Research on Satellite Network Traffic Prediction Based on Improved GRU Neural Network

Liu, Zhiguo; Li, Weijie; Feng, Jianxin; Zhang, Jiaojiao

doi:10.3390/s22228678

Cited by 24 publications

(10 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…(3) GRU (Liu et al [ 40 ]): This study proposed a method for predicting satellite network traffic based on gated recurrent units. The method can fully exploit the time dependent features of the input sequences by a GRU neural network and mine the similar correlations between data sequences.…”

Section: Methodsmentioning

confidence: 99%

Milling Surface Roughness Prediction Based on Physics-Informed Machine Learning

Zeng

2023

Sensors

View full text Add to dashboard Cite

Surface roughness is a key indicator of the quality of mechanical products, which can precisely portray the fatigue strength, wear resistance, surface hardness and other properties of the products. The convergence of current machine-learning-based surface roughness prediction methods to local minima may lead to poor model generalization or results that violate existing physical laws. Therefore, this paper combined physical knowledge with deep learning to propose a physics-informed deep learning method (PIDL) for milling surface roughness predictions under the constraints of physical laws. This method introduced physical knowledge in the input phase and training phase of deep learning. Data augmentation was performed on the limited experimental data by constructing surface roughness mechanism models with tolerable accuracy prior to training. In the training, a physically guided loss function was constructed to guide the training process of the model with physical knowledge. Considering the excellent feature extraction capability of convolutional neural networks (CNNs) and gated recurrent units (GRUs) in the spatial and temporal scales, a CNN–GRU model was adopted as the main model for milling surface roughness predictions. Meanwhile, a bi-directional gated recurrent unit and a multi-headed self-attentive mechanism were introduced to enhance data correlation. In this paper, surface roughness prediction experiments were conducted on the open-source datasets S45C and GAMHE 5.0. In comparison with the results of state-of-the-art methods, the proposed model has the highest prediction accuracy on both datasets, and the mean absolute percentage error on the test set was reduced by 3.029% on average compared to the best comparison method. Physical-model-guided machine learning prediction methods may be a future pathway for machine learning evolution.

show abstract

Section: Methodsmentioning

confidence: 99%

Milling Surface Roughness Prediction Based on Physics-Informed Machine Learning

Zeng

2023

Sensors

View full text Add to dashboard Cite

show abstract

“…The literature [5] captures the spatiotemporal correlation of satellite network traffic to predict the future network state and designs routing strategies based on the load levels of neighboring satellites. To improve the accuracy of traffic prediction, the literature [23] improved the Gated Recurrent Unit (GRU) neural network to more effectively analyze the self-similarity and long correlation between data series by introducing an attention mechanism, to improve the prediction accuracy. In addition, the Particle Swarm Optimization (PSO) algorithm is introduced to automatically obtain the best neural network model hyperparameters, thus improving the prediction efficiency.…”

Section: Related Researchmentioning

confidence: 99%

“…In the context of the existing limitations in satellite network traffic prediction methods, which fail to fully exploit the long-range correlations among data, resulting in significant forecasting errors and low prediction accuracy, the literature [23] proposes an innovative satellite network traffic prediction strategy based on an improved Gated Recurrent Unit (GRU) neural network. This forecasting method cleverly integrates attention mechanisms with the GRU neural network to more comprehensively capture the relationships and hidden states among traffic data sequences, thereby significantly improving prediction accuracy.…”

Section: Forwarding Control Mechanismmentioning

confidence: 99%

“…Based on the current traffic value Curr_tra f f ic i , each satellite node utilizes the traffic prediction model from the literature [23] to forecast the traffic value Next_tra f f ic i for the next time step. In the G-AODV protocol, individual satellite nodes broadcast HELLO packets to neighboring nodes to probe the network traffic load in their vicinity.…”

Section: Forwarding Control Mechanismmentioning

confidence: 99%

See 1 more Smart Citation

Traffic-Predictive Routing Strategy for Satellite Networks

Liu,

Wang

et al. 2023

Electronics

Self Cite

View full text Add to dashboard Cite

To address the issues of uneven satellite network load and unstable link connections, a globally adaptive satellite network routing strategy based on traffic prediction (G-AODV) is proposed by enhancing the existing Ad hoc On-Demand Distance Vector Routing (AODV) protocol. To prevent heavily loaded nodes from becoming intermediate nodes, G-AODV introduces a traffic prediction mechanism in the route discovery phase. The routing request packet adopts the corresponding forwarding control policy based on the comparison between the predicted traffic load value at the next moment and the dynamic threshold. At the same time, a path replacement strategy is adopted to replace paths before node congestion occurs to achieve load balancing. Considering the unstable characteristics of the satellite chain that easily breaks, the route maintenance phase selects a path repair method by judging the node stability to avoid secondary breaks in the route. The simulation results show that in scenarios with different packet delivery rates, compared with the three comparative routing strategies, the packet delivery rate of G-AODV is increased by up to 20%, the packet loss rate is reduced by 22%, and the end-to-end delay is significantly reduced. In scenarios with different communication connection pairs, G-AODV’s packet delivery rate increased by up to 20%, the packet loss rate decreased by 18%, and the end-to-end delay was still reduced.

show abstract

“…However, this method has obvious gradient vanishing/explosion problems, which cannot be used for the prediction under long time series conditions. Comparatively, the Long Short Term Memory (LSTM) networks [17][18][19], and Gated Recurrent Unit (GRU) [20,21] can make up for the problem of gradient vanishing/explosion to some extent.…”

Section: Introductionmentioning

confidence: 99%

Network Traffic Prediction Based on the Multi-time Granularity GRU-BP Neural Network

haipeng,

Zhang

2023

Preprint

View full text Add to dashboard Cite

A multi-time granularity GRU-BP neural network is proposed in this study for network traffic prediction. In the method, the network traffic data is fitted firstly by the cubic spline curve, and the fitted data is extracted according to different time granularities. Then, several GRU neural networks corresponding to specific time granularities are used for network traffic pre-prediction. Finally, the pre-prediction results of all the GRU neural networks are fed into the fully connected neural network. The fully connected neural network outputs the final network traffic prediction results. Comparison results show that the proposed method can improve the calculation accuracy by 9.0% and reduce the calculation time by 28.6%.

show abstract

Research on Satellite Network Traffic Prediction Based on Improved GRU Neural Network

Cited by 24 publications

References 21 publications

Milling Surface Roughness Prediction Based on Physics-Informed Machine Learning

Milling Surface Roughness Prediction Based on Physics-Informed Machine Learning

Traffic-Predictive Routing Strategy for Satellite Networks

Network Traffic Prediction Based on the Multi-time Granularity GRU-BP Neural Network

Contact Info

Product

Resources

About