汽车功能安全：面向敏感指令攻击场景的自主驾驶车辆路径规划风险缓解控制

A risk mitigation control strategy for autonomous driving vehicles is proposed based on the state decoupling and real-time model predictive control(MPC) to address the problem of path planning failure and tracking instability resulting from sensitive command attacks. Firstly, the potential risks in the lateral, longitudinal, and heading states are decoupled to design network risk index and network block risk index. These indices quantitatively assess the risk of sensitive command attacks of an autonomous vehicle. Secondly, the network risk index is introduced into the local path planner based on MPC, which real-time modifies the penalty functions and plan the locally optimal reference path with consideration of the dynamic network attack risks. Additionally, a threshold-driven redundant bus switching mechanism is added to the trajectory tracking layer to mitigate the harmful effects of sensitive command attacks on vehicle control ability. Finally, three common scenarios of sensitive command attack sets are used to validate the effectiveness of the proposed strategy. The results show that the proposed strategy can mitigate approximately 31% of sudden speed changes and prevent longitudinal driving risk and lateral instability in the scenarios of acceleration and braking command attacks compared with the non-mitigated scheme. Furthermore, for steering command attacks, the proposed strategy can halt the incorrect steering process and effectively prevent collision accidents.