Core Concepts
Efficiently tuning controller parameters using a guided Bayesian optimization algorithm with a digital twin.
Abstract
This article introduces the Guided Bayesian Optimization algorithm for data-efficient controller tuning using a digital twin. The methodology is model-free and suitable for nonlinear plants, improving data efficiency by utilizing available information in the closed-loop system. Experimental results demonstrate superior performance compared to traditional Bayesian optimization methods.
Introduction to Guided Bayesian Optimization for controller tuning.
Methodology overview focusing on the use of a digital twin.
Analysis of experimental results showcasing improved data efficiency.
Comparison with traditional Bayesian optimization techniques.
Stats
"Our method requires 57% and 46% fewer experiments on the hardware than Bayesian optimization to tune the control parameters of the linear and rotary motor systems."