自动驾驶认知能力测试评价研究综述  

Review on testing and evaluation of cognitive abilities for autonomous vehicles

作　　者：杨澜[1] 赵祥模[1] 王润民[1] 王振 房山 瞿广跃 YANG Lan;ZHAO Xiangmo;WANG Runmin;WANG Zhen;FANG Shan;QU Guangyue(School of Information Engineering,Chang’an University,Xi’an 710018,China)

机构地区：[1]长安大学信息工程学院,西安710018

出　　处：《汽车安全与节能学报》2025年第1期1-15,共15页Journal of Automotive Safety and Energy

基　　金：国家自然科学基金项目(52472446,52441205);陕西省留学人员科技活动择优资助项目(2023001)。

摘　　要：对交通动态情境的准确认知是自动驾驶(AV)智能性的关键体现,因此全面、合理、高效的测试评价方法对于验证其效能至关重要。为了跟踪自动驾驶认知能力测试评价的研究进展,该文首先从宏观、中观和微观3个层面深入剖析了自动驾驶测试领域存在的核心问题,深入探讨了自动驾驶与人类驾驶在认知方面的关联性;其次,从基于“金字塔”模型的自动驾驶测试体系架构入手,全面回顾了在关键测试场景生成、虚拟仿真测试、虚实融合测试、实际道路测试以及认知能力评价等领域的最新研究成果;最后,指出了自动驾驶认知能力测试评价领域面临的挑战和发展趋势。该综述研究成果将为自动驾驶技术的迭代演进和功能验证提供重要参考。Accurate understanding of dynamic traffic scenarios is a crucial manifestation of the intelligence in autonomous vehicle(AV).Therefore,it is essential to validate its effectiveness through comprehensive,rational,and efficient testing and evaluation methods.To keep abreast of the research progress in test and evaluation on the cognitive capabilities of autonomous driving,this paper first delves the core issues existing in the field of AV test from macro,meso and micro perspectives.It explores in depth the cognitive correlations between AV and human driver.Secondly,based on the“pyramid”model architecture for AV test,it comprehensively reviews the latest research findings in key test scenario generation,virtual simulation test,hybrid virtual-real test,realroad test and cognitive capability evaluation.Finally,it highlights the challenges faced in the field of test and evaluation for AV cognitive capabilities and outlines future development trends.This comprehensive review will provide an important reference for the iterative evolution and functional validation of AV technology.

关 键 词：自动驾驶(AV) 认知能力 测试场景 虚实融合 性能评价 

分 类 号：U491[交通运输工程—交通运输规划与管理]