Nonlinear Model Learning for Compensation and Feedforward Control of Real-World Hydraulic Actuators Using Gaussian Processes

This paper presents a robust machine learning framework for modeling and control of hydraulic actuators. We identify several important challenges concerning accurate models the dynamics real machines, including noise uncertainty in state measurements, nonlinear effects, input delays, data-efficiency. In particular, we propose dual-Gaussian process (GP) model architecture to learn surrogate actuator, showcase accuracy predictions against piecewise neural network that have been widely used literature. addition, provide techniques inverse controllers by batch GP inference an automated, seamless computationally fast manner. Finally, demonstrate performance trained real-world feedforward tracking applications.