Active Security Resilient Optimal Strategy for Safe and Stable Autonomous Driving under CAN Bus Attacks

The Controller Area Network (CAN) is the core component of in-vehicle network communications, which is highly vulnerable to malicious cyberattacks due to the lack of authentication and encryption, leading to serious threat to the safety of autonomous driving. To solve this issue, this paper proposes an innovative active security resilient optimal strategy for distributed drive electric vehicles (DDEVs) to defend against various CAN bus attacks for safe and stable autonomous driving. First, a data-driven system modeling approach for DDEVs is designed to describe the nonlinear dynamics accurately under three typical CAN attacks on vehicle position and heading states. Then, a novel multi-sensor-information-based detection method with sliding window cumulative residual is developed to detect the attack timely. When the CAN attack is detected, the abnormal signals are removed and replaced with image-based state estimates before being fed into the controller. Once the attack defense is activated, a novel integrated optimal control method is proposed, jointly optimizing the steering and four-wheel torques to achieve precise path tracking and stable vehicle motion control. Extensive simulations on a MATLAB/Simulink and CarSim co-simulation platform validate the effectiveness of the proposed approach on the attack detection, the secure state estimation, and the optimal control, and hardware-in-the-loop experiments further confirm the real-time feasibility of the proposed control algorithm.