A Survey on AI Technologies for UAV Swarms: Advancement of AI and the Evolution of Multi-Agent Architecture; [UAV 군집 AI 기술 동향: 인공지능 발전과 다중 에이전트 아키텍처의 진화 관점에서]

  • Kang, Guhyeon; 
  • Shin, Oh-Soon; 
  • Shin, Yoan; 
  • Lee, Jong-Ho; 
  • Kwon, Minhae
Citations

SCOPUS

0

초록

Unmanned Aerial Vehicle (UAV) swarm systems are a next-generation aerial robotics technology in which multiple UAVs autonomously collaborate to carry out complex missions, and their research scope has steadily expanded alongside advances in Artificial Intelligence (AI). This survey systematically reviews UAV swarm AI research from two perspectives. The first traces the evolution of AI techniques, from meta-heuristic swarm intelligence to Deep Reinforcement Learning (DRL) and Multi-Agent Reinforcement Learning (MARL), and further to Large Language Models (LLMs) and generative AI. The second analyzes the architectural transition from centralized control to semi-decentralized structures based on Centralized Training with Decentralized Execution (CTDE), and to fully decentralized structures enabled by federated learning. We also discuss the limitations of existing research and directions for future work. © 2026, Korean Institute of Communications and Information Sciences. All rights reserved.

키워드

Control Architecture; Multi-Agent Reinforcement Learning; Swarm Intelligence; UAV Swarm
제목
A Survey on AI Technologies for UAV Swarms: Advancement of AI and the Evolution of Multi-Agent Architecture; [UAV 군집 AI 기술 동향: 인공지능 발전과 다중 에이전트 아키텍처의 진화 관점에서]
저자
Kang, Guhyeon; Shin, Oh-Soon; Shin, Yoan; Lee, Jong-Ho; Kwon, Minhae
DOI
10.7840/kics.2026.51.8.1677
발행일
2026-08
유형
Article
저널명
한국통신학회논문지
권
51
호
8
페이지
1677 ~ 1691