Abstract
The Zero-shot learning and supervised learning approaches for modeling and predicting external influences on unmanned aerial vehicles (UAVs) are considered. Supervised learning provides accurate predictions for conditions represented in the training dataset, while Zero-shot learning enables the system to adapt to novel, previously unseen situations, including unexpected weather events and external disturbances. The paper presents examples using linear regression and zero-shot classification to estimate the impact of various external conditions on UAV flight parameters, including drift angle. The proposed approaches can enhance UAV resilience to unpredictable environmental factors and improve simulation quality