Abstract

Self-driving cars are on the horizon, making it necessary to consider safety assurance and homologation of these autonomously operating vehicles. In this study, we systematically review literature that proposes new methods for these areas. The available methods were categorized into a novel taxonomy, dividing them into the strategies of combinatorial testing, robustness testing and search-based testing. We analyzed the literature in regard to modeling capabilities, targeted automation subsystem, targeted driving task level and the metrics used for criticality evaluation and coverage of the scenario space. We found that there are significant differences and shortcoming in the modeling capabilities of the existing research and that methods of each strategy usually target a specific driving task level. Additionally the criticality assessment of scenario-based validation methods was examined, revealing the need for more comprehensive metrics to assess complex scenarios. The developed taxonomy furthers the understanding in different scenario-based testing approaches for automated vehicles and serves as a guide for future research.

摘要

自动驾驶汽车即将出现，因此有必要考虑这些自动驾驶汽车的安全保证和认证。在这项研究中，我们系统地回顾了为这些领域提出新方法的文献。可用的方法被归类为一种新的分类法，将它们分为组合测试、稳健性测试和基于搜索的测试策略。我们分析了有关建模能力、目标自动化子系统、目标驾驶任务级别以及用于关键性评估和场景空间覆盖的指标的文献。我们发现现有研究的建模能力存在显着差异和不足，并且每种策略的方法通常针对特定的驾驶任务级别。此外，还检查了基于场景的验证方法的重要性评估，揭示了需要更全面的指标来评估复杂的场景。开发的分类法进一步了解了自动驾驶汽车的不同基于场景的测试方法，并作为未来研究的指南。