移动传感网节点位置预测方法研究

Method of node location prediction for mobile sensor network

针对现有稀疏传感网中移动节点位置预测精度较低,提出一种基于深度信念网络的移动未知节点位置预测方法。首先利用深度信念网络强大的特征学习能力,分析不同信号强度向量样本集;其次在深度信念网络的最后一层级联一层支持向量机,将所学习到的信号强度分布特征作为顶层支持向量机的输入,构建距离预测模型;最后预测未知节点与其相邻节点之间的距离,判断其可能位置所在区域,计算得出未知节点的预测位置。仿真实验结果表明,文中所提出的位置预测方法与RBF神经网络位置预测方法相比,预测精度提高了19.3%;与支持向量机预测方法相比,预测精度提高了23%;与改进的MCL相比,预测精度提高了33.4%,且有较强的鲁棒性,适用于稀疏传感网络节点位置预测。

In allusion to the low accuracy of mobile node location prediction in existing sparse sensor networks,a method of mobile unknown nodes location prediction based on deep belief network is proposed.Firstly,the different signal strength vector sample sets are analyzed based on the powerful feature learning ability of deep belief network.The last layer of the deep belief network is cascaded with one layer of support vector machine,and the learned signal strength distribution features are used as the input of the top⁃layer support vector machine to construct distance prediction model.The distance between the unknown node and its neighbors are predicted,the region where it might be located is judged,so as to calculate the predicted position of the unknown nodes.The simulation experiment results show that in comparison with the RBF neural network location prediction method,the support vector machine prediction method and the improved MCL,the prediction accuracy of the location prediction method proposed in this paper improves by 19.3%,23%and 33.4%,respectively,and it has strong robustness,which is suitable for node location prediction in sparse sensor networks.