自动驾驶系统逻辑场景全覆盖测试用例生成方法

The scenario-based testing method is the mainstream means to verify the safety of the automated driving system (ADS). However, the logical scenario uses the form of parameter space to describe the scenario. It is difficult for the third-party detection organizations to use the same test case to ensure the test fairness and test coverage when the performance of the system under test is different. For this reason, this paper proposes a full coverage test cases generating method for ADS in logical scenario based on the test case representativeness. First, a systematic full coverage testing framework tailored for ADS is established. Subsequently, a quantitative evaluation method is introduced to assess the representativeness of test cases by jointly analyzing naturalistic driving probability distributions and hazardous event characteristics. Finally, an optimization calculation method for achieving full coverage of the differentiated sample combination space is developed, based on a heat-driven hierarchical greedy algorithm integrated with a genetic algorithm, enabling the efficient acquisition of representative parameter combinations that achieve full coverage of the logical scenario parameter space. The proposed approach is empirically validated using a lead-vehicle cut-in scenario. The results indicate that the proposed method achieves a logical scenario parameter space coverage rate of 100% and a boundary fitting error of 0.08, both of which significantly outperform current mainstream approaches, including the Monte Carlo method (coverage rate: 84.3%, fitting error: 0.19) and combinatorial testing (coverage rate: 86.5%, fitting error: 0.14). These findings demonstrate the methodś potential to effectively support testing organizations in developing a fair and efficient scenario generation framework.