Data-Driven ILC for Trajectory Tracking in Nonlinear Dynamic Systems

The aim of this work is to propose a data-driven ILC algorithm that features fast convergence for nonlinear dynamic systems. This idea utilizes adaptive filtering that implicitly identifies the time-varying system inverse along the trajectory being tracked. By feeding the error signal through the copied inverse filter, it results in a rapidly convergent inversion-based ILC. This approach is compared to a nonlinear extension of the data-driven ILC that uses system adjoint as the learning filter. The developed algorithm is validated through simulation on a fully actuated 2 DOF Furuta pendulum.