Sensor and actuator attacks, which propagate the injected false data through CAN bus, make real threats to the security of the control of autonomous vehicles. This paper attempts to solve the secure event-triggered path following control problem of autonomous vehicles subject to both sensor and actuator attacks. Firstly, a more general physical security model is established to characterize the impacts of sensor and actuator attacks on measurements and control actions. Then, feedback linearization method based on Lie derivative is exploited to linearize the established nonlinear attacked model. Thus, the nonlinear path following control is transformed into a linear case which will facilitate to the following networked-based analysis and design. Based on the established secure model, the event-triggered control scheme is utilized to alleviate communication burden under sampled-data control framework. Furthermore, the sliding control design is augmented to mitigate the malicious impacts caused by such false data injection attacks due to its inherent robustness to abnormal attack signal. Remarkably, the proposed secure sliding control can be plugged into the original control scheme rather than changing its existing control structure. Criteria for formal stability analysis and controller design are also provided and expressed in terms of some numerically tractable linear matrix inequalities. Finally, both simulation and experimental verification cases are shown to verify the composite control scheme.
Read full abstract