Issue |
JNWPU
Volume 42, Number 6, December 2024
|
|
---|---|---|
Page(s) | 1030 - 1038 | |
DOI | https://doi.org/10.1051/jnwpu/20244261030 | |
Published online | 03 February 2025 |
Decentralized and autonomous behavior decision-making for UAV cluster
面向去中心化的集群自主行为决策研究
1
School of Astronautics, Northwestern Polytechnical University, Xi’an 710072, China
2
Shanghai Institute of Mechanical and Electrical Engineering, Shanghai 201109, China
Received:
23
October
2023
It is difficult for conventional methods like the diagram theory in a complex environment to carry out modeling and calculation so as to make large-scale cluster behavior decisions. Hence, this paper studies small fixed wings and establishes the decentralized behavior decision-making model for a UAV cluster that has communication limitations and scale ceiling effects. The idea of swarm intelligence is combined with the decoupling multi-agent deep deterministic strategy gradient (DE-MADDPG) for the constructed model to do adaptive learning. Finally, the optimal behavior decision of the UAV cluster is made. Simulations are carried out to verify the model. The consistent movement of the UAV cluster and the maneuvering obstacle avoidance behavior in complex environments are realized. Compared with the MADDPG, the DE-MADDPG exhibits superior precision and real-time capability.
摘要
为解决复杂环境下图论等传统方法在面临大规模集群行为决策时难以建模和计算的问题, 以小型巡飞弹为研究对象, 针对无人机集群通信受限和规模天花板效应等挑战, 建立了去中心化的无人机集群行为决策模型, 结合群体智能的思想和解耦型多智能体深度确定性策略梯度(DE-MADDPG)策略, 对构建的模型进行自适应学习, 得出无人机群体的最优行为决策。通过开展仿真验证, 实现了无人机集群的一致性运动以及复杂环境下的机动避障行为。相比于MADDPG, DE-MADDPG具有更强的准确性和实时性。
Key words: UAV cluster / autonomous behavior decision-making / multi-agent deep reinforcement learning / decentralization / consistent movement / obstacle avoidance
关键字 : 无人机集群 / 自主行为决策 / 多智能体深度强化学习 / 去中心化 / 一致性运动 / 避障
© 2024 Journal of Northwestern Polytechnical University. All rights reserved.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.