L4 autopilot and has offline the XC90 with L4 autopilot hardware for Uber testing.

On the other hand, more concepts such as L2 + and L2 + began to appear. It is estimated that Mercedes-Benz is the only one that has really landed the L3 at present. Mercedes-Benz has announced that its self-driving assistance system, Drive Pilot, will be ordered in Germany on May 17, and the S-Class and EQS models will have L3 self-driving capabilities.

Today, when autopilot technology is getting closer and closer to L3, domestic manufacturers seem to hover between L2 and L3 for a long time, which seems to be an insurmountable gap. The industry calls it the "Valley of Terror". Zhu Xisan once said, "instead of getting out of the valley of terror, autopilot is getting deeper and deeper in the valley of terror."

Why did this happen?

2. Responsibility definition in March 2018, a Tesla Model X automatically changed lanes while activating autopilot on Interstate 101 in Mountain View, California, and hit a ramp at 120km / h, killing the driver. In March 2019, a Tesla Model 3 in Florida failed to identify the truck ahead, neither automatically changing lanes nor prompting the driver to pass under the truck at a speed of 110km / h, killing the driver.

At that time, Tesla was not equipped with a fully self-driving (FSD) chip developed by himself, but an Eye Q3 chip from Mobileye, an established self-driving company. Who will "carry the pot" has become the biggest difference between Tesla and Mobileye.

Amnon Shashuya, chairman and CEO of Mobileye, believes that Mr Musk has assured himself to his face that Autopilot will introduce a mechanism to ensure that drivers must hold the steering wheel by hand. In other words, in the grading standard specified by the American Institute of Automotive Engineers (SAE), this corresponds to L2-level auxiliary driving technology.

In his view, Tesla in the promotion of "autopilot" (corresponding to L3 or above) as the biggest selling point, mistakenly used L2 for L3 radical route, is the culprit that led to the tragedy.

This dispute became the fuse for Tesla and Mobileye to go their separate ways. In March 2020, the National Transportation Safety Board (NTSB) determined after an investigation that the responsibility for the accident was attributed to drivers who overrelied on L2 autopilot systems.

Since then, the legal dilemma faced by L3 has become one of the reasons why some car companies skip L3 and only do L2 or L4.

Autopilot is usually divided into six levels, ranging from L0, L1 to the highest L5. Different levels of technology correspond to different attributions of responsibility. Among them, the L0-L2 stage is considered as auxiliary driving, the driver is the dominant, while L4 and L5 are machine-led self-driving stage, the responsibility is relatively clear.

Although the L3 autopilot system, which is in the transitional stage, is considered to be a "watershed" of self-driving ability, marking the formal transfer of driving power from "human" to "autopilot system", it is this positioning that puts L3 in an awkward position. Since both drivers and machines need to participate in it, which side should be held responsible?

For many manufacturers, laws and regulations are not in place and the division of responsibilities is not clear. no matter how advanced the technology is and no matter how great the interests are, they dare not act rashly. In addition to being afraid to bear a large number of responsibility for accidents, it is also a devastating blow to one's own reputation.

In addition, L3 autopilot is not only the transformation of accident responsibility from "human" to "car", but also puts forward strict requirements for the reliability and safety of the system. To ensure the safety of L3 autopilot system to a greater extent, we need redundancy, redundant sensors, redundant actuators, redundant power supply, redundant communication and so on. In order to achieve "redundancy", the vehicle must be stacked with components in hardware, software and other aspects, and the final result is that the cost of the whole vehicle is too high.

3. L3, there is a long way to go in recent years, many countries around the world are actively promoting the commercial application of high-level autopilot.

In 2021, Japan took the lead in testing the L3, and Toyota cars equipped with L3 autopilot system were sold in Japan and hit the road. In the same year, Germany approved L3 autopilot on the road at the legal level, and Mercedes-Benz became the first European car company to comply with the regulations. In the past two years, there has also been a frequent message in South Korea that L3 autopilot will be allowed to be commercialized.

On June 19 this year, Britain allowed driverless cars to be used commercially for the first time. Before that, the United States pushed self-driving trucks on the road. On June 1, the California legislature said it planned to enact a new bill on self-driving trucks. The bill requires that when self-driving trucks with a total weight of more than 10,000 pounds are tested or transported on public roads, they must be equipped with safety guards. At present, the bill has been approved by the California state legislature, if approved by the Senate and the governor, it will formally take effect.

The rapid development and application of high-level autopilot technology around the world also requires China to liberalize the commercial operation of L3 and above. In this regard, Xu Haidong, deputy chief engineer of the China Association of Automobile Manufacturers, said that for the market-oriented application of high-level self-driving, China cannot bind its hands and feet, but must release it at the right time to speed up its marketization process.

Zhu Xisan points out that if you stay at L2 or L2 +, it will cost Tesla the volume, but it will not be enough: "for the same system, Tesla's pure vision scheme, the hardware BOM cost of HW3.0 will not exceed 10,000 yuan, but some cars with urban NGP landed this year use high-precision maps, lidar, and so on, and the BOM cost is basically between 40,000 and 70,000."

Therefore, he believes, "We should adhere to the technical route of realizing the L3, overcome the 'valley of terror' through L3 autopilot, and avoid the competition with Tesla in cost."

Relevant people pointed out that the landing of domestic L3 intelligent driving has landmark significance, on the one hand, it represents the national policy's attention and determination to the intelligent driving industry, or clears the obstacles for the follow-up development of the industry; on the other hand, accelerating the application of AI innovation in the automobile side may soon usher in the real commercial inflection point of autopilot.

Hong Yong, an associate researcher at the Electronic Commerce Institute of the Research Institute of the Ministry of Commerce, believes that the release of the L3 self-driving standard is not only conducive to the healthy development of the entire industry, but also helps to eliminate interoperability problems between different regions and manufacturers, so that self-driving vehicles can operate seamlessly around the world. At the same time, it can also reduce the cost of using autopilot technology and provide consumers with more choices.

However, in order to achieve large-scale commercial mass production, L3 autopilot still has a long way to go. According to che Baizhi Ku (Wechat: EV100_Plus), there are mainly the following two points.

First of all, the development of autopilot can not be achieved overnight. Landing L3 involves road traffic law and other certifications. first of all, instructions should be issued at the regulatory level and pilot projects should be carried out to promote them throughout the country. At least in the next two or three years, the possibility of a real full landing on the L3 is less likely.

Secondly, although some regions and countries have legislated to test the waters of L3, but from the perspective of the overall market, the market share of legislation is still small. If car companies choose to launch L3 self-driving cars, it means that the technology can only be landed in some markets, and the sales and promotion of cars will be limited.

However, with the introduction of more groundbreaking and leading policies, China's self-driving industry may take the lead in this round of technology competition.

[full text reference]

[1] "L3: a detour in the smart driving track? Zheng drove and talked about AI.

[2] "Autopilot ushered in a policy that is good for new energy car companies who can land L3 quickly and on a large scale", CCTV

[3] "Zhu Xisan: strategic determination is needed to realize L3 self-driving", Automobile Home

This article comes from the official account of Wechat: che Bai think Tank (ID:EV100_Plus). Author: Zhou Shuangjiang.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.