用于协同感知的分布式聚类方法研究

Distributed Clustering Method for Cooperative Sensing

为了解决无中心协同感知条件下的多目标信息处理问题,提出了一种新的用于协同感知的分布式聚类方法。首先,构造了新颖的一致性类中心计算数学模型,使得各协同感知设备可并行计算自身采集的数据,各邻近设备之间仅仅需要交换暂态类中心,有效保护数据隐私性;然后,在感知网络达到稳态时所有感知设备获得一致的类中心和聚类结果。实验结果证实,本文所提出的分布式聚类方法可有效利用各协同设备的存储、计算、通信能力,在不交换大量数据的前提下完成协同聚类任务。

This paper presents a new distributed clustering method for cooperative sensing to solve the problem of multi-target information processing under the condition of decentralized cooperative sensing.Firstly,it formulates a new mathematical model for consensus class center computation with locally averaged data.Secondly,it enables all the sensing equipment to compute in parallel and exchange only the information with its neighboring nodes to protect the privacy of the captured data.Finally,all the equipment obtain the consensus class center and the clustering results when the network is stable.The experiment results prove that the distributed clustering method can make full use of the store,computation,and communication ability of the equipment to complete the clustering task without exchanging massive data.