Feedforward Motion Control: From Batch-to-Batch Learning to Online Parameter Estimation

Feedforward control is essential in high-performance motion control. The aim of this paper is to develop a unified framework for automatic feedforward optimization from both batch-wise data sets as well as real-time data. A statistical analysis is employed to analyze the effect of noise, i.e., an iteration varying disturbance, on feedforward controller performance. This provides new insights, both potential advantages as well as possible hazards of real-time estimation are considered. Finally, a case study confirms and illustrates the results.