"From the perspective of the industry, autonomous driving is not so hot."
Text / Ba Jiuling (WeChat public account: Wu Xiaobo channel)
At the end of April, Musk made a surprise visit to Beijing, and the outside world generally analyzed: this is to clear the policy obstacles for Tesla's self-driving system FSD V12 to enter China. FSD V12 is currently the world's most advanced autonomous driving system.
As early as April 25, Hangzhou announced that from May 1, the main urban area of Hangzhou will be open to unmanned driving, with an open area of 3474 square kilometers, accounting for more than 20% of Hangzhou's area and serving a population of more than 10 million. From May 1, the main urban area of Hangzhou will be open to unmanned driving. The open area is 3,474 square kilometers, which is more than 20% of the area of Hangzhou.
In Hangzhou, unmanned buses run in the Xianghu Lake Scenic Area
Many people wonder: In the future, on the streets of Hangzhou, will you see unmanned cars running around? Hangzhou wants to be the first big city in the world to allow unmanned vehicles on the road on a large scale?
Then, on May 6, Zhang Ying, director of the Shanghai Municipal Commission of Economy and Information Technology, publicly stated in the media that there have been 1,002 unmanned test roads in Shanghai, with a test mileage of 2,000 kilometers.
On May 7, the Shenzhen Municipal Transportation Bureau released the "Sixth Batch of Open Road Catalogue for Road Testing and Demonstration Applications of Intelligent Connected Vehicles in Shenzhen", with a total of 43 open roads and a total mileage of about 106.01 kilometers. At present, the total mileage of Shenzhen's test opening has reached 944 kilometers.
The three cities "made a move", and the competition to take the lead in autonomous driving has a strong meaning.
Specifically, each city has its own advantages. Hangzhou's advantage is the full opening of the main urban area and fast legislation, the implementation of May 1 "Hangzhou Intelligent Connected Vehicle Testing and Application Promotion Regulations" is the first city in the country to clarify the specific process of autonomous vehicles on the road with local legislation, in addition to the special economic zone, and also the first city in the country to legislate for low-speed unmanned vehicles.
Shanghai has the advantage of Tesla's proximity to the water. In addition, its comprehensive strength ranked second in the country in the "Comprehensive Evaluation Index and Best Practices for the Development of the Urban Intelligent Connected Vehicle Industry (2022)".
Shenzhen has the strongest foundation for the autonomous driving industry. According to Qichacha data, there are 921 autonomous driving companies in Shenzhen, which is far more than other cities (previously Beijing took the lead).
Although other cities have not "let go", there are many that have laid a good foundation. In the Comprehensive Evaluation Index and Best Practices for the Development of the Urban Intelligent Connected Vehicle Industry (2022), Beijing ranked first in the overall ranking; Wuhan issued 1,581 road test and operation licenses for various types of autonomous driving, and 387 intelligent networked vehicles were regularly tested and operated, the largest number of license plates, normalized test and operation vehicles, and cumulative travel service orders in the country.
However, from the perspective of the industry, autonomous driving is not so hot.
In fact, from the perspective of the two major markets of the United States and China, the autonomous driving industry has fallen into a low ebb in recent years. Last year, for example, Cruise Origin, one of the twin stars of Silicon Valley's self-driving companies and once valued at more than $30 billion, completely failed, its Robotaxi (driverless taxi) operation license was revoked, and the self-driving model has been discontinued.
The direct cause is: On the evening of October 2, 2023, after a woman was hit by a car in San Francisco, USA, a Cruise driverless taxi happened to pass by, and the Cruise driverless taxi successfully braked, but then made a wrong judgment and caused a second run over the injured woman when pulling over, and dragged her for about 6 meters.
He Xiaopeng, the founder of Xpeng, once said: "Completely unmanned driving cannot see the complete logic, and may even need to find another way." ”
China can see a sharp decline in the number of new autonomous driving companies, starting in 2022, the number of new companies has halved compared with previous years.
Does this mean a resurgence of the autonomous driving industry? The minibus contacted the relevant person in charge of the Hangzhou Economic and Information Bureau, as well as senior experts in the industry, to make an interpretation for us.
The relevant person in charge of Hangzhou Economic and Information Bureau
Intelligent networked vehicles are the trend of the times, and all vehicles in the future will be intelligent networked vehicles.
The Hangzhou government has always been at the forefront of the trend. So we took the lead in doing this in the country. In the future, within the area of the eight main urban areas, unmanned vehicle companies with licenses can choose the route to run the test, and the line is a record system, we do not do approval, just tell us about it.
Three or four years ago, in many cities, including Hangzhou, the government had too much administrative intervention in enterprises, which was actually detrimental to the development of enterprises.
The road test routes we have drawn for them, such as the ones, are not of interest to the enterprises. In those closed environments, such as some logistics parks, the technology of autonomous driving is already very mature. Now the government does not interfere with the behavior of enterprises, and enterprises can make their own choices. Moreover, we also believe that both manned and unmanned vehicles have the right to pass on the road.
The purpose of this legislation is to make the division of responsibilities very clear. For example, for the traffic police, the accident is not terrible, and the responsibility is clear.
On the other hand, we have established a safety monitoring platform, which will access the video data and driving trajectory positioning of unmanned vehicles. There are three cameras in front of the car, on the left side of the driver, and on the right side of the driver, which can also be used as the basis for us to determine responsibility.
Of course, because all unmanned enterprises take the route of lidar, one advantage of lidar is that its safety factor is relatively high, which can reach 95% in the city. However, the marking of daily road conditions may be more unclear, which is a challenge for enterprises.
At present, there have been no active accidents in Hangzhou's unmanned vehicles (with a cumulative safety test and application mileage of more than 1.2 million kilometers). From a national point of view, there have been active accidents, and there should be about a dozen or so in total.
The country's first driverless online car-hailing high-speed rail station demonstration operation line
Now many enterprises have called to consult, which is also an opportunity and challenge for Hangzhou.
We "make an exception", and companies that do unmanned driving will come to Hangzhou from other places, for example, we are negotiating with Baidu, because we can give many scenarios for it to apply. They must require a wealth of application scenarios to continuously iterate algorithms. If the government is not open, there is no way to iterate on the algorithm. Now various cities are also gradually relaxing road test restrictions, such as Shanghai.
At the end of the year, we may drive unmanned vehicles into the West Lake Scenic Area, which is equivalent to taking the lead in the country to realize unmanned vehicles entering the scenic spot. This also needs to be evaluated with the traffic management department in all aspects of traffic flow and passenger flow, including the technical level of preparation, involving all aspects of things, and comprehensive evaluation and demonstration.
Hangzhou will also try to apply more unmanned driving modes in the future. For example, the unmanned buses at night that are being practiced in Guangzhou generally take routes from high-speed railways and airports to urban areas. There are also night elevated unmanned sweepers, which are worth learning.
Jussi
Professor, School of Automotive, Tongji University
Director of the Institute of Automotive Safety Technology
I think it's too early to talk about autonomous driving in cities.
Liability for accidents caused by autonomous driving remains a serious problem. For example, if a drone hits and kills someone, who goes to court? This question seems easy, but it is prone to kicking the ball. Similar laws in Shenzhen and Shanghai Pudong have been issued for a year or two, but are unmanned cars running all over the streets?
Is China's unmanned driving a single vehicle intelligent or a vehicle-road cloud realization? It's an unknown.
Tesla is a bicycle and does not need road-side awareness and high-definition maps. Tesla's Robotaxi (driverless taxi), which may be launched on August 8, is pictureless and driverless. In China, it is a high-precision map plus smart car, plus smart road. We call it Cheluyun, that is, we need to renovate the road-end facilities, and only the road sections after the road facilities have been renovated can be operated.
There is also a technical route, which is Waymo, a subsidiary of Google in the United States, which takes the route of maps and smart cars, and Waymo also needs high-precision maps, which need to be equipped with enough map mapping vehicles in the area where the driverless car operates, and the map should be updated every hour. If it doesn't refresh the map with high precision, it won't dare to run. So it's the L4 route. Tesla is the L5 route.
Tesla's FSD has a total of 1.25 billion miles (about 2 billion kilometers) in real road conditions, but it admits that there are still accidents (minibus note: Tesla Autopilot has caused a total of 736 accidents in the United States since 2019, with a total of 17 deaths, according to a Washington Post report in June last year).
Musk believes that about 6 billion miles of self-driving testing may be needed to achieve full self-driving.
The first problem with unmanned vehicles on the road is responsibility. The second problem is not making money. For safety, a bunch of people are there to supervise. The cost of 1km is 30 yuan, and it still loses money. Most of them have to eat subsidies.
Therefore, let's do another three years of technological exploration, and in the case of immature technology, the government will "pull out the seedlings to help it grow", and some small companies will come out to "cheat and subsidize" to demonstrate, and in the end they will not be able to mass-produce, unable to generate tax revenue, and will make the city poor. Unmanned driving has not yet reached the mass promotion period.
5G autonomous vehicles under testing
The first thing to overcome is the technical difficulties of unmanned driving, and it still can't be solved now, which is the basic level. The second is to pass the legal hurdle, and if the foundation can't be passed, it will not be held accountable unless it is like the U.S. government.
The U.S. approach is this: you buy $5 million insurance for each test car, and then put it on a black box. In the event of an accident insurance claim, the negative impact (such as market value) is borne by the enterprise itself. The government does not pursue criminal liability and product liability. Liability waivers for autonomous vehicles are the key to the real implementation of autonomous driving.
However, Cruise Origin, one of the twin stars of Silicon Valley's self-driving company, also collapsed last year, and died down after a serious injury accident. In the United States, only Waymo is left running. California, with its Silicon Valley, does not dare to call itself a "driverless city".
China's AI capabilities are still insufficient, including: basic computing power (the Ministry of Industry and Information Technology will have 300E computing power construction in 2025, and Tesla's Dojo will be close to 100E computing power), vehicle-side high-computing power chips, AI cloud offline large models, vehicle-side online model training, and billions of kilometers of AI training data collection......
Now that FSD is likely to enter China, we can see whether foreign monks will chant scriptures, see if Lao Ma can run through this matter, and then see if Tesla's Robotaxi listing on August 8 can make unmanned driving that has fallen into the cold winter popular again and re-inspire investor confidence.
In the future, in terms of financial resources and technology, I think Huawei has the most potential to make autonomous driving.
In the field of autonomous driving, Huawei has built a complete chain of chips, computing power, and data. Huawei's AI chip Ascend 910 is benchmarked against NVIDIA A100, and it is expected to ship more than 400,000 pieces this year. Without computing power and data, it is impossible to have a good AI model.
The author of this article | Limpo | Editor-in-Charge | Xu Tao
Editor-in-Chief | He Mengfei | Image source | VCG