| 摘要: |
| 近年来，微小型无人机系统已成功应用于战场侦察与打击领域。多机协同平台能够对中大型、高机动目标开展序贯作业，进而提升战术打击效能。然而，在复杂对抗环境下，目标外观与运动状态频繁变化，易导致表征更新不稳定及跟踪漂移；同时，分布式协作带来的灵活性也可能使目标因规避而脱离视野。为此，提出一种面向序贯作业的高效协同式目标跟踪方法。通过自适应选择模块动态筛选关键特征层，实现轻量化设计并增强对高机动目标的实时感知。引入不对称注意力机制建模搜索区域与局部目标关系，以突出关键表征、抑制干扰。结合跨无人机映射策略，使失效平台能够快速重定义搜索区域。通过多无人机单目标检测数据集的实验表明，所提方法的准确率和精度相比基线模型分别提升了7.1%与3.9%。 |
| 关键词: 无人机集群 目标跟踪；分布式系统；序贯作业；边缘计算 |
| DOI：10.20079/j.issn.1001-893x.251120011 |
|
| 基金项目:国家自然科学基金面上项目（62571185）；国家自然科学基金青年科学基金项目（62201209） |
|
| A Distributed Cooperative Target Tracking Method for Sequential Missions with Multiple UAVs |
| LI Zijiana,WEI Jianxina,LI Qingpenga,QIN Yuliangb,DUAN Li |
| (1a.School of Artificial Intelligence and Robotics;1b.College of Computer Science and Electronic Engineering,Hunan University,Changsha 410082,China;2.School of Electronic Engineering,Naval University of Engineering,Wuhan 430033,China) |
| Abstract: |
| In recent years,micro unmanned aerial vehicle(MUAV) systems have been successfully applied in battlefield reconnaissance and strike missions.Multi-UAV collaborative platforms can perform sequential operations against large,highly maneuverable targets,thereby enhancing tactical strike effectiveness.However,in complex adversarial environments,frequent variations in target appearance and motion often lead to unstable feature updates and tracking drift.Meanwhile,the flexibility introduced by distributed collaboration may cause targets to evade and leave the field of view.To address these challenges,an efficient collaborative target tracking method for sequential missions is proposed.Specifically,an adaptive selection module is designed to dynamically filter key feature layers,achieving a lightweight design while enhancing real-time perception of highly maneuverable targets.An asymmetric attention mechanism is introduced to model the relationship between the search region and local targets,highlighting key representations and suppressing interference.Furthermore,a cross-UAV mapping strategy is incorporated to allow a failing platform to quickly re-define its search region.The experiment conducted on the multi-drone single-target detection dataset MDOT shows that the accuracy and precision of the method are improved by 7.1% and 3.9%,respectively,compared with that of the baseline model. |
| Key words: UAV swarm target tracking distributed systems sequential operations edge computing |
《电讯技术》编辑部 