Adaptive fractional-order non-singular terminal sliding mode control for omnidirectional quadrotors based on WRBF neural network.

Abstract

This paper presents a novel robust six-degree-of-freedom trajectory tracking control strategy for tilt-rotor quadrotors operating under uncertainties and disturbances. The key contribution lies in a unified framework that synergistically co-designs a Fractional-Order Nonsingular Terminal Sliding Mode Controller (FONTSMC) with an adaptive Wavelet Radial Basis Function (WRBF) neural network, establishing a deeply integrated architecture rather than a simple combination of independent modules. This co-designed structure introduces three fundamental advances: first, the WRBF network enables precise online estimation and compensation of unstructured uncertainties while the fractional-order nonsingular terminal sliding surface ensures fast finite-time convergence without singularity; second, the Mexican Hat wavelet activation function significantly enhances local approximation accuracy, learning speed, and noise robustness compared to conventional Gaussian RBF networks; third, a parallel control structure integrated with Moore-Penrose pseudo-inverse-based allocation efficiently maps synthesized 6-DOF commands to redundant actuators. Closed-loop stability is rigorously guaranteed through Lyapunov analysis. Comprehensive simulations demonstrate that the proposed controller outperforms conventional NTSMC and RBF-FONTSMC methods in tracking accuracy, convergence speed, control effort, and response smoothness, confirming its superior capability for complex UAV operations.