基于联合仿真的自动驾驶控制执行系统预期功能安全测试方法研究

Co-Simulation-Based Safety of the Intended Functionality Testing Method for Automated Driving Control Systems

摘要:

自动驾驶控制执行系统的预期功能安全（SOTIF）测试是保障自动驾驶安全的重要活动，基于潜在触发条件构建测试场景有助于通过仿真测试完成SOTIF流程的闭环。然而，目前面向自动驾驶控制执行系统SOTIF测试的研究还相对有限。为此，首先基于现有标准和相关实践，提出了控制执行系统潜在触发条件的分类框架，以提升触发条件分析的系统性和全面性。然后，结合自动驾驶仿真软件和多体动力学模型，建立了联合仿真平台。最后，构建包含多种来源不同的触发条件的测试场景，在提出的平台上进行了仿真实验。实验结果表明，所提出的基于联合仿真的测试方法可以满足控制执行系统SOTIF仿真测试需求；与侧风用例相比，同时复合了低附着触发条件的用例导致系统误差增加了88.6 %以上。为面向控制执行系统的SOTIF测试研究提供了新的方法。

Abstract:

Testing the safety of the intended functionality （SOTIF） of automated driving control systems is essential to ensuring the overall safety of automated driving systems. Constructing test scenarios based on potential triggering conditions can help to close the loop of the SOTIF process through simulation. However， research on SOTIF testing for automated driving control systems is still relatively limited. Therefore， this paper first proposes a categorization framework for potential triggering conditions of control systems， based on existing standards and relevant practices， to improve the systematic and comprehensive analysis of triggering conditions. Next， it develops a co-simulation framework by integrating autonomous driving simulation environments and multi-body dynamics models. Finally， it conducts simulation experiments on the proposed platform with test scenarios that incorporate triggering conditions from multiple sources. The experimental results demonstrate that the proposed co-simulation method can meet the requirements for SOTIF simulation testing of automated driving control systems. Compared to those with isolated crosswind stimuli， scenarios compounded with low-adhesion triggering conditions result in an 88.6 % increase in system errors. This paper presents a novel method for conducting SOTIF testing of automated driving control systems， offering enhanced insights into the evaluation of safety-critical scenarios.