A Neural Network-Based Fixed-Time MAV/UAVs Formation Tracking Control Scheme Design and Applications

This article proposes a heterogeneous control scheme to enhance the tracking capability of multiple uncrewed aerial vehicles (UAVs) subswarms, whose tracking targets are determined in advance by a manned aerial vehicle (MAV). The proposed system employs reinforcement learning to develop a fixed-time formation tracking controller that addresses limitations of conventional nonlinear dynamic inversion while enhancing steady-state performance. Specifically, the critic neural network is designed to assess the implementation reward of the actor controller, which ensures that the formation tracking error converges within a fixed time. In the actor nonlinear dynamic inversion controller (ANDI), a torque and force are applied to each individual to stabilize their positional error dynamics and attitude error dynamics. Finally, stability analysis using Lyapunov theory rigorously proves fixed-time convergence of the closed-loop system. The proposed control scheme is demonstrated to be effective and superior by numerical simulation and experimental results.