Full Coverage Testing Method for Automated Driving System in Logical Scenario Parameters Space

Scenario-based testing is a mainstream approach for evaluating the safety of automated driving systems (ADS). However, logical scenarios are defined through parameter spaces, and performance differences among systems under test make it difficult to ensure fairness and coverage using the same concrete parameters. Accordingly, an automated driving system testing method is proposed. Guided by the established full-coverage testing framework, a quantitative evaluation method for scenario representativeness is first proposed by jointly analyzing naturalistic driving probability distributions and hazard-related characteristics. Furthermore, a hybrid algorithm integrating heat-guided hierarchical search and genetic optimization is developed to address the non-uniform full-coverage problem, enabling efficient selection of representative parameters that ensure complete coverage of the logical scenario space. The proposed method is validated through empirical studies in representative use cases, including lead vehicle braking and cut-in scenarios. Experimental results show that the proposed method achieves 100% coverage of the logical scenario parameter space with an 8% boundary fitting error, outperforming mainstream baselines including monte carlo (84.3%, 19%), combinatorial testing (86.5%, 14%) and importance sampling (72.0%, 7%). The approach achieves exhaustive coverage of the logical scenario space with limited concrete scenarios, and effectively supports the development of consistent, reproducible and efficient scenario generation frameworks for testing organizations.