Iterative learning control using prediction with arbitrary horizon

A novel Iterative Learning Control algorithm is derived based on previous optimal algorithms. The algorithm is characterised by an optimisation criterion that takes the present and also future trials into account, in a similar way to Model Predictive Control. One of the important properties, which are shown in the paper, is a monotonically decreasing error with a fast rate of convergence that can also be controlled by choice of the prediction horizon and choice of design weights. The numerical implementation uses classical Linear Quadratic Regulator methods and a number of parallel plants. Simulation results visualise the properties of the algorithm.