多轨分割网络大规模节点计算效率控制仿真

Multi-track Segmentation Network Large-Scale Node Computing Efficiency Control Simulation

当前方法无法高效提升节点计算效率同时降低计算能耗,提出基于HNCTC的多轨分割网络大规模节点计算效率控制方法。以节点度、网络链路、网络密度、聚类系数和熵等影响因素分析为依据,引入HNCTC算法实现节点计算效率控制。考虑到网络运行时各节点承受的负载不同,规定条件下,簇头在簇内消耗的能量大致相同,距离计算簇头比较近的数据簇能量消耗比较大,全网能量消耗不均等,采用非均匀最小均方减法分簇,依据簇内节点位置信息筛选出簇头,基于新簇能量消耗信息确定更新簇头能量阈值,通过减少计算节点在簇内能耗,均衡参与计算的数据簇负载,根据均衡数据簇负载调整节点计算效率。在簇头节点收到一帧数据之后进行数据计算融合,将融合后的数据传输至下一个簇头节点,判断簇头节点能量是否低于能量阈值,假设为是,则启动簇头更新机制,进入新一轮节点计算效率控制流程。实验结果表明,上述方法节点计算效率和能耗控制效果均优于当前方法。

The current method cannot improve the computing efficiency of nodes and reduce the computational energy consumption. Therefore, this article presented a method to control the computing efficiency of large-scale node in multi-track segmentation network based on HNCTC . Based on the analysis of influencing factors such as node degree, network link, network density, clustering coefficient and entropy, we introduced HNC TC algorithm to realize the control of node computing efficiency. The load on each node was different when the network was running. Based on the specified condition, the energy consumed by the cluster head in the cluster was roughly the same. The energy consumption of data cluster which was relatively close to the cluster head was relatively large, and the energy consumption of whole network was heterogeneous. According to the position information of nodes in the cluster, we used uniform minimum mean square subtraction declustering to filter out the cluster head. Based on the new cluster energy consumption information, we determined the energy threshold updating cluster head. By reducing the energy consumption of the computing nodes in cluster, we balanced the data cluster load in the calculation. According to the equilibrium data cluster load, we adjusted the node computing efficiency. After the cluster head node received a frame of data, the data was calculated and merged. After that, we transmitted the fused data to the next cluster head node, so as to determine whether the energy of cluster head node was lower than the energy threshold. If it is yes, we started the cluster head update mechanism and entered a new round of node calculation efficiency control. Simulation results show that the node computational efficiency and energy control effect of the proposed method are better than that of current method.