Neural Network Based Singularity-Free Adaptive Prescribed Performance Control of Two-Mass Systems

This paper focuses on the trajectory tracking control problem of two-mass systems, addressing the challenges posed by unknown system dynamics and unknown control gain. To handle these challenges, we first reformulate the system model into a singularity-free form and employ neural networks to approximate the unknown nonlinear functions. To ensure that the tracking errors are bounded by predefined performance boundaries and avoid the potential singularity problem inherent in other indirect adaptive control methods, we develop a singularity-free prescribed performance controller. Additionally, to simplify the controller design procedure, we adopt a high-order command filter and abandon the commonly used backstepping control approach. We employ the Lyapunov approach to analyze the stability of the identification and control algorithms, while simulation results demonstrate the efficacy of the proposed algorithms.