Local Bayesian Optimization for Controller Tuning with Crash Constraints
This work addresses controller tuning for robotics or control systems, but it is incremental as it builds on a recently proposed local BO method by adding crash constraints.
The paper tackled the problem of automated controller tuning in large, high-dimensional spaces with crash constraints, where evaluations fail outside an unknown feasible region, by extending a local Bayesian optimization method to handle these constraints, resulting in enhanced controller performance and reduced tuning time and resources in simulations and hardware experiments.
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.