Adaptive Sliding Mode Control for Uncertain Tilting Quadrotors Using Randomized Feedforward Neural Networks

This study explores a control strategy for a tilting quadrotor UAV (TQUAV) subject to model uncertainties and time‐varying mass. A novel sliding mode control (SMC) framework incorporating randomized feedforward neural networks (RFNNs) is developed to address the following adaptive challenges. First, for the uncertainty compensation, the adaptive laws based on RFNNs are designed to estimate unknown model uncertainties, time‐varying mass, and external disturbances. Second, for the time‐varying parameter adaptation, the neural network‐driven adaptive mechanisms are introduced for online adjustment of unknown or time‐dependent parameters within sliding manifolds. Third, for the approximation error mitigation, the conventional adaptive control techniques are employed to compensate for inherent neural network approximation errors. In addition, the closed‐loop system's stability is rigorously proven by Lyapunov stability theory, ensuring asymptotic convergence of both positional and attitude tracking errors. Comparative numerical simulations are given to validate the superior performance of the proposed adaptive control architecture over conventional methods.