面向协同感知的无人蜂群智能资源调度方案  

Intelligent resource scheduling scheme for UAV swarm collaborative sensing

作　　者：赵鹏程 李天扬 冷甦鹏 熊凯 ZHAO Pengcheng;LI Tianyang;LENG Supeng;XIONG Kai(School of Information and Communication Engineering,University of Electronic Science and Technology of China,Chengdu 611731,China)

机构地区：[1]电子科技大学信息与通信工程学院,四川成都611731

出　　处：《电信科学》2025年第3期17-26,共10页Telecommunications Science

基　　金：国家自然科学基金青年项目(No.62201122);四川省科技计划项目(No.2024YFHZ0321)。

摘　　要：随着低空经济的蓬勃发展,无人机在监测和感知领域得到广泛应用。然而,无人机有限的机载计算资源制约了感知数据的高效处理。此外,协同感知中产生的观测区域重叠进一步增加了冗余的计算负担。同时,无人机网络的高动态拓扑和节点资源的波动性大幅加剧了资源协同的难度。针对上述挑战,提出了一种面向协同感知的无人蜂群智能资源调度方案,通过自适应感知模式、分步卸载计算任务和竞价带宽策略,实现了通信、感知、计算(通感算)异构资源协同互补,提升协同感知效率。采用基于注意力机制的多智能体强化学习算法求解优化问题,以增强智能体提取环境关键特征的能力。仿真结果表明,与基准方案相比,该方案不仅有效降低了感知任务的执行时间,还提高了计算资源的利用率。With the rapid development of the low-altitude economy,unmanned aerial vehicles(UAV)have been widely applied in monitoring and sensing tasks.However,the limited onboard computing resources of UAV constrain the efficient processing of sensing data.Moreover,overlapping observation areas in collaborative sensing introduce additional computational redundancy.Meanwhile,the highly dynamic network topology and fluctuating node re sources significantly increase the complexity of resource coordination.To address these challenges,an intelligent re source scheduling scheme for UAV swarm collaborative sensing was proposed.Adaptive sensing mode selection,step wise computation offloading,and competitive bandwidth allocation were integrated to achieve heterogeneous re source coordination across communication,sensing,and computation(CSC),thereby enhancing collaborative sensing efficiency.Furthermore,a multi-agent reinforcement learning(MARL)algorithm with an attention mechanism was employed to solve the optimization problem,enabling agents to extract critical environmental features more effec tively.Simulation results demonstrate that,compared with benchmark schemes,the proposed scheme significantly re duces the execution time of sensing tasks while improving computational resource utilization.

关 键 词：协同感知 无人蜂群 资源调度 多智能体强化学习 注意力机制 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]