ChinaEV100: 2020 China Autonomous Driving Simulation Blue Book

Simulation testing is crucial for the safe implementation of autonomous driving. According to a study by the Rand Corporation, for autonomous driving algorithms to reach the level of human drivers, at least 17.7 billion kilometers of driving data are needed to improve the algorithms. If a fleet of 100 autonomous driving test vehicles were deployed and tested on the road 24 hours a day, at an average speed of 25 miles (40 kilometers) per hour, it would take more than 500 years to complete the target mileage. The time and cost spent on road testing are unbearable.

In addition, the lack of traffic regulations and insurance claims mechanisms for autonomous driving also restricts the large-scale road testing of autonomous vehicles. Since autonomous vehicles cannot guarantee absolute safety, the Chinese government is cautious about opening up road testing for autonomous driving. It is difficult to meet the strong testing demand for autonomous vehicles by only relying on some open roads and intelligent network test areas for road testing.

At present, simulation testing based on scenario libraries is an important way to solve the lack of autonomous driving road test data. Simulation testing mainly achieves closed-loop simulation testing of autonomous driving perception, decision-making planning, control and other algorithms by building a virtual scenario library. The scenario library is the basis of autonomous driving simulation testing. The higher the coverage of the scenario library to the real world, the more realistic the simulation test results.

Different stages of autonomous vehicle development have different requirements for scenario libraries, which require scenario libraries to implement different test functions:

According to the data in the Blue Book, currently about 90% of autonomous driving algorithm testing is completed on simulation platforms, 9% is completed on test sites, and 1% is completed through actual road testing. With the improvement of simulation technology and the popularization of its application, the industry aims to achieve 99.9% of testing volume completed on simulation platforms, 0.09% completed in closed tests, and the last 0.01% completed in actual road testing, making autonomous driving research and development more efficient and economical.

At the press conference, Zhu Xichan, director of the Institute of Automotive Safety Technology at the School of Automotive Engineering of Tongji University, said: "Without a virtual simulation platform, it is impossible to develop intelligent connected vehicles. If my country wants to realize its dream of moving from a large automobile country to a powerful automobile country, the virtual simulation platform as a core technology must be independently controllable." China is the world's largest automobile producer and consumer. Simulation software, as one of the most critical core technologies in the development of autonomous vehicles, must be independently developed to occupy a leading position in international competition. The total number of simulation software companies and units in the United States and Germany accounts for more than half of the global total. Chinese companies are still catching up in this field.

Based on this situation, entities including technology companies, car companies, autonomous driving solution providers, simulation software companies, universities and research institutions are all investing in the construction of virtual simulation platforms. Technology companies started relatively late in simulation and have less experience in exploring automotive functions, but they have big data advantages and strong software development capabilities. Currently, autonomous driving simulation technology companies mainly include Tencent, Baidu, Huawei, Alibaba, etc.

Take Tencent as an example. Based on its technological accumulation in maps, games, cloud computing, artificial intelligence and other fields, Tencent's autonomous driving virtual simulation platform TAD Sim can realize automated scene generation and high-concurrency testing in the cloud, improving the efficiency of autonomous driving test verification. It has now launched virtual-reality autonomous driving test cooperation with intelligent connected test sites in Changsha, Xiangyang, Shenzhen and other places.

Tencent Autonomous Driving Simulator (TAD Sim), a self-driving simulation platform, was released in 2018. It is a virtual-real, online-offline self-driving simulation platform that combines professional game engines, industrial-grade vehicle dynamics models, and virtual-real traffic flow technologies. In order to achieve a high degree of scene restoration, Tencent has made use of its experience and technology in the game field, and applied game technologies including scene restoration, atmospheric system, sensor simulation, physical engine, Agent AI, cloud game technology, and MMO synchronization.

In June 2020, TAD Sim 2.0 was released. The architecture was optimized and upgraded on the original basis. The data transmission and acceleration capabilities were greatly improved, and the resource usage was reduced by 30%. Currently, there are more than 1,000 scene types in the scene library, which can be generalized to generate more than 10,000 times richer scenes, and has a test capacity of more than 10 million kilometers per day.

At present, TAD Sim is cooperating with national intelligent connected vehicle test areas, national ministries and commissions, and domestic leading automobile companies in many places, including cooperating with the National Intelligent Connected Vehicle (Changsha) Test Area to rebuild the geographical overview of the test area with virtual simulation technology to support intelligent connected vehicle simulation testing; assisting the construction of the digital twin autonomous driving evaluation system of Xiangyang Da'an Automobile Testing Center and the Shenzhen Intelligent Connected Transportation Test Demonstration Platform, etc., to promote virtual and real simulation testing, accelerate the implementation of autonomous driving research and development, and promote the construction and operation upgrade of smart transportation.

The PDF version will be shared on 199IT Knowledge Planet, just scan the QR code below!