绘制自动驾驶“零事故”的安全线路图

Core Insights - The rapid development of intelligent connected vehicles in China has raised concerns about traffic safety incidents caused by intelligent driving systems, highlighting safety "pain points" in the process of automotive intelligence and connectivity [1] - The Ministry of Industry and Information Technology and eight other departments issued the "Automotive Industry Stabilization and Growth Work Plan (2025-2026)", which conditionally approves the production access of L3-level vehicles [1] - The 2025 World Intelligent Safety Conference held in Xi'an focused on the trends and safety challenges of autonomous driving technology, with experts emphasizing the need for collaboration in the industry to continuously improve the safety roadmap towards "zero accidents" [1] Group 1: Safety and Testing - The ultimate goal of autonomous driving systems is to achieve "zero traffic accidents," which can be guided by the evolution path of achieving "near-zero emissions" in automotive emissions [2] - Key points of safety testing differ from environmental testing, as some autonomous driving systems may function well during the day but fail at night due to a lack of necessary environmental factor testing and calibration during training [2] - A team from Tsinghua University has collaborated with companies like Changan Automobile and FAW Group since 2020 to develop a scene collection that includes 100 types of scenarios and evaluation standards, forming a Safety of the Intended Functionality (SOTIF) index [2] Group 2: Technological Innovations - To train AI effectively, a laboratory has been established using virtual simulation technology to simulate extreme sudden situations, helping AI accumulate experience and promote system iteration towards the "zero traffic accident" goal [3] - The evaluation system developed is expected to be adopted as a standard for production access in China, contributing to the improvement of international safety assessment regulations for autonomous driving systems [3] Group 3: Industry Development and Collaboration - The penetration rate of L2-level assisted driving in China's passenger vehicles reached 62.1% in the first half of the year, while the approval and testing trials for L3/L4-level autonomous driving are steadily advancing [4] - There is a need to promote the development of policies and regulations that adapt to industry safety requirements, strengthen safety testing and evaluation technology research, and build a global collaborative industrial ecosystem [4] - The development of safety chips and network security for intelligent connected vehicles is crucial, with efforts in Shaanxi focusing on the autonomous research and development of vehicle-grade safety chips and the commercialization of L4-level intelligent heavy trucks [4]