摘要
汽车自动驾驶的场景建模及仿真分析是确保自动驾驶系统安全性的关键,对于提高整个系统的安全性至关重要,已经引起工业界和学术界的广泛关注。场景建模是场景仿真、测试、验证分析的基础,能够有效帮助发现场景中存在的危险,是自动驾驶系统设计开发面临的挑战性问题之一。然而,现有的场景模型生成方法,主要采用数据驱动式,结合路采数据和数字孪生技术构建具体场景,无法有效应对ADS场景复杂、类型多样、测试效率不高的问题。同时,为了提高场景的覆盖率,迫切需要从逻辑场景的层次设计一种可视化的场景建模语言,可用于构建逻辑场景,并具有丰富的语义模型,能够支持逻辑场景的测试及验证分析。针对以上问题,我们提出一种模型驱动式场景建模方法用于构建逻辑场景模型,并结合ADS领域知识及ISO 26262功能安全标准,提出了一种逻辑场景建模语言SML4ADS,给出了其相应的语法及语义模型。此外,设计并实现了一种可扩展的、灵活的场景模型仿真方法,能够支持多种场景仿真工具。结合实际案例分析,展现了本文所提出的模型驱动式场景建模及仿真方法的有效性,为后续场景模型的测试、验证与分析奠定基础。
Scenario modeling and simulation analysis for autonomous driving cars is a crucial aspect to ensuring system safety.It has garnered widespread attention in both industry and academia,as it is vital to improving the safety of the entire system.Scenario modeling forms the foundation for scenario simulation,testing,and validation analysis,and can effectively help identify hazards present in the scenario.This is one of the challenging problems faced in the design and development of autonomous driving systems.However,the existing methods for generating scenario models mainly rely on data-driven approaches,utilizing road acquisition data and digital twin technology to construct specific scenarios.These methods are inadequate when dealing with complex and diverse ADS scenarios,and they have low testing efficiency.Furthermore,to improve scenario coverage,there is an urgent need to design a visual scenario modeling language at the logic scenario level that has a rich semantic model and can support testing and validation analysis of logic scenarios.To address these issues,we propose a model-driven scenario modeling method for constructing logic scenario models,combining ADS domain knowledge and the ISO 26262 Road vehicles-Functional safety,and introduce the SML4ADS logic scenario modeling language with a corresponding syntax and semantic model.In addition,we design and implement an extensible and flexible scenario modeling and simulation method that can support multiple scenario simulation tools.Combined with actual case analysis,we demonstrate the effectiveness of the model-driven scenario modeling and simulation method proposed in this paper,laying the foundation for subsequent scenario model testing,validation,and analysis.
作者
周文韬
郑成行
万宜萱
杜德慧
Zhou Wentao;Zheng Chenghang;Wan Yixuan;Du Dehui(Software Engineering Institute,East China Normal University,Shanghai 200062;Shanghai Key Laboratory of Trustworthy Computing(East China Normal University),Shanghai 200062;International Joint Laboratory of Trustworthy Software of Ministry of Education(East China Normal University),Shanghai 200062)
出处
《中国汽车》
2023年第8期50-57,共8页
China Auto
基金
国家自然科学基金面上项目(编号:61972153)。
关键词
自动驾驶
场景建模
领域特定建模语言
场景仿真
autonomous driving
scenario modeling
domain-specific modeling language
scenario simulation