Review of Evaluation Metrics for the Credibility of Autonomous Driving Test Scenarios

Authors

• Jiwei Li, School of Information Engineering, Chang'an University, Xi'an 710018, China
• Runmin Wang, School of Information Engineering, Chang'an University, Xi'an 710018, China; Collaborative Innovation Center for Western Traffic Safety and Intelligent Control by Province and Ministry, Chang'an University, Xi'an 710018, China
• Yu Zhu, School of Information Engineering, Chang'an University, Xi'an 710018, China
• Jingjun Cheng, School of Information Engineering, Chang'an University, Xi'an 710018, China; Collaborative Innovation Center for Western Traffic Safety and Intelligent Control by Province and Ministry, Chang'an University, Xi'an 710018, China
• Xiangmo Zhao, School of Information Engineering, Chang'an University, Xi'an 710018, China; Collaborative Innovation Center for Western Traffic Safety and Intelligent Control by Province and Ministry, Chang'an University, Xi'an 710018, China

Abstract

Abstract: Currently, scenario-based simulation testing has become a core technical approach for verifying autonomous driving system functionality. However, due to the absence of unified evaluation standards for simulation scenarios, significant differences exist among the metrics constructed and adopted by different studies, each with its own limitations. This hinders mutual recognition of research outcomes and prevents scientific measurement of scenario quality. This study systematically reviews relevant research from the past five years, categorizing autonomous driving test scenarios into four types: parameter-descriptive, trajectory-driven, environment-simulation, and perception-data-driven. It clarifies the core characteristics, application objectives, and generation mechanisms of each scenario type. Subsequently, focusing on credibility as the core criterion, it systematically organizes existing evaluation metrics developed in various studies across five dimensions: challenge, realism, coverage, human-likeness, and acceleration. Third, through systematic integration of existing research, a unified credibility evaluation framework for multi-type scenarios is established. Finally, current challenges in this field are summarized, and future research directions are explored. This study provides a theoretical foundation for credibility verification of autonomous driving simulation scenarios and offers guidance for advancing simulation testing standardization and credibility evaluation system development.

Recommended Citation

Li, Jiwei; Wang, Runmin; Zhu, Yu; Cheng, Jingjun; and Zhao, Xiangmo (2026) "Review of Evaluation Metrics for the Credibility of Autonomous Driving Test Scenarios," Journal of System Simulation: Vol. 38: Iss. 9, Article 3.
DOI: 10.16182/j.issn1004731x.joss.25-1067
Available at: https://dc-china-simulation.researchcommons.org/journal/vol38/iss9/3

First Page

2464

Last Page

2485

CLC

TP391.9

Recommended Citation

Li Jiwei, Wang Runmin, Zhu Yu, et al. Review of Evaluation Metrics for the Credibility of Autonomous Driving Test Scenarios[J]. Journal of System Simulation, 2026, 38(9): 2464-2485.

Corresponding Author

Wang Runmin

DOI

10.16182/j.issn1004731x.joss.25-1067