Scholarlink

<p>In precision farming, the combination of Swarm Intelligence (SI) and Deep Reinforcement Learning (DRL) to enhance cooperation of Unmanned Aerial Vehicles (UAVs) in smart agriculture systems has been introduced and developed as a revolutionary method. In this paper, a multi-agent deep reinforcement learning (MADRL) is proposed that allows UAVs to autonomously collaborate, adapt, and optimize resource allocation (water fertilization, pest control), crop monitoring, and pest control of vast farmlands. The system improves task scheduling, coverage efficiency and decision-making in dynamic agricultural environments by adopting bio-inspired swarm intelligence (SI) techniques including Ant Colony Optimization (ACO) and Particle Swarm Optimization (PSO). Through simulations, we show that our approach can outperform heuristic and rule-based models that do not use DRL in terms of energy efficiency, task completion time, and real-time adaptability. When validated on a real-world agricultural dataset, the robustness of the model is confirmed; the operational efficiency of the models outputs exceeds that of conventional approaches by over 20–30%.</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

MADRL-SI: a multi-agent deep reinforcement learning-based swarm intelligence approach for coordinating multiple UAVs in smart agriculture systems

  • 期刊论文

  • 发表日期：

    2025年11月22日

查看全文
  • Adel Saad Assiri,
  • Fahad kamal Alsheref,
  • Mostafa Al Ghamdi,
  • Mohamed Abbassy

摘要

In precision farming, the combination of Swarm Intelligence (SI) and Deep Reinforcement Learning (DRL) to enhance cooperation of Unmanned Aerial Vehicles (UAVs) in smart agriculture systems has been introduced and developed as a revolutionary method. In this paper, a multi-agent deep reinforcement learning (MADRL) is proposed that allows UAVs to autonomously collaborate, adapt, and optimize resource allocation (water fertilization, pest control), crop monitoring, and pest control of vast farmlands. The system improves task scheduling, coverage efficiency and decision-making in dynamic agricultural environments by adopting bio-inspired swarm intelligence (SI) techniques including Ant Colony Optimization (ACO) and Particle Swarm Optimization (PSO). Through simulations, we show that our approach can outperform heuristic and rule-based models that do not use DRL in terms of energy efficiency, task completion time, and real-time adaptability. When validated on a real-world agricultural dataset, the robustness of the model is confirmed; the operational efficiency of the models outputs exceeds that of conventional approaches by over 20–30%.