摘要
在搜救任务中,移动障碍物的出现以及环境的突变等因素,显著增加无人机集群执行搜救任务时的安全风险。因此,提出基于深度强化学习的无人机集群协同搜救任务智能分配方法。构建基于深度强化学习的无人机集群协同搜救任务智能分配框架,该框架通过智能体、经验池、评价网络、策略网络等进行交互,实现无人机集群协同搜救任务智能分配。实验结果显示,该方法的无人机集群协同搜救任务智能分配结果中,无人机间均能够保持安全距离,并且各无人机快速覆盖不同搜救目标点,未出现路径交叉情况,实现高鲁棒性协同搜救任务分配。且在安全性、实时性和任务完成效率三个维度均表现出色,充分证明该方法在动态复杂环境中的优越性能。
In search and rescue missions,the presence of moving obstacles and sudden changes in the environment significantly increase the safety risks of drone swarms during search and rescue operations.Therefore,a deep reinforcement learning based intelligent allocation method for collaborative search and rescue tasks in drone clusters is proposed.Build an intelligent allocation framework for collaborative search and rescue tasks in drone clusters based on deep reinforcement learning.This framework interacts with intelligent agents,experience pools,evaluation networks,strategy networks,etc.to achieve intelligent allocation of collaborative search and rescue tasks in drone clusters.The experimental results show that in the intelligent allocation of drone cluster collaborative search and rescue tasks using this method,safe distances can be maintained between drones,and each drone quickly covers different search and rescue target points without path crossing,achieving high robustness in collaborative search and rescue task allocation.And it performs well in three dimensions:security,real-time performance,and task completion efficiency,fully demonstrating the superior performance of this method in dynamic and complex environments.
作者
宋蓓蓓
余战秋
Song Beibei;Yu Zhanqiu(School of Computer and Art,Anhui Technical College of Industry and Economy,Hefei,Anhui 230051,China)
出处
《黑龙江工业学院学报(综合版)》
2025年第10期118-123,共6页
Journal of Heilongjiang University of Technology(Comprehensive Edition)
基金
安徽省教学研究一般项目“基于多源异构数据的知识图谱在计算机网络教学中的应用研究”(项目编号2023jyxm1502)。
关键词
深度强化学习
无人机
搜救任务
智能分配
Q学习算法
奖励收益
deep reinforcement learning
unmanned aerial vehicle
search and rescue mission
intelligent allocation
Q-learning algorithm
reward income
