System Analysis of Architectural Solutions and Algorithmic Approaches to Integrating Multimodal Artificial Intelligence in Autonomous Unmanned Aerial Vehicles

The unmanned aerial vehicle (UAV) sector is shifting from deterministic GNSS-based navigation to cognitive systems with multimodal AI.Traditional methods fail in GPS-denied environments, while vision-language models (VLMs) enable semantic understanding and natural language tasks.This analysis evaluates architectures, models, and state of the art (SOTA) projects in this field.Recommendations focus on low-latency video links, quantization, and simulation for reliable UAV control.