Local Bayesian Optimization for Controller Tuning with Crash Constraints

TL;DR

This paper introduces a local Bayesian optimization method that efficiently tunes controllers with crash constraints, significantly reducing tuning time and resources through simulations and hardware experiments.

Contribution

It extends local Bayesian optimization to handle crash constraints in controller tuning, addressing high-dimensional search space challenges.

Findings

Enhanced controller performance demonstrated in simulations.

Reduced tuning time and resource consumption.

Effective handling of crash constraints in real-world experiments.

Abstract

Controller tuning is crucial for closed-loop performance but often involves manual adjustments. Although Bayesian optimization (BO) has been established as a data-efficient method for automated tuning, applying it to large and high-dimensional search spaces remains challenging. We extend a recently proposed local variant of BO to include crash constraints, where the controller can only be successfully evaluated in an a-priori unknown feasible region. We demonstrate the efficiency of the proposed method through simulations and hardware experiments. Our findings showcase the potential of local BO to enhance controller performance and reduce the time and resources necessary for tuning.