Finite-time neural adaptive control for high-order nonlinear switching multi-agent system based on switching linear state observer

The unmanned multi-agent system (MAS) necessitates the ability to seamlessly switch subsystems, controllers, and communication topologies to respond in real time to evolving task demands and dynamic environments. This adaptability maximizes resource utilization and enhances resilience. However, such frequent and arbitrary switching can lead to deviations from stable system states, introducing various security risks. To address the challenges of state distortion and discontinuity caused by these random switches, this paper proposes a neural adaptive control scheme based on a Switching Linear State Observer (SLSO). This scheme restores system convergence and consensus within a finite time after state transitions, ensuring secure cooperative control. The SLSO is designed to estimate higher-order states from the system’s first-order state, while neural networks are employed to account for uncertainties and unknown disturbances. Additionally, a second-order sliding mode differentiator is utilized to mitigate the “derivative explosion” issue in higher-order terms during differentiation, thereby simplifying controller design. This approach achieves stability and consensus amid frequent switches, demonstrating semi-global finite-time consensus and boundedness. Numerical simulations are presented to validate the effectiveness of the proposed controller.