UAV Formation Stability Control Strategy Based on Improved Radial Basis Function

Abstract

Stable formation control of multiple unmanned aerial vehicles (UAVs) is extremely challenging in practical missions due to factors such as strong nonlinearity, parameter fluctuations, wind disturbances, and communication between UAVs under constrained conditions. This paper proposes a distributed formation stabilization strategy, in which each UAV generates a local formation error based solely on information from its neighbors, and derives control measures from a consistent second-order error dynamic. To compensate for unknown nonlinear terms and unconsidered dynamic characteristics, an improved radial basis function (RBF) algorithm is integrated into the controller. Compared to traditional fixed-core RBF networks, the proposed method uses normalized basis functions to constrain the regressor, introduces a real-time width adjustment mechanism to improve approximation accuracy under time-varying uncertainties, and employs a robust drift learning law with leaky projection to prevent parameter divergence under stable perturbations. All closed-loop signals remain bounded, and the generated errors eventually become uniformly bounded. This paper defines a clear bound related to the approximate residual and the perturbation level. Comparative simulation results under mass disturbances, gusts, and topology changes show that the proposed method has faster convergence speed, lower steady-state generation error, and lower power consumption compared to distributed controllers without neural compensation and standard adaptive methods based on radial basis functions.