城市自动驾驶决策系统安全分析与策略设计

Safety Analysis and Strategy Design for Decision System of Urban Autonomous Driving Vehicles

摘要:

基于系统理论过程分析（system theory process analysis， STPA）， 提出了一种面向高等级自动驾驶决策系统的安全性开发方法。该方法应用在一个城市自动驾驶决策系统的原型开发阶段， 通过安全分析得到系统的70个不安全控制行为。针对其中3个功能状态， 分析得到10个不安全控制行为原因， 提出9个安全策略。应用其中一个典型安全策略进行系统改进， 通过仿真试验对其进行了验证。试验结果表明，基于所提出方法设计的安全策略有效可行， 提出的方法能够提高自动驾驶决策系统的安全性。

Abstract:

Based on system theory process analysis （STPA）， a safety development method for high level autonomous driving decision systems was proposed. The proposed method was applied to the prototype of an urban autonomous driving decision system. 70 Unsafe Control Actions （UCAs） for the entire system were found through safety analysis. Focusing on 3 function states of the system， 10 reasons for UCAs and 9 safety strategies were put forward. One strategy got implemented and validated through model-in-the-loop simulations. Test results show， the safety strategy is effective and the proposed method can enhance the safety of autonomous driving decision systems.