Adaptive Neural Network with Dynamically Adjusted Activation Function Based Robust Control of Robot Manipulators

This paper proposes a hybrid control strategy for the trajectory tracking of robotic manipulators operating under time-varying uncertainties and external disturbances. The control framework combines a robust controller with a neural network-based estimator to effectively manage both structured and unstructured uncertainties. The robust component attenuates bounded disturbances and a portion of the model uncertainties, while the remaining unknown dynamics are estimated by a neural network term without requiring a known regressor matrix. To enhance estimation accuracy, a novel activation function with a dynamically adjusting center is introduced. Unlike conventional fixed-form activation functions, the proposed structure generates a unique activation function for each input instance, allowing the network to better capture complex and time-varying system behaviors. This adaptive mechanism significantly improves the network's ability to track changes in system dynamics in real time. The stability of the closed-loop system is ensured through Lyapunov-based analysis, and simulation results confirm the enhanced tracking performance and robustness of the proposed method under challenging and uncertain conditions.