Performance iterative learning control on nonlinear systems for tracking varying-scale and morphologically similar targets

This paper proposes a novel three-dimensional (3D) performance iterative learning control (PILC) scheme on nonlinear systems for tracking varying-scale and morphologically similar targets. The condition and performance weighted K-nearest neighbour (KNN) and the historical database update algorithms are developed to match the historical control inputs closest to the current operating conditions, and further realise the learning process along the performance domain. The control inputs of the PILC scheme can be updated by utilising the control information obtained from the time, iteration and performance dimensions. Also, the convergence property of the proposed framework is proved via the contraction mapping principle. Finally, compared with the existing ILC works, the illustrative simulation and experimental results are presented to illustrate that the PILC scheme has better tracking performance and robustness, smaller initial error, and faster convergence speed.