부산대학교 도서관

This paper studies the problem of secure and resilient distributed cooperative control for autonomous vehicles under false data injection attacks on the vehicular communication net-work. We propose a model-based scheme to detect unknown data falsification attacks using the predicted traj ectory in a distributed nonlinear model predictive control problem. First, as a local agent, each autonomous vehicle stores the information received from its neighboring vehicles to clarify anomalies. The recorded data is investigated over a temporal moving observation window which contains different prediction sequences of the distributed nonlinear model predictive control problem in the previous times. In the next step, a cumulative sum discriminator is employed to compare the predicted values from the observation window and measured values. Then, a recovery mechanism is designed to ful-fill secure control of the connected automated vehicular platoons. Finally, a comparative simulation demonstrates the effectiveness and capability of the proposed strategies.