Deep Learning Aided Spectrum Prediction for Satellite Communication Systems

Ding, Xiaojin; Feng, Lijie; Zou, Yulong; Zhang, Gengxin

doi:10.1109/tvt.2020.3043837

Cited by 24 publications

(11 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Baselines and Evaluation Metrics: For comparison, 10 prediction models based on deep learning are considered as baselines, which are LSTM [ 18 ], GRU [ 32 ], BiLSTM [ 33 ], CNN [ 34 ], CNN-LSTM [ 35 ], CNN-BiLSTM [ 25 ], CNN-BiLSTM-attention [ 36 ], seq2seq-LSTM [ 37 ], seq2seq-LSTM-attention [ 38 ], and seq2seq-Bi-ConvLSTM [ 39 ]. In order to fairly compare the performance of models, the input information and output information of all prediction models are all the same and the baseline models use the multi-step principle to output results.…”

Section: Methodsmentioning

confidence: 99%

“…The reason for choosing the above two models for comparison is as follows. According to the results of the literature [ 25 ], CNN-BiLSTM-attention achieves better performance for time–frequency spectrum prediction than LSTM, BiLSTM, GRU, and CNN-LSTM. In addition, seq2seq-LSTM-attention has good performance on long-term time series prediction.…”

Section: Methodsmentioning

confidence: 99%

“…In [ 24 ], a scheme was formulated with CNN-LSTM for multi-dimensional spectrum prediction. Xiaojin Ding et al [ 25 ] combined CNN and bidirectional LSTM (BiLSTM) for predicting the state of multi-channels with higher accuracy than LSTM, BiLSTM, and CNN-LSTM. Graph convolution network (GCN) has a stronger ability to extract spatial features than CNN.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

A Deep Long-Term Joint Temporal–Spectral Network for Spectrum Prediction

Wang,

Hu,

Jiang

et al. 2024

Sensors

View full text Add to dashboard Cite

Spectrum prediction is a promising technique to release spectrum resources and plays an essential role in cognitive radio networks and spectrum situation generating. Traditional algorithms normally focus on one-dimensional or predict spectrum values in a slot-by-slot manner and thus cannot fully perceive the spectrum states in complex environments and lack timeliness. In this paper, a deep learning-based prediction method with a simple structure is developed for temporal–spectral and multi-slot spectrum prediction simultaneously. Specifically, we first analyze and construct spectrum data suitable for the model to simultaneously achieve long-term and multi-dimensional spectrum prediction. Then, a hierarchical spectrum prediction system is developed that takes advantage of the advanced Bi-ConvLSTM and the seq2seq framework. The Bi-ConvLSTM captures time–frequency characteristics of spectrum data, and the seq2seq framework is used for long-term spectrum prediction. Furthermore, the attention mechanism is used to address the limitations of the seq2seq framework that compresses all inputs into fixed-length vectors, resulting in information loss. Finally, the experimental results have shown that the proposed model has a significant advantage over the benchmark schemes. Especially, the proposed spectrum prediction model achieves 6.15%, 0.7749, 1.0978, and 0.9628 in MAPE, MAE, RMSE, and R2, respectively, which is better than all the baseline deep learning models.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A Deep Long-Term Joint Temporal–Spectral Network for Spectrum Prediction

Wang,

Hu,

Jiang

et al. 2024

Sensors

View full text Add to dashboard Cite

show abstract

“…Tumuluru et al [35] designed a spectrum prediction model based on neural network, which can accurately identify spectrum holes in cognitive networks. Ding et al [36] first preprocessed historical spectrum occupancy data, and then designed a deep learning-based fusion network to predict spectrum occupancy. This network effectively combines multiple prediction results to improve prediction accuracy.…”

Section: Related Workmentioning

confidence: 99%

Securing Radio Resources Allocation With Deep Reinforcement Learning for IoE Services in Next-Generation Wireless Networks

Peng

Xue

Zhu

et al. 2022

IEEE Trans. Netw. Sci. Eng.

View full text Add to dashboard Cite

The next generation wireless network (NGWN) is undergoing an unprecedented revolution, in which trillions of machines, people, and objects are interconnected to realize the Internet of Everything (IoE). with the emergence of IoE services such as virtual reality, augmented reality, and industrial 5G, the scarcity of radio resources becomes more serious. Moreover, there are hidden dangers of untrusted terminals accessing the system and illegally manipulating interconnected devices. To tackle these challenges, this paper proposes a securing radio resources allocation scheme with Deep Reinforcement Learning for IoE services in NGWN. First, the solution uses a BP neural network based on multi-feature optimized Firefly Algorithm (FA) for spectrum prediction, thereby improving the prediction accuracy and avoiding interference between unauthorized and authorized users with efficient radio utilization. Then, a spectrum sensing method based on deep reinforcement learning is proposed to identify the untrusted users in system while fusing the sensing results, to enhance the security of the cooperative process and the detection accuracy of spectrum holes. Extensive simulation results show that the proposal is superior to the traditional solutions in terms of prediction accuracy, spectrum utilization and energy consumption, and is suitable for deployment in future wireless systems.

show abstract

“…The authors of [147] proposed a joint beam hopping and power control scheme for maximizing the throughput of LEO SCSs, while preserving the signal quality of GEO SCSs. A deep learning aided spectrum prediction method was proposed in [148] for mitigating the inter-system interference. A sophisticated combination of a convolutional neural network and of a carefully dimensioned bespoke memory was harnessed for data mining from the historical spectrum usage of the GEO SCSs.…”

mentioning

confidence: 99%