Speed Controller design for Walking Piezo Actuator Based on Physical Guided Neural Network

Walking piezo actuators (WPAs) employ multiple groups of piezoelectric feet, each driven by a specific waveform, to achieve walking motion with nanometer precision and millimeter-scale motion range. Due to rate-dependent displacement errors inherent in the walking motion, the WPA speed exhibits a nonlinear relationship with the actuation frequency, which impacts its speed precision. To address this challenge, this paper presents a WPA speed model based on a physical guided neural network (PGNN), which effectively captures this nonlinear relationship with high precision and low computational complexity. Building upon this model, an open-loop speed controller is designed. The proposed controller adjusts the step length during walking motion, thereby establishing a linear relationship between actuation frequency and WPA speed. Experimental results demonstrate that the proposed controller reduces the speed error to below 1.2% across a speed range of 0-470 μm/s, representing a 25% improvement compared to the no-compensation scenario.