Abstract: To address the functional anomalies and deficiencies in autonomous driving systems caused by the performance limitations of onboard sensors such as cameras and radar, this study aims to establish a universal sensor testing and evaluation framework. This framework is intended to provide a basis for OEMs' sensor selection and performance validation. Leveraging practical experience in data-driven closed-loop engineering, a comprehensive theoretical and methodological system is constructed. This system covers the entire process from in-vehicle configuration and data acquisition to data processing and perceptual performance testing, thereby elevating sensor testing from isolated functional verification to systematic and quantifiable performance evaluation. The research resulted in a complete and actionable methodology for sensor testing and evaluation. This framework enables the systematic assessment of key sensor performance indicators, accurately revealing their performance boundaries and limitations. The proposed methodological system provides a universal testing standard for the industry. It can not only effectively guide OEMs in sensor selection decisions and enhance the reliability of autonomous driving systems but also holds significant reference value for advancing the development of environmental perception technology across the sector.

Key words: intelligent and connected vehicles; data closed loop; data collection; perception testing

摘要： 为解决因摄像头、毫米波雷达等车载传感器感知性能局限导致的自动驾驶系统功能异 常与不足问题，研究旨在建立一套通用的传感器测试评价体系，为主机厂传感器选型与性能 验证提供依据。基于实际数据闭环工程经验，构建了一套覆盖实车配置、数据采集、数据处 理到感知性能测试的全流程理论方法体系。该方法体系将传感器测试从孤立的功能验证，提 升为系统化的、可量化的性能评价。研究形成了一套完整且可落地的传感器测试评价方法论。 该体系能够系统性地评估传感器的关键性能指标，从而准确揭示其性能边界与局限性。文章 所提出的方法体系为行业提供了一种通用的传感器测试标准，不仅能有效指导主机厂的传感 器选型决策，提升自动驾驶系统的可靠性，也对推动整个行业的环境感知技术发展具有重要 参考价值。

关键词: 智能网联汽车；数据闭环；数据采集；感知测试