UAV path planning, as a key technology in the field of automatic control and intelligent systems, has demonstrated significant potential in various applications, including logistics and distribution, environmental monitoring, and emergency rescue. A comprehensive reassessment of the existing representative literature reveals that most reviews in this field focus on specific aspects and are largely confined to methodological investigations, primarily qualitative analyses that lack empirical data to support their conclusions. To address this gap, this study employs the mapping knowledge domain (MKD) method of bibliometrics, utilizing CiteSpace, VOSviewer, and Bibliometrix R package to analyze a total of 4416 documents from the Web of Science Core Collection (WOSCC) spanning from 2000 to 2024. Through retrospective analysis and scientific knowledge mapping, we first review the development of UAV path planning and categorize it into four distinct stages. Secondly, we identify key external features of the field. Using techniques such as co-citation analysis and keyword clustering, we then identify research trends, burst papers, and hotspots. Finally, we highlight five typical application scenarios of UAV path planning. The results of the study indicate that the field of UAV path planning has made significant advancements over the past two decades, particularly since 2018. These studies encompass various disciplinary areas, underscoring the increasing necessity for the integration of multidisciplinary approaches to UAV path planning in recent years. The aim of this study is to provide researchers with a comprehensive reference and new research perspectives while offering technical guidelines for professionals working in related applications.

Defending against adversarial unmanned aerial vehicle (UAV) swarms presents a critical challenge for modern security systems, requiring coordinated strategies under partial observability. Existing multi-agent reinforcement learning (MARL) and spatio-temporal graph neural networks (ST-GCN) methods typically treat agents as homogeneous nodes, failing to effectively model large-scale, dynamic spatio-temporal dependencies to support cooperative decision-making, and lacking semantic interpretation of adversarial tactics. To address this, we propose the Spatio-Temporal Attention Graph-Enhanced policy optimization (STAGE). Unlike ST-GCN that uses fixed or predefined graph structures, STAGE models the diverse swarm topology through a learnable dynamic adjacency matrix and a multi-hop neighborhood aggregation mechanism to capture dependencies at different ranges. Moreover, to overcome the black-box nature of existing adversarial analysis, we design a dual-path intent recognition framework, which concatenates cluster-level features from graph attention networks with computed tactical metric vectors, and trains a classifier for four types of swarm tactical intents, enabling explicit recognition of enemy intent by the defenders. The policy is optimized end-to-end within the MAPPO framework through a structured multi-objective reward function. Extensive experiments in multi-scale area defense scenarios demonstrate that STAGE outperforms state-of-the-art methods in task performance and tactical comprehension robustness.