Distributed Neural Network Controller Based on Feature Enhancement With Flexible Time Prescribed Performance for Multi-UAV Swarms

This article focuses on the design of a distributed neural network controller to address the unknown disturbance and error convergence problems in multiple unmanned aerial vehicle (multi-UAV) swarms. The multi-UAV swarm is organized through a communication topology that connects individual UAVs with virtual leaders. To address the problem of unknown disturbances, a feature enhancement layer is designed for neural networks utilizing the Hadamard product and feature matrix, combined with a state predictor and a distributed learning strategy. This approach improves the approximation performance of the traditional radial basis function neural network in handling unknown disturbances within multi-UAV swarms. In addition, to address the error convergence problem, the flexible time prescribed performance algorithm is integrated into the controller design, ensuring that trajectory tracking errors remain within predefined bounds over a specified time period $T$. Using Lyapunov theory, Barbalat's lemma, and Gronwall inequality, it is proven that the tracking errors of the multi-UAV swarm in the overall closed-loop system converge to an arbitrarily small region and remain uniformly ultimately bounded. Finally, simulation results further demonstrate the effectiveness of the proposed control algorithms.