Hybrid neural network adaptive fuzzy sliding mode online compensatory control for robots with global stability

• Published in: Neurocomputing (Amsterdam) Vol. 657; p. 131632
• Main Authors: Xiao, Guocheng, Liu, Bei, Li, Yufeng, Yin, Haibin
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 07.12.2025
• Subjects: Adaptive fuzzy sliding mode control; Hybrid neural network; Non-linear robotic system; Trajectory tracking; Trajectory tracking; Hybrid neural network; Adaptive fuzzy sliding mode control; Non-linear robotic system
• ISSN: 0925-2312
• DOI: 10.1016/j.neucom.2025.131632

Due to parameter variations, joint friction, and external disturbances, nonlinear robot system models may suffer from modeling inaccuracies or parameter identification difficulties. In addition, traditional control methods make it difficult to estimate the uncertainty of the system model accurately, and the uncertainty of the internal model parameters of the system will also seriously affect the tracking control accuracy of the robot. To solve the above problems, this paper proposes a hybrid neural network adaptive fuzzy sliding mode online compensation control method for robots with global stability. The technique uses an RBF neural network to estimate the uncertainty and dynamically compensates for it by introducing it into an adaptive fuzzy sliding mode controller. Compared to the traditional approximation network, the method ensures the stability of the global consistent final boundedness of the system signals. Also, it achieves the convergence of the neural network weights to the ideal values. To effectively suppress the system jitter, this paper proposes a scheme based on the combination of hyperbolic tangent function sliding mode and fuzzy control. The focus of this paper is on the simulation verification of the proposed control method. The simulation experiment results verify the effectiveness of the proposed method. The accuracy of the integral squared error of the technique in the presence of internal uncertainty and maximum friction moment disturbance is 3.24e-5 rad, which greatly improves the dynamic tracking performance of the robot.