Preliminary Study of Self-adaptive Constrained Output Iterative Learning Control for Robotic Manipulators

Abstract

The Iterative Learning Control (ILC) can be applied within the control of industrial robots with uncertainties in their mathematical model for accomplishing precise trajectory tracking. The method calculates the tracking error and corrects the output control signals in accordance with a predefined learning operator. This research will focus on the influence of the ILC learning operator over the performance of the ILC. The proposed in this preliminary study algorithm for Self-adaptive Constrained Output ILC (SCOILC) will try to analyze how the learning operator can be optimized for the given task and afterwards to use this analysis in order to enhance the ILC performance when the task is slightly changed. The algorithm extends the convergent Constrained Output ILC (COILC) method. SCOILC adds an initial step of regression analysis of the executions of the COILC for selected trajectory with different learning operators. This allows the ILC procedure to self-adapt to the specific type of trajectories needed for execution of a specific domain of tasks. The conducted experiments led to a conclusion that this initial adaptation process reduces the overall convergence time of the ILC when it is applied to a modification of the trajectory which was used during the adaptation step.