Adaptive Sliding Mode Control for Quadrotor Position Tracking Using a Novel Neural Feedback‐Error‐Learning Approach and a Flexible Sigmoid Activation Function

ABSTRACT This paper presents a novel neural adaptive sliding mode control (ASMC) for robust quadrotor trajectory tracking under severe nonlinearities, measurement noise, and external disturbances. The proposed strategy integrates a neural feedback‐error‐learning (FEL) framework that adaptively tunes both the sliding surface and reaching law gains online via a dual‐cost optimization scheme, thereby simultaneously reducing tracking errors and control effort. Two key contributions form the cornerstone of this work. First, the neural FEL framework enables the online optimization of control gains by employing distinct cost functions that separately address tracking performance and control effort. Second, the integration of a flexible sigmoid activation function in the ANN's output layer guarantees strictly positive control gains, which enhances system stability. Simulation studies demonstrate that, compared with a similar ANN‐based ASMC approach, the proposed controller achieves approximately a 26% reduction in maximum overshoot, nearly a 49% decrease in settling time, and roughly a 64% reduction in steady‐state error. In addition, significant improvements in tracking performance are observed, as evidenced by reduced root mean square error (RMSE) and mean absolute error (MAE) while maintaining comparable control effort. These quantitative enhancements underscore the potential of combining adaptive neural tuning with sliding mode control for high‐performance and reliable quadrotor operation in complex, uncertain environments.