Abstract: Settle time, ts, is defined as the elapsed time from rest until the measured position is contained within an acceptable distance from the target position. The objective of this research is to minimize ts for short unsaturated motions for each plant within a population of servomechanisms. The original motivating application was repetitive single track "seeks" in disk drives, operating in modes such as sequential data transfer, servo track writing, and scans for detecting media surface defects resulting from manufacturing. In each of these operating modes, it is desirable to have each drive achieve its minimum ts to increase data throughput or decrease manufacturing time and cost. Standard practices within the disk drive industry instead attempt to use a single robust seek control design that achieves a minimum level of performance across the entire population. Therefore, many of the units sacrifice performance in order for the population outliers to perform adequately.

Alternatively, our research focuses on adaptive methods which exploit the repetitive nature of the operating modes to minimize ts for each unit. Our solution procedure relies on a combination of adaptive model-inversion, typically used for output tracking, and adaptive desired trajectory generation to minimize settle time. In the pursuit of our research objective, we have made the following contributions:

• We provide a quantitative framework for evaluating model-inversion algorithms and architectures in a settle application, with particular attention paid to inversion algorithms for nonminimum phase systems. Algorithm complexity is also heavily weighted in this analysis, given the limited computational ability in the disk drive applications.
  
• We study the effects of parametric and nonparametric modeling uncertainty on settle performance when using inversion algorithms. We extend this analysis to experimental validation of these inversion algorithms in a settle application using disk drive hardware.
• We are currently working on devising an adaptive model inversion technique which makes use of repetitive measurements to minimize settle time in the presence of parametric uncertainty. We hope to show significant experimental settle improvements for each unit in the population using this technique by the fall of 2007.

While disk drives were the original motivating application for this research, minimizing settle time over a population is a ubiquitous objective. Applications as varied as atomic force microscopy, space-based imaging, and silicon wafer fabrication benefit from increased settle performance. We are currently investigating applications in some of these other areas.

This research has been partially funded by Maxtor Corporation and the National Science Foundation (CMS-0201495). My graduate advisers are Lucy Pao and Dale Lawrence with the University of Colorado at Boulder .