Neural Network-Augmented Iterative Learning Control for Friction Compensation of Motion Control Systems with Varying Disturbances

TL;DR

This work tackles robust friction compensation in repetitive motion tasks for a Lorentz force actuator under time-varying disturbances. It fuses Iterative Learning Control with a simple neural network and a Kalman-filtered estimator to adapt the ILC effort to changing friction and reference commands. The nonlinear component of the ILC is learned by a lightweight neural network, precomputing converged ILC effort and reducing iterations, while the Kalman filter improves robustness and signal quality. Experimental and simulation results show improved tracking accuracy and faster convergence across multiple trajectories compared to conventional ILC.

Abstract

This paper proposes a robust control strategy that integrates Iterative Learning Control (ILC) with a simple lateral neural network to enhance the trajectory tracking performance of a linear Lorentz force actuator under friction and model uncertainties. The ILC compensates for nonlinear friction effects, while the neural network estimates the nonlinear ILC effort for varying reference commands. By dynamically adjusting the ILC effort, the method adapts to time-varying friction, reduces errors at reference changes, and accelerates convergence. Compared to previous approaches using complex neural networks, this method simplifies online training and implementation, making it practical for real-time applications. Experimental results confirm its effectiveness in achieving precise tracking across multiple tasks with different reference trajectories.

Figures (10)

• Figure 1: Sample figure illustrating the ILC framework.
• Figure 2: Learning dynamics of ILC.
• Figure 3: Integration of ILC and neural network with Kalman filtering.
• Figure 4: Mechanical and electrical components of the test setup.
• Figure 5: ILC-based friction compensation for a single task.