确定性网络5G-A终端时延预测

Deterministic Network 5G-A Terminal Time Delay Prediction

工业控制场景下5G-A终端传输时延是确定性网络能力的直接表征之一,时延预测对提高网络确定性至关重要。由于传输时延序列的不稳定性和随机性,单一模型难以准确预测。针对该问题,提出一种基于优化变分模态分解(Variational Mode Decomposition, VMD)和卷积注意力长短时记忆网络(Convolutional Attention Long Short Term Memory Network, CA-LSTM)的传输时延预测方法。为提高VMD的分解性能,利用相关系数检验法确定时延序列分解的模态数,并利用蝗虫优化寻优分解的惩罚因子和保真度系数;设计CA-LSTM网络,借助卷积滤波器以及注意力机制使得网络具备分辨时延特征重要程度的能力;将各模态预测值重建成一维时延值得到预测结果。实验研究结果表明,优化VDM能够将5G终端传输时延序列有效分解,结合CA-LSTM模型相比于经典LSTM在MSE、RMSE和MAE上分别提升了37.1%、21.3%和23.6%。

The transmission delay of 5G-A terminals in industrial control scenarios is one of the direct characterizations of deterministic network capabilities,and delay prediction is crucial to improving network determinism.Due to the instability and randomness of the transmission delay sequence,a single model is difficult to predict accurately.To solve this problem,a new delay prediction method based on optimal Variational Mode Decomposition(VMD)and Convolutional Attention Long Short Term Memory Network(CA-LSTM)is proposed.Firstly,in order to improve the decomposition performance of VMD,the correlation coefficient verification method is used to determine the modal number of the time-delay sequence decomposition,and the grasshopper optimization algorithm is used to determine the penalty factor and fidelity coefficient of the decomposition.Secondly,a CA-LSTM network is proposed,which uses convolution filters and an attention mechanism to make the network have the ability to distinguish the importance of time-delay sequences in a certain period.Finally,each modal prediction value is reconstructed into a one-dimensional delay value to get the prediction results.The experimental results show that the optimized VMD can effectively decompose the 5G terminal transmission delay sequence.Compared with the classical LSTM,the CA-LSTM improves the MSE,RMSE and MAE by 37.1%,21.3%and 23.6%respectively.