Iterative Learning Control algorithm with self-adaptive steps

Most Iterative Learning Control (ILC) algorithms being used currently still follow the essence of the original idea of Arimoto, in which the control error with its derivative and integral are composed to deduce the iterative-learning-rate. Although these ILCs gain a wide range of successful applications in various control fields, they are still with intrinsic shortcomings of conventional ILC such as huge computation, big storage capacity demanded for algorithm data, and the control law is sensitive to control errors. To solve these problems and enhance ILC performances, a novel ILC with self-adaptive steps is proposed as a new approach. In this approach, the control step for each iteration is set according to the error sum of the last time iteration, and the step value will decrease as the iteration goes which could obtain the merits of quick convergence speed at early stage and high control precision at late stage of control process; while the sign of the step is decided by the instant sampling value of the control error. The stability of the algorithm is analyzed using newly designed adaptive steps. First the validity and robust of the approach is tested through simulation on a object with randomly set disturbance, secondly the feasibility of the approach to solve real problem is verified through trajectory tracking control on a robot joint driven by valve-serving-cylinder with time-variant parameters, just like a real hydraulic system in practice. Both simulation and experimental results demonstrate that the self-adaptive steps method proposed in this paper could reduce iterative calculation and storage size of the algorithm, it also avoids error amplification in error differentiating. Moreover, the method is with easy-to- set steps within a wide range.