Neural Network-Based Switching Output Regulation Control for High-Speed Nano-Positioning Stages

This study establishes a high-speed nano-positioning stage composed of a symmetrically driven structure with multiple parallel-bonded thin piezoelectric ceramic layers capable of performing micro- or nano-scale manipulations. Accordingly, a neural-network-based switching output regulation controller (NN-SORC) was developed to compensate for the associated hysteresis nonlinearity. To address the challenges of slow floating-point computation speeds and low compilation efficiency, a closed-loop control system with a field-programmable gate array–central processing unit (FPGA-CPU) dual-layer data-processing framework was developed. A feedback linearization method was designed to linearize the hysteresis nonlinearity of the framework, resulting in a switching-tracking error system. With the assistance of Lyapunov theory and an average dwell time technique, sufficient conditions were derived to ensure the asymptotic stability of the NN-SORC governing closed-loop system using the switching reference signals often encountered in realistic micro-/nano-scale detection and manufacturing processes. Finally, extensive comparative experiments were conducted to verify the effectiveness and superiority of the proposed NN-SORC scheme.