Driving scene data is the basic data resource for the research and development and testing of intelligent connected vehicles, an important "case library" and "exercise set" for evaluating the functional safety of intelligent connected vehicles, and a key data basis for redefining the level of intelligent vehicles. The driving scene test cases are mainly reproduced through the virtual simulation environment and tool chain. Therefore, the construction of a virtual scene database is a key bridge connecting scene data and scene applications. The virtual scene database has the characteristics of infinity, scalability, batching, and automation. The Data Resource Center of China Automotive Technology Research Center Co., Ltd. has been accumulating natural driving scene resources since it carried out driving scene data collection and analysis in 2015, and has collected more than 320,000 kilometers of natural driving mileage data, covering Beijing, Tianjin, and Shanghai. The working conditions cover key areas such as highways, cities, villages, and parking lots. The environment covers various weather such as sunny, rainy, snowy, haze, etc., and the scope covers multiple types of typical scenes, corner scenes, accident scenes, etc. , Has been built into a leading Chinese characteristic driving scene database. After years of experience accumulation, the Data Resource Center has gradually formed a complete theoretical system such as data collection specifications, data processing procedures, feature extraction methods, scene database structure specifications, test case data formats, and driving scene virtual simulation test methods. In order to give full play to the application value of the existing driving scene data of the data resource center, cater to the scene needs of enterprises in the research and development and verification of intelligent connected cars, and solve the industry's pain points in localized functional safety evaluation, the data resource center is proposed to be based on driving scene data With the accumulation of technology in construction, special reports on "Research and Application of Intelligent Networked Vehicle Driving Scene Database" from multiple levels such as data collection, processing analysis, virtual simulation and evaluation system, and then provide practical technical support for the industry. The special report will be divided into 8 issues. This issue focuses on the overall thinking and latest achievements in the construction of virtual scene database. Virtual scene database construction Virtual scene database: Driving scene data is the basic data resource for the development and testing of intelligent connected vehicles, an important "case library" and "exercise set" for evaluating the functional safety of intelligent connected vehicles, and a key data basis for redefining the level of intelligent vehicles . The driving scene test cases are mainly reproduced through the virtual simulation environment and tool chain. Therefore, the construction of a virtual scene database is a key bridge connecting scene data and scene applications. The characteristics of the virtual scene database: the virtual scene database has the characteristics of infinity, scalability, batching, and automation. 1. Infinite: The virtual scene database is mainly obtained from test cases through virtual simulation modeling. Test cases are derived from functional scenes and logical scenes. Due to the continuity of scene parameter distribution and the diversity of scene elements, test cases cannot be exhausted. For example, with the continuous accumulation of the number of scenes, the virtual scene database is constantly enriched, and the virtual scene database is also unlimited. 2. Scalability: The key elements that make up the scene include static elements, dynamic elements, and driver behavior elements. The different permutations and combinations of elements and traversal values ​​expand the boundaries of the virtual scene library more abundantly, making the number of virtual scene libraries appear Proportional growth. For example, the same test scenario can expand a richer test case by changing the weather conditions, lighting conditions, the number and location of traffic participants. 3. Batch: With the help of virtual simulation tool chain to develop standard driving scene data interface, it can realize batch import and modeling of test cases, and use high-performance simulation server to realize batch simulation test, saving time and labor cost. 4. Automation: Automated testing is another feature of the virtual scene database. The evaluation rules of test cases will be written into the database. When the simulation test is over, combined with the performance of the tested object, comprehensive evaluation results and indicators are automatically given. The construction of virtual scene database: scene data format, virtual simulation tool chain and test case evaluation system are the main links of constructing virtual scene database. The test case first needs to be defined as a standard data format. The current international common data formats include OpenDrive and OpenScenario. The Data Resource Center is also actively developing hierarchical scenario data formats that meet the characteristics of Chinese driving scenarios to meet the interface requirements of multiple tool chains; The virtual simulation tool chain is the key to the construction of the virtual scene library. At present, there are more than 10 virtual simulation tools at home and abroad that can realize the static/dynamic feature modeling, environment rendering, and real-time simulation of the driving scene, which greatly enriches the application mode of the virtual scene library. And application fields, choosing the right tool chain is an important part of building a typical, universal and representative virtual scene library. The construction of the evaluation system of virtual test cases is the top priority of the scene data application. The data resource center comprehensively proposes an automatic driving virtual simulation test evaluation system from many aspects such as legal requirements, functional safety, and algorithm validity, which enhances the support of driving scenes. The practical significance of R&D and testing of intelligent networked vehicles. Classification of the virtual scene database: The data resource center divides the simulation scenes into four categories: natural driving scenes, dangerous working conditions, legal standard scenes, and parameter reorganization scenes, including different natural conditions (weather, light, etc.), and different road types (road conditions) , Lane line type, etc.), different traffic participants (vehicles, pedestrian location speed, etc.), different environment types (high-speed, residential, shopping malls, villages, etc.), including multiple types of virtual simulation test cases. Natural driving simulation scenario-fully tested scenario The natural driving simulation scene comes from the driving scene database collected by the data center and the road test scene of the enterprise. The natural driving simulation scene can well reflect the randomness, complexity and typical regional characteristics of the test. At present, the data center has collected 320,000 kilometers of natural driving scene data, and generated thousands of typical test cases through mature scene division methods. Based on the construction mechanism of daily updated test case data, the natural driving simulation scene library is continuously enriched and improved. Dangerous conditions simulation scenarios-necessary test scenarios Dangerous conditions simulation scenarios mainly cover three types of simulation scenarios: severe weather environments, complex road traffic, and typical traffic accidents. From a large number of natural driving scene databases, the data resource center extracts test cases for dangerous working conditions under different influencing factors through parameterized statistical analysis of the scenes, including weather light, geographic terrain, traffic congestion, road structure, and special obstacles Cases of risk-prone scenarios caused by other factors. In addition, the data resource center analyzes and inputs data from these dangerous conditions and builds simulation scenarios, and parameterizes the dangerous conditions for the extended generation of more extreme and edge scenarios. Standard regulations simulation scenario-basic test scenario Standard regulatory test scenarios are the basic scenarios that autonomous driving functions need to meet in the R&D and certification stages. The data center has always followed the development of autonomous driving policies and has built more than 20 standards based on ISO, NHTSA, ENCAP, CNCAP and other standards and evaluation procedures. A standard simulation test scenario supports the simulation verification of multiple automatic driving functions such as AEB, ACC, LKA, APA, and at the same time runs through the automated test process of the standard scenario. Parameter reorganization simulation scenario-supplementary test scenario The parameter reorganization simulation scenario aims to parameterize the existing simulation scenarios and complete the random generation or automatic reorganization of the simulation scenarios, thereby supplementing a large number of test scenarios with unknown operating conditions and effectively covering the blind areas of automatic driving function testing. The simulation scenarios of parameter reorganization can be legal scenarios, natural scenarios and dangerous scenarios. The legal scene can be reorganized through the parameter settings of different traffic elements; the natural scene can be reorganized using the parameter random generation algorithm; for the reorganization of the dangerous scene, the data resource center finds the edge scene through automated testing, calculates the parameter weight of the edge scene, and expands the risk of high weight The factor parameter range can realize the automatic generation of more dangerous simulation test scenarios. High Current Terminal Blocks,Panel Terminal Block,Feed Through Terminal Block,Heavy Power Terminal Block Sichuan Xinlian electronic science and technology Company , https://www.sztmlch.com
