Iterative Learning Control in Precision Motion Systems

Iterative Learning Control (ILC) is a specialised methodology developed to enhance the performance of systems that repetitively execute the same trajectory or task. By harnessing data from successive trials, ILC algorithms update control inputs to minimise tracking errors, thereby achieving improved precision and faster convergence over iterations. This approach is particularly valuable in high-accuracy applications such as wafer scanners, nanopositioning stages, semiconductor lithography, and additive manufacturing, where positioning errors on the order of nanometres can have profound effects on product quality. ILC techniques accommodate model uncertainties and unmodelled dynamics by refining feedforward profiles—often in combination with feedback loops—and by incorporating robust filtering schemes to ensure stability. Recent advances include convex-optimisation frameworks for data-driven learning filters and refined filter designs that balance convergence speed against noise amplification. Across industrial and laboratory settings, ILC has demonstrated the capacity to reduce settling time, suppress periodic disturbances, and deliver sub-micrometre tracking performance, underlining its global significance in precision engineering and its role in enabling next-generation manufacturing and metrology systems.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Research from all publishers

Recent studies have introduced a data-driven ILC framework formulated as a convex-optimisation problem. By utilising measured frequency-response data rather than low-order parametric models, the method designs learning filters that guarantee convergence while accounting for unmodelled dynamics. A case study on a power-converter system for a particle accelerator demonstrated robust tracking with minimal iteration count, underscoring the approach’s adaptability to diverse hardware.

Another advancement centres on a model-based ILC algorithm incorporating a novel Q-filter configuration. Traditional low-pass Q-filters, while improving robustness, tend to compromise transient performance. The new design yields a superior compromise between noise attenuation and learning gain, leading to faster convergence and enhanced tracking of complex wafer-stage trajectories. Experimental validation on a semiconductor wafer-scanning platform confirmed both theoretical predictions and practical viability, with significant reduction of repetitive error components across iterations.

Iterative Learning Control in Precision Motion Systems publication trend

The graph below shows the total number of articles in iterative learning control in precision motion systems across all publications each year (not limited to Nature Index journals).

Technical terms

Iterative Learning Control (ILC): A control strategy for systems executing identical tasks, which refines future control inputs by learning from past iteration errors to improve tracking accuracy.

Q-filter: A low-pass filter within ILC algorithms that attenuates high-frequency update components, ensuring stable convergence in the presence of noise and unmodelled dynamics.

Feedforward Control: A proactive control approach that computes inputs based on a system model to achieve the desired output trajectory before error occurs.

Tracking Error: The instantaneous deviation between the desired reference trajectory and the actual system output during a single execution.

Convergence Rate: The speed at which the ILC algorithm reduces tracking error across successive iterations towards an acceptable threshold.

References

1. Data‐driven approach to iterative learning control via convex optimisation. IET Control Theory and Applications (2020).
2. Model‐Based ILC with a Modified Q‐Filter for Complex Motion Systems: Practical Considerations and Experimental Verification on a Wafer Stage. Complexity (2018).