Zhai Jun: Vehicle-road cooperative autonomous driving technology and implementation

From the horse-drawn carriages of long ago to the invention of steam engines, the history of human travel has continued to evolve. In today's era, with the support of new electronic and electrical architectures, the development of intelligent technology is driving industrial change, and intelligence and networking have become important directions for the development of the industry. Among them, autonomous driving has become a battleground in the automotive industry, and as one of the mainstream technical routes for intelligent driving, vehicle-road collaboration and vehicle-network integration have become new industry trends.

On June 15, at the smart driving session on the second afternoon of the 16th China Automotive Blue Book Forum, Zhai Jun, Chairman of Wanjie Technology, delivered a speech entitled "Vehicle-Road Collaborative Autonomous Driving Technology and Implementation".

This year marks the 30th anniversary of the establishment of Wanji Technology. After years of research and development and practice, Wanji has accumulated a large number of independent innovative technologies in many fields such as Internet of Vehicles, big data, cloud platform, edge computing and autonomous driving, and has developed a series of products such as laser radars at both ends of the road, V2X vehicle-road collaboration, intelligent networked roadside intelligent perception systems, intelligent networked cloud control platforms, ETC, dynamic weighing, etc.

The financial report shows that Wanji Technology achieved operating income of 909 million yuan in 2023, a year-on-year increase of 4.11%. In the first quarter of 2024, the company achieved operating income of 159 million yuan, a year-on-year increase of 50.64%. The company's dedicated short-range communication, intelligent network and laser radar business revenue has grown steadily, and the overseas region achieved revenue of 20.1371 million yuan, an increase of 237.17% over the same period last year.

Working in the same direction and resonating with national policies is the key for Wanjie Technology to seize every market opportunity.

In this speech, Zhai Jun pointed out that under the guidance of policies, it can be clearly seen that the vehicle-road-cloud integrated technology route has been highly recognized and supported by the country. Vehicles and roads, new energy intelligent connected vehicles and cities are moving in both directions, and intelligent connected new energy vehicles are about to usher in large-scale development.

At the same time, he said that autonomous driving requires vehicle-road collaborative perception, collaborative decision-making and collaborative control to achieve enhanced environmental perception, enhanced scene understanding capabilities, enhanced game decision-making capabilities, etc., to make autonomous driving absolutely safe.

He also predicted that with the deepening of the "vehicle-road-cloud integration" strategy, the intelligent connected vehicle industry and urban intelligent roadside infrastructure are about to usher in a golden period of vigorous development, and high-level autonomous driving will achieve a breakthrough in L4 within the next 3-5 years.

The following is the transcript of Zhai Jun’s speech, which has been slightly edited.

Dear experts and guests, good afternoon! I am honored to have the opportunity to discuss the development of autonomous driving technology with you today. Next, I will start from national policies and industry development needs, combined with some exploration and research of Wanji Technology in the past five years, and share with you our insights on vehicle-road collaborative autonomous driving technology.

My report is divided into four parts: industry background, vehicle-road collaborative autonomous driving technology architecture, practice and implementation, and Wanji full-stack capabilities.

Industry Background of Vehicle-Road Collaborative Autonomous Driving

In June 2024, the Ministry of Industry and Information Technology and other "four ministries" announced the first batch of "intelligent connected vehicle access and road traffic pilot" consortium list: BYD, NIO, Changan, GAC, SAIC, BAIC Blue Valley, FAW, SAIC Hongyan, Yutong Bus and other 9 car companies were shortlisted, which indicates that my country is accelerating the establishment of a sound and complete intelligent connected vehicle production access and road traffic safety management system.

On January 17, 2024, the Ministry of Industry and Information Technology and other five ministries jointly issued a notice on the pilot application of "vehicle-road-cloud integration" for intelligent connected vehicles, vigorously promoting the industrialization of intelligent connected new energy vehicles. More than 40 cities across the country have actively applied, and the final 20 approved cities will take the lead in carrying out large-scale demonstration applications of "vehicle-road-cloud integration" for intelligent connected vehicles and exploring new business models.

At present, many cities such as Beijing, Fuzhou, Ordos, and Shenyang have started bidding, and the domestic vehicle-road-cloud integration market is ushering in a climax of development. Under the guidance of policies, we can clearly see that the vehicle-road-cloud integration technology route has been highly recognized and supported by the country. Cars and roads, new energy intelligent networked vehicles and cities are moving in both directions, and intelligent networked new energy vehicles are about to usher in large-scale development.

Vehicle-road cooperative autonomous driving technology architecture

Based on the policy background, I will introduce the vehicle-road collaborative autonomous driving technology architecture to you.

As we all know, autonomous driving has a long tail problem. It has been at the level between L2+ and L3 for quite a long time. To achieve a higher level of autonomous driving, vehicle-road collaboration is needed. High-level autonomous driving requires the coordinated perception, coordinated decision-making and coordinated control of the vehicle and the road to achieve enhanced environmental perception, enhanced scene understanding ability, enhanced game decision-making ability, etc. Vehicle-road collaboration makes autonomous driving absolutely safe.

This is the overall architecture of vehicle-road cooperative autonomous driving (see the figure above). At the bottom layer, we divide it into the architecture of autonomous driving vehicles and the architecture of intelligent transportation roads. The vehicle side and the road side achieve high-precision digital time and space alignment through GPS or Beidou timing and high-precision maps. In terms of cloud platforms, there are edge clouds, regional clouds and central clouds; from the perspective of cloud control application platforms, it covers vehicle-road cooperative services, vehicle supervision and test evaluation.

Vehicle-road cooperative autonomous driving can empower autonomous vehicles online, realize vehicle-road cooperative perception, regulation and control game assistance, and global optimal path planning. It can also empower autonomous vehicles offline and provide massive scene data for algorithm training iteration and intelligent driving test simulation.

The software architecture of vehicle-road collaborative autonomous driving is divided into perception layer, communication layer, software layer and application layer, and time and space alignment is achieved at the software layer; the enhancement of perception, decision-making and planning enables high-confidence road-side perception to participate in vehicle-side decision-making and planning.

The vehicle-road collaborative autonomous driving system integrates the vehicle brain, cloud brain and road-side edge computing MEC to form a vehicle-road-cloud collaborative fusion algorithm; it realizes collaborative perception, collaborative decision-making planning, collaborative control and collaborative positioning.

The roadside smart base station collects traffic scene trajectory data from a bird's eye view 24/7, covering urban intersections, roundabouts, highways, tunnels, toll booths, service areas, etc. The massive trajectory data is mined, and these high-quality scene data can be used for autonomous driving algorithm model training iteration and test evaluation. In terms of test evaluation, it can be used for ADAS assisted driving testing, autonomous driving testing, V2X testing, etc.

The vehicle-road collaborative autonomous driving test system can serve the test research of vehicle-road collaboration and autonomous driving; the system supports multi-modal traffic participants such as autonomous driving vehicles, domain controllers, driving simulators, and simulated vehicles; the test platform integrates twin monitoring, scenario library management, simulation management, test management, V2X server and other functions.

In terms of roadside perception, the overall architecture of the roadside perception system provides autonomous driving with a bird's-eye view of beyond-visual-range global information and abnormal information through the deployment of various roadside sensors and access to other roadside equipment, and integration through the MEC algorithm.

The current laser video fusion holographic perception system, through dedicated roadside radar design and cost-effective edge computing units, can achieve a target positioning accuracy of up to 30 centimeters and an overall perception delay of less than 100 milliseconds. According to the technical requirements and test method standards for vehicle-road collaborative roadside perception systems, it can reach the highest SL3 level and meet the requirements for serving autonomous driving.