CAERI(SH:601965) Jin Rong Jie·2025-11-29 11:46Core Viewpoint - The patent application titled "An Integrated Training and Testing Method for Autonomous Driving Algorithms" aims to enhance the efficiency and performance of autonomous driving algorithms through a systematic approach to training data generation and algorithm integration [1]. Group 1: Patent Details - The patent is jointly applied by Tongji University and China Automotive Engineering Research Institute Co., Ltd., with the publication number CN121029622A and an application date of October 2025 [1]. - The invention focuses on a unified architecture that enables targeted generation of training data, accelerates algorithm integration training, and provides a closed-loop iteration for testing and diagnosing algorithm defects [1]. Group 2: Methodology and Features - The method includes constructing a standardized training data link, proposing targeted generation and generalization methods for training data, which ensures efficient supply and balanced distribution of data [1]. - It establishes an algorithm integration training platform that supports modular and end-to-end technical routes, promoting collaborative training of algorithms to accelerate development [1]. - A closed-loop defect testing and diagnosis mechanism for autonomous driving algorithms is designed based on virtual perception injection into real vehicle testing, providing targeted feedback and optimization for data generation [1]. Group 3: Implications for Autonomous Driving - The integrated training and testing method aims to accelerate the iteration and upgrade of algorithms, facilitating their real-world application [1]. - This approach is expected to provide systematic support for building safe, intelligent, and efficient autonomous driving systems [1].