TIME-SERIES PREDICTION AND PERFORMANCE EVALUATION OF OPTICAL COMMUNICATION CHANNEL STATES BASED ON AN IMPROVED LSTM MODEL
Keywords:
Optical communication, Channel state prediction, CNN-BiLSTM-Attention, Performance evaluationAbstract
Optical communication channel states are highly dynamic and are easily affected by noise, nonlinear impairments, received power fluctuations, atmospheric turbulence, and link instability. Accurate time-series prediction of channel states is therefore important for link quality assessment, performance optimization, and intelligent communication control. To improve prediction accuracy and stability, this study proposes an improved CNN-BiLSTM-Attention model for time-series prediction and performance evaluation of optical communication channel states. In the proposed framework, convolutional neural networks extract local fluctuation features from multidimensional channel state sequences, bidirectional long short-term memory networks capture temporal dependencies, and an attention mechanism emphasizes critical time steps related to future channel degradation. Channel state indicators, including received optical power, OSNR, SNR, Q-factor, BER, and EVM, are constructed as model inputs or evaluation variables. Experimental results show that the proposed model outperforms ARIMA, SVR, GRU, LSTM, and BiLSTM in terms of MAE, RMSE, MAPE, and R2. In addition, the prediction-assisted evaluation results indicate that the proposed method can reduce BER and outage probability while improving Q-factor, demonstrating its potential for intelligent optical link monitoring and adaptive communication control.References
[1] Amirabadi M A, Nezamalhosseini S A, Kahaei M H, et al. A comprehensive survey on machine and deep learning for optical communications. IEEE Access, 2025, 13: 88794-88846. DOI: 10.1109/ACCESS.2025.3569559.
[2] Cao M, Xia J, Wang H, et al. LSTM attention neural-network-based signal detection for hybrid modulated faster-than-Nyquist optical wireless communications. Sensors, 2022, 22(22): 8992. DOI: 10.3390/s22228992.
[3] Hu X, Niu Z, Xiao J, et al. Inter-channel nonlinearity compensation in long-haul optical fiber communication using overlapped-dense LSTM network: complexity reduction and visual understanding. Optics Communications, 2024, 556: 130274. DOI: 10.1016/j.optcom.2024.130274.
[4] Li Y, Zhang H, Li L, et al. Multistep ahead atmospheric optical turbulence forecasting for free-space optical communication using empirical mode decomposition and LSTM-based sequence-to-sequence learning. Frontiers in Physics, 2023, 11: 1070762. DOI: 10.3389/fphy.2023.1070762.
[5] Pu Z, Jiang L, Yan L, et al. Joint modulation format identification and OSNR monitoring based on Stokes vector distribution features for digital coherent optical receivers. Chinese Optics Letters, 2024, 22(5): 050602. DOI: 10.1364/COL.22.050602.
[6] Vuong Q P, Nguyen V D, Ho D T L, et al. An optimized LSTM-based equalizer for 100 Gigabit/s-class short-range fiber-optic communications. AIMS Electronics and Electrical Engineering, 2024, 8(4): 404-419. DOI: 10.3934/electreng.2024019.
[7] Wang H, Zhou X, Chen W, et al. Laser chaotic synchronization communication based on LSTM. Acta Photonica Sinica, 2023, 52(6): 0606003. DOI: 10.3788/gzxb20235206.0606003.
[8] Wang Y, Leeson M, Liu Z, et al. Machine learning-based models for optical fiber channels. Optics Communications, 2025, 591: 132099. DOI: 10.1016/j.optcom.2025.132099.
[9] Zhang L. Channel equalization in ultraviolet communication based on LSTM-DNN hybrid model. Scientific Reports, 2025, 15: 17226. DOI: 10.1038/s41598-025-02159-9.
[10] Zhou L, Cao M, Zhang X, et al. CNN-LSTM-assisted joint channel estimation and signal detection for optical wireless communications. Applied Optics, 2025, 64(35): 10640-10648. DOI: 10.1364/AO.574868.