Fixed-Time Passivity-Based Model-Free Adaptive Control of Robotic Manipulator Using Zeroing Neural Network

Abstract

This paper addresses the challenges of controlling robotic manipulators characterized by high nonlinearity and uncertainty. Traditional model-based approaches often require precise knowledge of robot dynamics, which can lead to instability when assumptions are incorrect. To overcome this limitation, we propose a fixed-time passivity-based model-free adaptive control scheme utilizing Zeroing Neural Network (ZNN). Unlike conventional methods, our approach estimates the robot dynamics solely from input data, eliminating the need for prior knowledge of internal parameters. By leveraging the passivity property of the Euler-Lagrange equation, the ZNN enhances convergence speed and robustness against disturbances. Lyapunov function was employed to conduct a comprehensive analysis of the stability of the proposed model-free adaptive controller. We then compare the performance of the proposed ZNN-based adaptive control scheme with traditional Radial Basis Function Neural Network (RBFNN) and demonstrates superior tracking precision. The proposed method is rigorously validated through extensive simulations, highlighting its effectiveness in achieving high-precision control in dynamic environments.