Data-efficient Autotuning with Bayesian Optimization: An Industrial Control Study
arXiv (PDF)
Bayesian optimization is proposed for automatic learning of optimal controller parameters from experimental data. A probabilistic description (a Gaussian process) is used to model the unknown function from controller parameters to a user-defined cost. The probabilistic model is updated with data, which is obtained by testing a set of parameters on the physical system and evaluating the cost. In order to learn fast, the Bayesian optimization algorithm selects the next parameters to evaluate in a systematic way, for example, by maximizing information gain about the optimum. The algorithm thus iteratively finds the globally optimal parameters with only few experiments. Taking throttle valve control as a representative industrial control example, the proposed auto-tuning method is shown to outperform manual calibration: it consistently achieves better performance with a low number of experiments. The proposed auto-tuning framework is flexible and can handle different control structures and objectives.
| Author(s): | Matthias Neumann-Brosig and Alonso Marco and Dieter Schwarzmann and Sebastian Trimpe |
| Links: | |
| Journal: | IEEE Transactions on Control Systems Technology |
| Volume: | 28 |
| Number (issue): | 3 |
| Pages: | 730--740 |
| Year: | 2020 |
| Month: | May |
| Project(s): | |
| Bibtex Type: | Article (article) |
| DOI: | 10.1109/TCST.2018.2886159 |
| State: | Published |
| Electronic Archiving: | grant_archive |
BibTex
@article{NeuMarSchTri18,
title = {Data-efficient Autotuning with Bayesian Optimization: An Industrial Control Study},
journal = {IEEE Transactions on Control Systems Technology},
abstract = {Bayesian optimization is proposed for automatic
learning of optimal controller parameters from experimental
data. A probabilistic description (a Gaussian process) is used
to model the unknown function from controller parameters to
a user-defined cost. The probabilistic model is updated with
data, which is obtained by testing a set of parameters on the
physical system and evaluating the cost. In order to learn fast,
the Bayesian optimization algorithm selects the next parameters
to evaluate in a systematic way, for example, by maximizing
information gain about the optimum. The algorithm thus iteratively
finds the globally optimal parameters with only few
experiments. Taking throttle valve control as a representative
industrial control example, the proposed auto-tuning method is
shown to outperform manual calibration: it consistently achieves
better performance with a low number of experiments. The
proposed auto-tuning framework is flexible and can handle
different control structures and objectives.},
volume = {28},
number = {3},
pages = {730--740},
month = may,
year = {2020},
slug = {bayesianindustrial2018},
author = {Neumann-Brosig, Matthias and Marco, Alonso and Schwarzmann, Dieter and Trimpe, Sebastian},
month_numeric = {5}
}