面向自动驾驶产品准入的仿真可信度评估体系OA

To satisfy the practical requirements of the autonomous driving product type approval,and promote the standardization of simulation credibility assessment,this paper proposes a simulation credibility assessment system based on the"Management-Analysis-Verification-Validation"paradigm,aligned with relevant standards such as UN R171.First,by integrating regulatory requirements,a standardized workflow for autonomous driving system simulation credibility assessment is established.Subsequently,a systematic methodological solution is developed:at the management level,documentation essentials are determined;at the analysis level,a simulation criticality analysis method is constructed;at the verification level,based on numerical analysis,software quality assurance,and error analysis,methods for code verification,calculation verification,and sensitivity analysis are proposed;and at the validation level,quantitative metrics for credibility assessment are established.Through a practical project case,an implementable credibility assessment system is developed for manufacturers,and its applicability is verified.The case study reveals that:① due to differences in dimensions and physical interpretations among various metrics,each metric is suitable for credibility characterization under different error scales;therefore,the selection of metrics should be diversified to reduce evaluation bias;② uncertainty in key sensitive input parameters must be controlled within expected bounds to prevent error propagation and amplification;③ the numerical smoothing of models may cause significant local errors under certain driving cycles,thus the application scope of the simulation model should be clearly defined.The research conclusion can provide a reference for the standardization and industrial application of autonomous driving simulation credibility assessment.