Temperature and humidity calibration chambers, which provide controlled environments for instrument testing and validation, are widely applied in the aerospace and biomedicine fields. However, traditional fixed calibration points fail to adapt to complex operational requirements and exhibit problems including a limited coverage range and low efficiency. To address these challenges, this study develops a Gaussian Process-based Multi-Fidelity Bayesian Optimization (GP-MFBO) framework for optimal selection of temperature and humidity calibration points. The framework integrates the following three key components: (1) a three-layer progressive multi-fidelity modeling system comprising physical analytical models, computational fluid dynamics (CFD) numerical simulations, and experimental verification; (2) a systematic uncertainty quantification system covering model uncertainty, parameter uncertainty, and observation uncertainty; and (3) an adaptive acquisition function that balances uncertainty penalty mechanisms and multi-fidelity information gain evaluation. The experimental results demonstrate that the proposed GP-MFBO method achieves optimal calibration point combinations with a temperature uniformity score of 0.149 and humidity uniformity score of 2.38, approaching theoretical optimal solutions within 4.5% and 3.6%, respectively. Compared to standard Gaussian process, Co-Kriging, two-stage optimization, polynomial regression, and traditional single-fidelity methods, GP-MFBO achieves uniformity score improvements of up to 81.7% and 76.3% for temperature and humidity, respectively. The prediction confidence interval coverage reaches 94.2%, outperforming all comparative methods. This research provides a rigorous theoretical foundation and technical solution for the scientific design and reliable operation of large-space temperature and humidity calibration systems.
Accurate real-time positioning of Unmanned Aerial Vehicles (UAVs) is critical for navigation and mapping but remains challenging in complex environments due to signal blockages and multipath effects. This study presents a comparative framework for real-time error prediction of the Global Navigation Satellite System (GNSS), evaluating two machine learning models (Random Forest and XGBoost) and a deep learning model (Long Short-Term Memory network) against an Extended Kalman Filter baseline. A high-precision total station provides ground-truth coordinates, enabling the derivation of positioning error labels from synchronized GNSS raw data. Among the evaluated models, the tree-based XGBoost model achieves a significantly lower Mean Squared Error (MSE) and a considerably higher Coefficient of Determination () score than other models in predicting positioning deviations. The high-accuracy error predictions from the optimal model establish the core of a software-only solution for positioning integrity. The framework demonstrates that reliable, real-time error estimates can be derived directly from observation data, providing the essential input required for future compensation systems without necessitating additional hardware.