Cooperative Game Theory Unlocks Better ILC Performance for Untrackable Repetitive Tasks
Standard Iterative Learning Control (ILC) assumes perfect trackability — that an input exists capable of driving system output to match a desired trajectory. This assumption often fails in practice, leaving conventional ILC suboptimal. This paper proposes an end-to-end ILC framework for closed-loop systems where both the reference input and the feedforward ILC input are jointly updated trial-to-trial using measurement data, modeled as a two-player cooperative game. The authors derive a sufficient condition under which this two-player scheme achieves strictly lower cost than single-player Norm-Optimal ILC (NOILC), backed by numerical validation.