One-shot optimization for vehicle dynamics control systems
André Thomaser, Anna V. Kononova, Marc-Eric Vogt, Thomas Bäck · Proceedings of the Genetic and Evolutionary Computation Conference Companion · 2022
Many real-world black-box optimization problems from industry are computationally expensive. Due to the advantage in wall-clock time, fully parallel sampling (one-shot search) is therefore often chosen over iterative search and adaptive sampling approaches. Our contribution shows how using a surrogate model (one-shot optimization with surrogate) can enhance the best solution found within the initial sample, requiring no further problem evaluations.