An Underwater Node Clock Synchronization Algorithm Based on Spatiotemporal Attention Graphs by Convolutional Neural Networks
Clock synchronization is the core mechanism of underwater wireless sensor networks.Real-time and accurate node movement speed is an important guarantee for constructing high-precision clock synchronization algorithm.To solve the problem of time-varying propagation delay and high energy consumption caused by node movement in synchronization process,this paper proposes a clock syn-chronization algorithm based on the combination of attention mechanism and graph convolutional neural network.Firstly,the Lagrange current model is used to simulate the trajectory of the node,and the ve-locity of the node is roughly estimated from the ocean current model.Node velocity,underwater envi-ronment information set and time ratio are fused to be used as network input features.Secondly,the spatial and temporal attention weight matrix is constructed using the attention mechanism combined with the input features,and the weight matrix is adjusted adaptively according to the feature data.The reunion graph convolutional neural network captures the spatial characteristics between node velocities and displacements.On this basis,the stack standard convolutional layer further merges the node infor-mation of adjacent time to obtain the time correlation,and then constructs the node movement model to effectively predict the node movement speed in real time,and finally quickly calculates the node dy-namic propagation delay to complete the clock synchronization.Experimental results show that the pro-posed algorithm improves the accuracy of TSHL algorithm,D-sync algorithm and K-sync algorithm by 26％,20％and 11％respectively.In terms of energy consumption,the proposed algorithm is also bet-ter than the existing clock synchronization algorithm.
Clock synchronizationocean current modelattention mechanismgraph convolutional neural network
NETL
NSTL
万方数据
