Nonlinearity Cancellation-Based Linear Quadratic Tracking Control for a Piezo-Actuated Fast Steering Mirror in High-Speed Scanning Tasks

This article aims to tackle the critical challenges posed by uniaxial hysteresis and biaxial coupling in high-speed scanning tasks involving piezo-actuated fast steering mirrors (piezo-FSMs). First, a deep coupling dynamical model based on feedforward neural networks (FNNs) is developed to predict the dual-axis angular output of the piezo-FSM using real measurements. The model's architecture incorporates direct connections from the input layer to the output layer, enhancing prediction accuracy and training efficiency. Second, leveraging the insights from this deep coupling model, a receding horizon control (RHC) is designed for high-precision tracking of high-speed scanning trajectories in simulation. Considering the heavy computational burden of the RHC, the third step involves designing a nonlinearity cancellation-based linear quadratic tracking (NC-LQT) control to efficiently approximate its optimal solution. Experimental validation demonstrates the efficacy and superiority of both the deep coupling model and NC-LQT control.