A convergence analysis of D-ILC algorithm

ILC is an emerging technique for learning control. The D-ILC algorithm is the generic ILC scheme which captures the error trend of a batch to update the control input for the next batch or batches. The 2-dimensional nature requires in-depth convergence analysis of the algorithm. This paper addresses these issues in detail. This paper deals with the convergence properties of ILC algorithms with emphasis on control input. Discrete-time linear state space representation of a linear time-invariant system has been considered along with usual assumptions which ensure D-type ILC algorithm converges in terms of output error. The convergence for control input sequence is investigated up to component level.