Neural Observer-Based Formation for Multi-Uavs Against Deception and Desired Trajectory Attacks

This paper investigates secure formation control for multi-UAV systems in the presence of deception and desired trajectory attacks. It is crucial to note that previous papers studying the secure formation control problem for multi-UAV systems usually apply to cases where the multi-UAV systems are subject to sensor attacks and actuator attacks. In this paper, we propose a formation control scheme to address both desired trajectory and deception attacks, incorporating attack modeling, estimation, and compensation at its core. First, dynamic models are constructed for multi-UAV systems and cyber attacks, respectively. In particular, desired trajectory and deception attacks are mapped to control input channels during modeling. Subsequently, an adaptive neural network observer is introduced to reconstruct desired trajectory attack signals. The online updating of neural network weights avoids the need for manual parameter selection. Next, we propose an event-triggered secure formation controller with an attack compensation approach aimed at reducing transmission resources under an event-triggered mechanism. The uniform ultimate boundedness of the system is derived. Finally, the formation tracking protocol is substantiated and validated by providing simulation results.