Optimal ILC

Abstract

In common with many other applications, control law design based on minimizing a suitable cost function has continued to be an active area in iterative learning control (ILC) research and applications. This chapter considers norm optimal iterative learning control (NOILC), where the cost function penalizes the difference between the control input vectors on successive trials. In NOILC, the weighting matrices Q and R adjust the balance, respectively, between trial-to-trial error convergence speed and robustness. It is generally recognized that there are three variables that are of particular importance when describing the performance of an ILC law, that is, trial-to-trial error convergence speed, minimum tracking error, and long-term performance. Predictive action in ILC design can be from trial-to-trial or along the trials. The chapter considers the former case, and then for the latter, control law development is considered with supporting experimental validation.