HOTLINE: 0086-021 52710299
Partners Partners & Customers
News Delegates’Testimonials

Development and Application of Autonomous Driving Test Scene Library

日期: 2019-08-28
浏览次数: 17

Shanghai, China, June 27, 2019 - In the 'The 2nd Annual China Autonomous Driving Test and Verification Technology Innovation Conference ', Mr. Chen Long, Director of the Technology Department of China Automotive Engineering Research Institute's Intelligent Network Testing and R&D Center, attended the session. The forum also gave a wonderful speech on the theme of “Development and Application of Intelligent Driving Test Scene Library”.


Due to a series of limitations such as high test cost, long test period and many test accidents, automotive companies are speeding up the construction of automatic driving simulation (software in the ring, hardware in the ring) testing capabilities. The autopilot virtual simulation test is based on virtual testing, combining real hardware systems with virtual environments to form a test tool chain, giving full play to the flexibility of the virtual environment, enabling rapid deployment and automated testing of test scenarios, and along with the toolchain. The number of real systems in the world continues to increase, and the number of test scenarios is reduced step by step.

Development and Application of Autonomous Driving Test Scene Library

In his speech, Minister Chen first explained that the scene is an important part of the automatic driving test. It is a comprehensive reflection of the driving behavior and driving environment of the car within a certain time and space. At the same time, a test scenario based on the definition of the product function is proposed. The classification method, that is, extracting the core factors affecting the performance and state of the product function from the perspectives of roads, environment, and traffic participants, and establishing a scene classification structure in a combined form. Taking AEB as an example, the scene classification was shown to the guests.Development and Application of Autonomous Driving Test Scene Library




Development and Application of Autonomous Driving Test Scene Library


Later, Minister Chen introduced the automated development tools from data to scene models built by China Automotive Research Institute. The camera and radar information are used to realize the acquisition of surrounding target information based on ontology, and the scenes of the traffic participants are classified into categories, and clustering and parameter statistics are performed for each category of data to obtain a scene model. In the integration of data acquisition equipment: integrated three types of data acquisition solutions, respectively, pure camera solution, millimeter wave radar and camera fusion solution and Mobileye, millimeter wave radar and camera fusion solution. Developed road test equipment and completed the monitoring platform solution. Completed the establishment of the largest and most extensive natural driving dataset in the country (500,000 km, 27 provinces and cities nationwide). Data tagging and fusion software are integrated in data management to form a structured data storage format. The scene advance algorithm, scene standard software and scene clustering algorithm of the system are developed in the scene data analysis.Development and Application of Autonomous Driving Test Scene Library

Next, Minister Chen gave a detailed explanation of the scene data analysis through practical cases: there were dangerous scenes, normal scenes and accident scenes in the scene extraction. Minister Chen showed us how they got from dangerous scenes, following scenes, lane changing scenes and moorings. Four scenes, such as car scene, through the screening of different algorithm criteria, the final synthesis of the scene segment; in the scene data annotation is divided into static element annotation and dynamic element annotation, and show us the dangerous scene extraction annotation software, parking scene extraction Marking software and lane line extraction annotation software; divided into cluster analysis and parameter analysis in the scene clustering algorithm, where Minister Chen continues to show us the cluster analysis of dangerous scenes, following scenes, lane changing scenes and parking scenes. Parameter distribution.


 Development and Application of Autonomous Driving Test Scene Library


Finally, Minister Chen introduced the automatic reconstruction method of scene based on the scene statistical model. After the speech, Minister Chen Long’s questions from the guests included: HIL could not achieve the problem of super-facts; the perception process could not achieve super-facts; how is VIL applied and applied in the scene development? How to solve the contradiction of VIL car is not the target car? And the authoritative answers to important questions such as the application of high-precision maps in testing.


Personal profile of the guest speaker: Chen Long, currently the head of the forward-looking technical department of the Intelligent Networking Automotive Test and R&D Center of China Automotive Engineering Research Institute, and also the ISO TC204 intelligent vehicle group expert and the ISO TC22 scene library standard group expert; and support the China Smart Vehicle Index Regulations. Formulated. Leading organizations have established 'i-VISTA intelligence' in collaboration with a number of domestic and foreign OEMs (Toyota, Honda, Hyundai, Dongfeng, Wuling, Jianghuai) and research institutes (Tsinghua, Jida, Tongji, Software Evaluation Center, Michigan, Leeds University). The International Joint Research Center for Network Testing and Evaluation of the Network has served as the deputy secretary-general of the center to promote the industrial application of scene-based automatic driving test evaluation technology. With the center, six car-enterprise horizontal cooperation projects were signed, and automated processing tools from data collection to simulation test cases were developed. A typical Chinese driving scene library was established, and cooperation with internationally renowned software and hardware vendors PreScan and VTD was formed. The productization of data has achieved sales revenue of more than 10 million in 2019. In 2019, based on the series of scene library products, it integrated the excellent HiL suppliers (NI, Konrad, VTD) at home and abroad to form a complete supply of automatic driving test equipment.

 Organizer: Shanghai Locka Automotive Technology Co., Ltd. (GRCC) is a technology consulting company engaged in technology development, technology transfer, technology consulting exhibition services, conference services and other business development in the field of automotive technology. The company provides industry information, business innovation development solutions, market research, business cooperation and network development platforms, personal career development, investment and financing consulting services to senior decision-makers at leading domestic and foreign companies (mainly Fortune 500 companies).



“'The 2nd Annual China Autonomous Driving Test and Verification Technology Innovation Conference  (CADT2019)”


Development and Application of Autonomous Driving Test Scene Library

Copyright ©2010 - 2016