Neural network based active disturbance rejection control for multi-joint robotic arm

Abstract

In this paper, an active disturbance rejection (ADRC) control method based on radial basis function (RBF) neural network is proposed and applied to the attitude control of multi-joint robotic arm. This approach is employed for enhancing the attitude control of a multi-joint robotic arm. The proposed method merges the nonlinear active disturbance rejection controller, based on an extended state observer (ESO), with the RBF neural network. The objective is to adaptively mitigate external interferences and internal friction. By employing RBF neural networks as the core of the control system, the active disturbance rejection controller's parameters are dynamically adjusted through the learning of the nonlinear mapping relationship within the system. This, in turn, bolsters the system's resistance to interference. Experimental validation is conducted, yielding results that underscore the effectiveness of the ADRC control approach grounded in the RBF neural network. It successfully attenuates the impact of external disturbances and internal friction, leading to heightened precision and stability in the robotic arm's attitude control..