Iterative Learning Control (ILC) fundamentally suffers from sensitivity to uncertainties, and poor cross-task generalization. Addressing these limitations, we propose the first unified Meta Learning-enhanced Iterative Learning Control (Meta-ILC) framework—integrating meta-learning principles with neural network-based adaptive control. Our approach replaces fixed ILC gain matrices with context-sensitive operators \(L_p(t)\) generated by deep residual networks, while a meta-optimizer extracts transferable knowledge across tasks to initialize near-optimal controllers. The framework autonomously adapts to system variations without model reliance or manual tuning, resolving the core deficiencies of conventional ILC. Experimental validation confirms significant advantages: accelerated convergence, minimal initial tracking errors, and robust performance across unseen trajectories, demonstrating transformative potential for high-precision applications including multi-axis robotics and semiconductor manufacturing.