Autonomous driving is on the way.
At the beginning of this month, Chongqing and Wuhan took the lead in releasing a pilot policy for the commercialization of autonomous driving, which in the eyes of the outside world indicates that my country's autonomous driving will usher in a high-speed growth inflection point.
Since 2017, the Ministry of Industry and Information Technology, the Ministry of Public Security, the Ministry of Transport, the National Standards Commission and other ministries and commissions have successively issued the "Management Specifications for Road Tests of Intelligent Connected Vehicles (Trial)" and "Management Specifications for Road Tests and Demonstration Applications of Intelligent Connected Vehicles (Trial)" and other documents.
In recent years, the central and local governments have successively launched a series of support policies to promote the advancement and commercialization of driverless technology.
As the core support for future travel, autonomous driving technology is like a golden key to open the door to emerging industries.
According to incomplete statistics, Beijing, Shanghai, Guangzhou, Shenzhen, Changsha and other regions have issued corresponding management measures or detailed implementation rules, and explorations have been carried out in multiple stages from road tests, manned tests, demonstration operations to unmanned operations.
At the same time, recent traffic accidents related to "autonomous driving" have been frequently searched.
In early August, a Xiaopeng P7 had a traffic accident when the LCC (Lane Centering Assist) function was turned on for assisted driving.
This is not an exception. In August last year, Lin Wenqin, the founder of the chain restaurant brand, died in a car accident while driving a Weilai ES8 and using the NOP pilot function, and the industry's disputes over the safety of the automatic driving system have never stopped.
At the beginning of August, Shenzhen provided the legal basis for the determination of liability for autonomous driving traffic accidents for the first time. Among them, if the car is completely unmanned, the owner will be responsible. If it is confirmed that the accident is caused by a vehicle defect, the owner can report to the manufacturer of the autonomous vehicle. seek compensation.
The era of autonomous driving may not be far away. Is this technology mature and what difficulties it faces? What causes frequent accidents, and what kind of regulatory norms the industry needs urgently.
In this regard, the Beijing News Shell Finance reporter interviewed Jie Fei, executive deputy director of the Beijing High-level Autonomous Driving Demonstration Zone Office, Zhu Xichan, professor of Tongji University and director of the Institute of Automotive Safety Technology, and the Urban Transportation Institute of the National Development and Reform Commission's Comprehensive Transportation Research Institute. Cheng Shidong, director of the center.
develop
Commercialization of high-level autonomous driving industry has a long way to go
Beijing News Shell Finance: During the testing of autonomous vehicles, the construction of cars, roads, clouds, networks, and maps is often mentioned. What role do they play?
Jiefei: The technical route of vehicle-road-cloud integration practiced in Beijing's high-level autonomous driving demonstration area is based on "smart cars", "smart roads", "real-time cloud", "reliable network", "accurate Figure" five systems to carry out infrastructure construction.
Through vehicle-road-cloud integration, we can realize the automatic driving vehicle-road coordination technology we usually mention, and solve some complex and dangerous scenarios that cannot be solved and processed by bicycle intelligence.
Build multi-source roadside perception infrastructure including cameras, millimeter-wave radar, lidar, etc. on the roadside to provide autonomous vehicles with a "God's perspective" that cannot be achieved by their own sensors, and improve the vehicle's fusion perception capability; in the cloud, Through the combination of edge cloud-regional cloud-central cloud, the accurate perception of all elements of digital intersections is completed, so as to coordinate global management and information services to meet the needs of different levels of autonomous vehicles; at the same time, by building a decimeter-level high-precision dynamic map platform Lane-level precise positioning relies on low-latency, high-reliability, and extremely stable network support to realize real-time data exchange.
While maximizing the safety of autonomous driving, it reduces the high cost of single-vehicle intelligence and promotes the large-scale implementation of high-level autonomous vehicles.
Beijing News Shell Finance: At present, what problems are still being broken through in the demonstration area?
Jiefei: First of all, at the technical level, there are still core key technologies that urgently need breakthroughs in the upstream and downstream of the industrial chain.
Based on its own automobile industry foundation, the demonstration zone focuses on supporting the industrial chain of core projects, focusing on "stuck neck" technologies such as chips and operating systems, and formulating special policies.
At the same time, as a new industry with integrated development, it is necessary to establish new construction standards and investment models.
The demonstration area adheres to the "Chinese solution" technical route of vehicle-road-cloud integration.
If the standards are not unified, the data will not be able to be interconnected, which will reduce the supporting role of the vehicle-road-cloud integration platform and the investment willingness of enterprises, increase related costs, and be unfavorable for large-scale promotion.
At present, the demonstration zone has released the first high-level autonomous driving demonstration zone standard system, and the development of relevant standards is being further advanced.
In addition, at this stage, the commercialization and scale of the high-level autonomous driving industry has a long way to go.
Therefore, the policy-pioneering zone creates a first-and-tested supervision sandbox, actively promotes commercialization exploration and the implementation of commercial scenarios, and continuously optimizes operation supervision to create a good business environment for enterprises in the demonstration zone to explore commercialization.
In the next step, the demonstration zone will strengthen the strategic direction of the development of intelligent connected vehicles, and help Beijing accelerate the construction of a benchmark city for the global digital economy by promoting the integrated development of technologies, policies and industries that have a major leading role.
By creating a benchmark for the implementation of intelligent networked vehicles in China, forming a Beijing practice model, gradually integrating the networked cloud-controlled automatic driving system into the smart city construction system, and promoting the application of new digital infrastructure and intelligent networked technology to more scenarios, enabling the general public Able to share the dividends of smart cities.
Beijing News Shell Finance: At present, the test of intelligent networked vehicles has entered the stage of "unmanned driving". From the perspective of the overall unmanned test, what are the main stages?
Jiefei: The road test of the intelligent connected passenger car is divided into four stages: the main driver, the co-driver, the rear seat and the remote driving.
In April 2021, relying on the demonstration zone to establish the Beijing Intelligent Connected Vehicle Policy Pilot Zone in the Economic and Technological Development Zone, through a series of policy pilots, relevant scenario tests have been opened successively, and enterprises have been promoted to carry out commercial exploration.
In April last year, the Municipal Autonomous Driving Office issued the first batch of road test licenses for passenger vehicles, opening the scenario test of intelligent network-connected passenger vehicles in the policy advance area.
In October, the unmanned road test was opened, and it was divided into three stages: the co-driver manned road test, the rear manned road test and the remote distance outside the car.
In November, the country's first commercial pilot of autonomous driving travel services was opened.
In April of this year, we opened the pilot program for the pilot of unmanned demonstration operation of the main driver in China for the first time.
In July, the first unmanned commercialization pilot in China was promoted to open in the policy pioneer area.
In the future, on the premise of ensuring safety, the industry will step into the "unmanned" stage of the whole vehicle in an orderly manner, and finally realize the transformation of the public travel service experience.
There is a need to push for autonomous bus testing
Beijing News Shell Finance: In August, the Ministry of Transport issued the "Guidelines for Self-Driving Vehicles Transportation Safety Services (Trial)" (draft for comments), what do you think of the introduction of this policy?
Cheng Shidong: Encourage the use of autonomous vehicles to engage in urban bus (electric) passenger transportation in scenarios such as closed bus rapid transit systems, mainly because different models have different requirements for driverless technology.
At present, cars, trucks, and low-speed unmanned delivery vehicles are all being tested, and it is also necessary to promote the self-driving test of buses.
Generally speaking, autonomous driving technology is universal, but there will be certain differences between different models. It is worthy of affirmation to promote the test according to the differences of different models.
Beijing News Shell Finance: The commercialization of autonomous driving has been accelerated in various places. What stage is the development of autonomous driving in my country at present?
Cheng Shidong: The technological development of each enterprise is different. As an industry insider, we can only judge the degree of technological development through some specific behaviors in the company's testing process, from the test site to the open road, from the presence of security personnel to the absence of security personnel. A manifestation of progress.
At present, both international and domestic, autonomous driving technology has taken a big step forward, which represents the company's confidence in its own technology.
ACCIDENT
Enterprises must define the conditions for the use of assisted driving
Beijing News Shell Finance: Recently, Xiaopeng Motors and Weilai Automobile have seen frequent accidents in assisted driving. What changes have been made in the publicity of enterprises, and how should the safety responsibility for such accidents be defined?
Cheng Shidong: In the first two years, companies may have promoted driverless or autonomous driving more, and in the past two years, they have emphasized assisted driving.
This is mainly because companies believe that it is unlikely that fully autonomous driving will be realized in the short term, so they turn to advanced assisted driving.
I think in the process of assisted driving, once an accident occurs, the division of safety responsibilities is very important.
The law now states that as long as the driver is driving the vehicle, even with assisted driving, all responsibility lies with the driver.
In fact, the use of assisted driving, such as the cruise control function, can distract the driver and require reaction time in the event of an emergency, which can increase safety risks.
Therefore, the current division of responsibilities is not reasonable enough.
Car manufacturers need to define the conditions for the use of assisted driving. Under these conditions, car companies should take responsibility for safety incidents.
On the contrary, if the driver uses the assisted driving scene under the specified conditions, the driver should be responsible after a traffic accident occurs.
The system becomes smarter, "L2+ is not autonomous driving after all"
Beijing News Shell Finance: In your opinion, what is the important reason for the frequent occurrence of "autonomous driving" accidents?
Zhu Xisan: The assisted driving function of some models has reached the L2+ level from the L2 level, and the automatic control capability is improving.
From the ADAS (Advanced Assisted Driving System) that existed a few years ago, the first generation of ADAS products used millimeter-wave radar and monocular cameras as environmental perception sensors, without artificial intelligence algorithms, using embedded electronic architecture, and no high calculation. The assisted driving system below the L2 level also does not use high-precision maps.
Drivers generally do not rely on this type of assisted driving system, because its actual performance is not good, so the L2-level assisted driving system is safe.
In contrast, the electronic architecture of the new generation of vehicles now uses a domain controller and a high computing power chip. The environment perception sensor of the vehicle has become a multi-purpose high-definition camera. With the support of the high computing power chip and AI algorithm, these high-definition The perception ability of the camera is greatly improved.
In addition, the new generation of models uses high-precision maps with centimeter-level accuracy and lane-level high-precision positioning in the intelligent driving domain controller.
At present, there is no L3-level autonomous driving standard in China, so the publicity of car companies basically refers to the L2+ level system as high-precision navigation intelligent driving (NOP, NGP, etc.).
The new generation of vehicles has strong computing power, and artificial intelligence algorithms are also used for environmental perception and camera processing. Each model is basically equipped with 8-10 cameras, which is indeed very capable.
Under normal circumstances, in the areas covered by high-precision maps, the intelligent driving systems of these vehicles basically have no problems, and there is almost no need for driver intervention in normal traffic conditions. Therefore, although the car companies are not misleading, car owners will accumulate accusations during use. It trusts and mistakenly thinks it is using self-driving features.
But the L2+ intelligent driving system is not autonomous driving after all.
Beijing News Shell Finance: There is a view that not correctly distinguishing between automatic driving and assisted driving is an important cause of accidents. What do you think of this view?
Zhu Xichan: We want to tell consumers that Chinese auto companies cannot sell self-driving cars yet, so it is impossible for users to buy cars with self-driving functions.
No matter how high computing power chips and how expensive lidars are on the vehicle, the current car intelligent driving systems that can be sold belong to assisted driving.
With the assisted driving system, the driver cannot be distracted, although it is in automatic control, the driver still needs to concentrate on driving.
technology
Perceiving static objects is still difficult for intelligent driving
Beijing News Shell Finance: Why has the automatic driving function not yet reached the L3 level?
Zhu Xichan: At present, the assisted driving of some models is supported by high-precision maps and high-power artificial intelligence algorithms, and the driving ability has indeed improved, but it is not autonomous driving after all.
At present, the assisted driving system vehicle has a collision accident, and it is basically easy to hit a static object.
The characteristics of machine driving and human driving are different.
For people, it is easy to identify static objects, but for smart cars, millimeter-wave radar has the ability to measure speed, and cameras use the optical flow algorithm to be sensitive to dynamic objects, but their ability to identify static objects is poor.
Judging from the accidents of smart cars, most of them hit vehicles or objects parked on the side of the road, which is why this type of smart car has not yet reached the L3 level.
Beijing News Shell Finance: How to solve the problem of identifying static objects in autonomous driving?
Zhu Xichan: The current intelligent driving system has poor ability to recognize static objects, and from a technical point of view, it may be a lidar.
This year, several newly launched models in China are equipped with lidar, such as the new versions of Xiaopeng P5 and P9, NIO ET7, and Great Wall Motor's Mocha.
As more models equipped with lidar come on the market, let's see if the problem of poor static object recognition ability can be solved.
At present, the vision and lidar fusion solution may be more suitable for the development direction of autonomous driving.
The perception of static objects may require lidar, and it may not be possible to rely solely on vision.
fall to the ground
L3 certification standards are being developed in China
Beijing News Shell Finance: How far is it from true autonomous driving?
Zhuxi Production: Germany is equivalent to having realized autonomous driving, and the Mercedes-Benz EQS equipped with L3-level autonomous driving function can be mass-produced and sold.
The domestic progress will also be very fast, and the domestic L3-level autonomous driving certification standard is also being formulated.
It should be noted that even if the L3 level automatic driving function is achieved, it does not mean that the hands can be taken off the steering wheel at any time.
The L3 level automatic driving function that has been released in Germany is the main lane cruising (ALKS) function (TJP) in the highway congestion scenario where the speed is limited to 0-60 km/h. Can't get off the steering wheel.
It is expected that next year, the new L3-level autonomous driving standard will be introduced, and the vehicle speed can reach 0-130 kilometers per hour (HWP) on the highway.
For consumers, consider whether it is cost-effective to buy a smart car with L3 capabilities.
If the vehicle function reaches the L3 level of autonomous driving, it is likely to be about 100,000 yuan more expensive than the current vehicle with the same type of L2 function, because the cost of the domain controller and lidar it is equipped with is still very high.
Beijing News Shell Finance: What is the difference between the main technical routes of autonomous driving at home and abroad?
Zhu Xichan: Germany has promulgated L3-level autonomous driving regulations to carry out L3-level autonomous driving certification. From this perspective, Germany is ahead.
From the perspective of domestic driving road conditions, many cars will be jammed when autonomous driving is dealing with congestion, which greatly increases the difficulty of autonomous driving, and may exit frequently under congested road conditions. At the same time, the current camera recognizes traffic lights, lane lines, pedestrians, traffic The accuracy of signs still needs to be improved, especially domestic traffic lights, not to mention cameras, I think it is a little difficult for us to understand.
At present, in terms of autonomous driving technology routes, ADAS driving assistance systems are the same globally, and there is no difference between L2+ level and L3 level technology routes at home and abroad.
However, in the direction of driverless driving, there are differences at home and abroad. The United States focuses on bicycle intelligence, while in China, it is doing network intelligence, road-side perception, and 5G communication to integrate the vehicle with the road and the cloud platform.
However, neither single-vehicle-intelligent unmanned driving nor network-connected-intelligent unmanned driving have passed the safety certification.
"Commercialization can only be achieved if the technology is mature"
Beijing News Shell Finance: When will my country achieve autonomous driving?
Cheng Shidong: It can only be said that it is getting closer, because from an international point of view, some companies have canceled security officers and tested this for a long time.
Domestic companies have also moved the safety officer from the main driver to the co-pilot, and even canceled the safety officer, which shows the trust in the safety of their own intelligent driving.
However, it cannot be ruled out that blind spots or key links may not be overcome. This process will take a long time, and it is currently impossible to predict when it will be realized.
Beijing News Shell Finance: Driverless taxis have started pilot projects, what are the prospects for commercialization?
Cheng Shidong: When the technology is immature, commercialization is not possible. Only when the technology is mature can it be commercialized.
Commercialization depends on cost-effectiveness and cost. Only if it can replace traditional taxis with drivers can we talk about market competitiveness.
At present, commercialization is more about testing and cognitive education for users, so that everyone is familiar with autonomous driving, and when the technology is mature, they have the confidence to ride. This is a kind of training in user psychology and understanding.
specification
Certification and OTA are the focus of industry supervision
Beijing News Shell Finance: How to further supervise the arrival of the era of autonomous driving?
Zhuxi Production: First, clarify the certification standards.
The self-driving system needs to be certified. It is necessary to formulate my country's self-driving car certification standards and conduct safety certification before allowing auto companies to sell cars with self-driving functions.
Second, OTA (over-the-air technology) supervision is required.
The intelligent driving software of the vehicle needs to be updated iteratively, that is, OTA update.
At present, the State Administration for Market Regulation has issued OTA regulatory requirements. If an enterprise wants to perform an OTA upgrade, it first needs to go to the State Administration for Market Regulation for filing.
Beijing News Shell Finance: What should car companies pay attention to in the promotion of intelligent driving and other aspects?
Zhu Xichan: At present, Chinese auto companies are still very disciplined on their official websites and user manuals, and they have not carried out excessive publicity on intelligent driving functions. To a certain extent, they have fulfilled their initial social responsibilities.
But some auto executives post misleading information about "autonomous driving" on social media, and car salespeople show or show off "autonomous driving" features to customers.
As a socially responsible automobile company, it should manage and restrain its employees, and should not spread such misleading information to expand the influence of its products.
There are also videos of smart car users showing off "autonomous driving" on the self-media, and these videos often have a high click-through rate, which can increase the click-through rate, and Internet celebrities will rush to produce and spread such misleading information in large quantities.
Entertainment platforms are full of these misleading information. Should auto companies issue some statements to prevent users from being misled and causing serious traffic accidents?
If it can really achieve this level, I think this enterprise is really putting social responsibility in the first place.
"To build a car, you must ensure safety."
Beijing News reporter Wang Linlin Chen Weicheng Wu Tingting Zhang Bing