In the east of the southern section of Tianfu Avenue in Chengdu, self-driving cars are already on the road.
Two weeks ago, Chengdu announced the designation of about 50 kilometers of public roads in Chengdu High-tech Zone and Longquanyi District as test roads for intelligent and connected vehicles in Chengdu. Two weeks earlier, 12 test vehicles had just received the number plates of the first batch of road test vehicles for intelligent and connected vehicles in Chengdu.
▲An autonomous driving test vehicle parked near the Chengdu Xinchuan Innovation and Technology Park
Open road testing is considered a key step in the development of autonomous vehicles. At this year's two sessions of the National People's Congress, Baidu Chairman and CEO Robin Li said bluntly that mainland autonomous driving has entered a critical period of landing, "it is necessary to further break through the policy bottlenecks that are incompatible with technology and industrial development, and stimulate innovation capabilities in the field of automatic driving."
The initiative of development has been handed over to the city. Therefore, it is imperative to build an "open network of self-driving cars". At present, local governments are more motivated and faster to explore autonomous driving test scenarios, and their competition has opened the curtain on intelligent networked vehicles.
01. Breakthrough policy
A few days ago, in front of the Xinchuan International Living Room in Chengdu High-tech Zone, we "encountered" a self-driving car.
Its appearance is no different from that of ordinary vehicles, except that the body is covered with blue circular columns and small black squares - these sensors are called the "eyes" of smart connected cars, which will replace people and complete the perception of the surrounding environment.
▲ Driverless bus
According to Tan Xin, project manager of Chengdu 5G Smart City Intelligent Driving Project of Chengdu High-tech Zone Electronic Information Industry Development Co., Ltd., they are the first batch of "branded" self-driving cars in Sichuan.
Before the "license plate", the vehicle was tested in a simulated scene at the temporary closed test site of Sino-German intelligent networked vehicles, and passed the review of the Chengdu Intelligent Connected Vehicle Road Test Expert Group.
The test road selected in Chengdu this time was initiated by the district (city) county government to test the road opening application, and then reviewed by the Chengdu Intelligent NetworkEd Vehicle Road Test Expert Committee, and approved by the Municipal Public Security Traffic Management Bureau, the Municipal Economic and Information Bureau, and the Municipal Transportation Bureau.
Key points - in vehicle preparation and road selection, municipal government departments mainly undertake the "review" function, while district (city) and county governments have greater autonomy.
Tan Xin mentioned that the "licensed" vehicles include 4 pure electric city buses and 8 pure electric cars, which need to be as close as possible to the reality of social open roads, and the road types are rich, the traffic conditions are better, and the L4 level automatic driving test is more realistic (that is, highly automatic driving, which can achieve driving without any operation by the driver), so the road section near Xinchuan was finally selected.
Previously, Shenzhen, Beijing, Shanghai and Guangzhou have successively opened autonomous vehicle license plates and test roads in recent years. Last year, Chengdu sent a special research team to Shenzhen to investigate artificial intelligence.
As more and more cities open up autonomous driving test sections, the consensus in the industry is that the key to opening up the gap will be to grasp the boundaries of openness. In October 2021, the Beijing Intelligent and Connected Vehicle Policy Pilot Zone in the Beijing Economic and Technological Development Zone allowed unmanned road testing of main vehicles under certain conditions. According to other news, the "Regulations on the Administration of Intelligent And Connected Vehicles in the Shenzhen Special Economic Zone (Draft for Comments)" has passed the "third trial" and is expected to be released within this year.
At the two sessions of the National People's Congress this year, Li Yanhong, member of the National Committee of the Chinese People's Political Consultative Conference and chairman and CEO of Baidu, also mentioned that at present, mainland autonomous driving has entered a critical period of landing, and technology has run and even led the world, but the development of high-grade autonomous vehicles in the mainland is still facing many problems such as not being able to enter the market, not being on the card, not being able to operate charges, and difficult to identify accident responsibility, and it is necessary to further break through the policy bottlenecks that are incompatible with technology and industrial development, and stimulate innovation in the field of automatic driving.
But even if it is difficult, someone has to break through. Tan Xin said, "For cities and enterprises, every policy landing will be a breakthrough in management wisdom and innovation. ”
02, long-term testing
Chengdu's first batch of "branded" self-driving cars have one thing in common: they are all based on the development of Baidu's Apollo autonomous driving platform.
Red Star News noted that previously, autonomous vehicles equipped with the Apollo platform have been driving on test roads in Shenzhen, Beijing, Guangzhou, Shanghai and other cities.
Du Yu, deputy general manager of Baidu's intelligent transportation Sichuan business department, mentioned that the company has been looking forward to the opening of test sections in Chengdu for a long time. At present, the technical problems of autonomous driving do not face a huge bottleneck, "more importantly, whether there are more scenarios to verify the vehicle's ability to automatically drive on the actual road."
For example, at present, the Apollo platform has the ability to automatically drive on different roads, including main roads, auxiliary roads, intersections, T-junctions, etc., but the system still needs to continuously carry out deep learning and iterative updates.
Du Yu explained that the road conditions are different in each city and even in each district. In urban roads, the road conditions at traffic light intersections are very complicated, including the identification of traffic lights, the identification of different vehicle states, and so on.
This also means that even if the policy of autonomous driving is different in different regions, long-term continuous testing is still a task that must be completed.
In the industry' view, road testing is equivalent to "feeding" data to the system, and continuously training optimization algorithms by collecting enough road samples to finally form a complete driverless decision-making system.
In addition, whether or not to carry people is also quite critical.
"At present, we are conducting a vehicle-road coordination non-manned test, and after each vehicle completes a test mileage of not less than 1,000 kilometers, it can be carried out with the policy iteration update and support." Tan Xin revealed that this also means that in the future, there will be no security officers in the car, and the "5G cloud cockpit" in the Xinchuan International Conference Hall can support the security officer to take over the vehicle remotely.
▲"5G Cloud Generation Cockpit" in the Xinchuan International Conference Room
03. Vehicle-road coordination
Another key point of autonomous driving is the transformation of urban roads.
Red Star News learned that the test section has been intelligently transformed, and the transformation work has been completed at 35 intersections, and traffic signal acquisition equipment has been deployed, as well as lidar, AI cameras, edge computing nodes, etc.
▲Intersection reconstruction information
"The traffic data of the intersection is collected by the roadside sensing device, and after the edge computing unit is processed, the system will feedback the road conditions of the intersection and the test area to the driverless vehicle in real time, thereby providing higher driving safety." Tan Xin said.
This model is called "vehicle-road collaboration" - relying on the development of the Internet of Vehicles, that is, in addition to the intelligence of the vehicle, sensors and other data processing equipment are installed on both sides of the road. For example, Du Yu said that an intersection, because there may be a situation where large vehicles are blocked in front of the autonomous vehicle, causes blind spots on the vehicle sensors, "If a high-level sensing device is installed on the side of the road, it can assist the autonomous vehicle to grasp the road situation beyond the visual distance, so as to make a more accurate driving judgment."
He also mentioned that in the future, the road data collected by roadside devices can be shared with all autonomous vehicles, which will reduce the dependence of bicycle intelligence on the type and number of on-board sensors, thereby reducing the cost of bicycle manufacturing and facilitating the promotion of commercialization.
"The technical path of the development of intelligent vehicles in the mainland is different, and the main use of bicycle intelligent systems in foreign countries is the mainland, while the mainland emphasizes vehicle-road coordination." Kong Ling'an, an associate professor at Xihua University's School of Automobile and Transportation, told Red Star News. This means that in addition to the intelligent system of the vehicle itself, it is also necessary to modify the road, install various signal sensors, and rely on the Internet to feedback various road information to the vehicle in real time.
However, Peng Yiqiang, dean of the School of Automobile and Transportation of Xihua University, also stressed that this involves a large-scale transformation of urban infrastructure, testing the city's determination to develop smart cars. For example, most of the intelligent transformation of some test sections in Chengdu is completed by enterprises themselves, "but the construction and application of infrastructure will definitely be oriented to the whole society, and we still need to find better costs in the future." ”
In this regard, Guo Xingrong, CTO of the Mushroom Car Federation, once said, "Enterprises in the field of automatic driving should not only be technology providers, but also deeply participate in the planning and construction of urban intelligent transportation, provide complete sets of solutions, and even participate in the operation of urban public services."
This points to city and business cooperation.
According to the planned development route, the development of intelligent roadsides is divided into smart infrastructure, smart system platforms, smart information services, smart road property management, smart traffic control, and smart auxiliary decision-making. Among them, the construction of smart infrastructure as a development focus needs to be deployed first. The natural advantages of various types of enterprises have also been involved in this. For example, 10 million technologies can provide "smart road solutions"; Apollo also has a layout in roadside perception sensor solutions, algorithms, V2X terminal hardware and software.
"Vehicle-road synergy is ultimately a challenge to synergy." Yiou Automobile once predicted that with the gradual maturity of the technology of all parties in the industrial chain and the improvement of standards and regulations, vehicle-road collaboration will usher in a "Chinese-style" development - led by the government, and jointly achieve smarter transportation and travel under the coordination of all parties.
Red Star News reporter Zou Yue
Edited by Tan Wangyu
(Download Red Star News, there are prizes for the newspaper!) ）