Norm optimal ILC with time-varying weighting matrices

In this paper, we focus on improving performance and robustness in precision motion control (PMC) of multi-axis systems through the use of time-varying weighting matrices. A Norm Optimal (N.O.) framework is used to design optimal learning filters based on design objectives. The general N.O. framework is reformatted to include time-varying weighting matrices which enable the controller to take the trajectory, position-dependent dynamics, and time-varying disturbances into consideration when designing the optimal learning controller. A general approach for designing the different weighting matrices is included. The time-varying weighting approach of this framework enables one to focus on individual components that affect the system at different times throughout the trajectory independently. The performance benefits of time-varying weighting matrices are illustrated through simulation and experimental testing on a multi-axis robotic testbed.