Application of Machine Learning for Adaptive Trajectory Control of UAVs Under Uncertainty

The article explores the potential of applying machine learning (ML) for adaptive trajectory control of unmanned aerial vehicles (UAVs) under uncertainty. The concepts of ML algorithms and the classification of UAVs by purpose, size, and weight are examined. To analyze control methods, theoretical approaches such as ensemble learning, neural networks, and probabilistic models are applied, enabling real-time adaptation of flight trajectories. Additionally, mathematical models are presented and illustrated with formulas describing the dynamics of interaction between the control system, external disturbances, and control inputs. Parameters such as system adaptability, trajectory correction accuracy, and stability under challenging conditions are studied to assess the accuracy and efficiency of the proposed algorithms. The study also investigates the impact of computational power limitations on the real-time performance of algorithms. The integration of data from various sensors is considered crucial for improving the accuracy and reliability of the control system. Special attention is given to the practical application of ML for environmental change prediction and flight trajectory optimization. Examples of real-world ML algorithm implementations include successful developments by Russian and foreign companies, demonstrating high levels of autonomy and adaptive control. The results show that ML significantly enhances UAV autonomy and safety, ensuring reliable trajectory corrections even under uncertain conditions. Further research could focus on developing collective control for UAV groups and improving real-time ML integration. This would expand UAV functionality, improve efficiency, and reduce resource consumption.