Scholarlink

<p>Unmanned aerial vehicle (UAV) swarms are increasingly deployed for surveillance, crowd sensing, and operations in environments inaccessible to humans. This work presents a deep Q-learning–based dynamic swarm pattern formation (DSPF) framework to enable autonomous and adaptive coordination among multiple UAVs. The proposed model integrates a speed control–based reinforcement learning (SC-RL) algorithm to achieve efficient and collision-aware pattern generation and switching. A decentralized coordinate calculation (DCC) algorithm is introduced to parallelize coordinate computation and significantly reduce pattern formation time, while a servo interrupt–based pattern switch (SIPS) control mechanism enables real-time pattern transitions. Simulation studies conducted with 100 UAVs demonstrate the effectiveness of the proposed approach in dense, collision-prone environments, achieving improvements of approximately 95.68 % in pattern formation time and 66.67 % in distance efficiency. The results confirm the feasibility and robustness of the DSPF framework for scalable UAV swarm operations.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Advanced RL-infused swarm dynamics for reconfigurable UAV formations

  • 期刊论文

  • 发表日期：

    2026年5月12日

查看全文
  • Venkatesan M.,
  • Senthil Kumar K.,
  • Vasantharaj Rajagopal

摘要

Unmanned aerial vehicle (UAV) swarms are increasingly deployed for surveillance, crowd sensing, and operations in environments inaccessible to humans. This work presents a deep Q-learning–based dynamic swarm pattern formation (DSPF) framework to enable autonomous and adaptive coordination among multiple UAVs. The proposed model integrates a speed control–based reinforcement learning (SC-RL) algorithm to achieve efficient and collision-aware pattern generation and switching. A decentralized coordinate calculation (DCC) algorithm is introduced to parallelize coordinate computation and significantly reduce pattern formation time, while a servo interrupt–based pattern switch (SIPS) control mechanism enables real-time pattern transitions. Simulation studies conducted with 100 UAVs demonstrate the effectiveness of the proposed approach in dense, collision-prone environments, achieving improvements of approximately 95.68 % in pattern formation time and 66.67 % in distance efficiency. The results confirm the feasibility and robustness of the DSPF framework for scalable UAV swarm operations.