Traffic digital twins provide realistic system-level testbeds for intelligent transportation system applications, including connected and automated vehicle impacts, energy analysis, and traffic control strategies. Their value depends on how well they reproduce the observed traffic conditions, making calibration against data essential. In practice, traffic digital-twin calibration is a black-box optimization problem in which each simulator evaluation is expensive and the parameter space may be high-dimensional. To address this challenge, we propose two complementary contributions within a Bayesian optimization framework: Memory-Guided Trust-Region Bayesian Optimization (MG-TuRBO) and an adaptive acquisition strategy. MG-TuRBO improves restart decisions by using search memory to identify basins that are both promising and under-explored, while the adaptive acquisition strategy uses the remaining evaluation budget to balance exploration and exploitation during candidate selection.We demonstrate the value of our method on transportation digital-twin calibration using two real-world traffic simulation problems with 14-dimensional (14D) and 84-dimensional (84D) calibration spaces. We compare the proposed framework against a genetic algorithm (GA), classical Bayesian optimization (BO), Trust-Region Bayesian Optimization (TuRBO), and Multi-TuRBO. The results show that Bayesian optimization methods reach target calibration quality much faster than GA in the 14D problem. In 84D, MG-TuRBO significantly outperforms Multi-TuRBO, GA, and BO, and is statistically on par with TuRBO, with the adaptive acquisition giving its best results. Our method is highly general and does not depend on the particulars of traffic simulation. This suggests that it is a promising, budget-aware approach for optimizing general high-dimensional black-box objectives from any field.
