UAV AUTONOMY: CLASSIFICATION, OPERATIONAL FEATURES AND ENHANCEMENT APPROACHES

Abstract

The paper presents a systematic study of autonomy levels of unmanned aerial vehicles and defines their main functional characteristics considering recent advances in robotics and artificial intelligence. Existing approaches to UAV autonomy classification are summarized according to the degree of human involvement in control, navigation, mission planning and decision making. The operational features of UAV systems at different autonomy levels are analyzed and their impact on mission effectiveness, operational reliability and adaptability to dynamic environments is determined. Recent scientific studies devoted to artificial intelligence, computer vision, multi sensor data fusion and intelligent control architectures are reviewed. Particular attention is paid to emerging technologies of 2025–2026 including the use of large language models for behavior generation, reinforcement learning for autonomous navigation, vision language navigation frameworks and distributed cloud edge computing solutions. Key technical constraints limiting autonomy growth such as energy capacity, onboard processing performance, sensor accuracy and communication robustness are identified. Promising directions for improving UAV autonomy are formulated including the development of autonomous decision making algorithms, cooperative swarm interaction, adaptive control strategies and energy efficient system design. It is shown that the integration of these technologies significantly improves situational awareness and enables long duration autonomous missions with minimal operator supervision. The obtained results may be applied in the design of next generation unmanned aerial platforms and advanced autonomous control software. The presented analysis also provides a basis for further research focused on cognitive autonomy and resilient UAV architectures. The study emphasizes the importance of combining algorithmic intelligence with reliable hardware components. The conclusions have practical engineering relevance. The findings are applicable to real UAV missions today.