Predictive gradient iterative learning control

Iterative learning control (ILC) is a control design method for high performance trajectory tracking. Most existing results achieve this by learning from information collected over past executions of the task (named trials). This paper proposes a novel gradient based ILC design which updates the control input by learning not only from past trials but also from the predicted future trials using plant model knowledge. It is shown that by including information from predicted future trials, the designed ILC controller is less short-sighted and therefore better performance can be achieved. Analysis of the algorithm´s properties reveals potentially substantial benefit in terms of convergence speed; the proposed algorithm also possesses distinct robustness features with respect to model uncertainty. Numerical simulations are provided to demonstrate the effectiveness of the proposed method.