In order to solve the problem of traffic burst due to the increase in access points and user movement in an FTTR network, as well as to meet the demand for a high-performance network, it is necessary to rationally allocate network resources, and accurate traffic prediction is very important for dynamic bandwidth allocation in such a network. Therefore, this paper introduces a novel traffic prediction model, named CPO-BiTCN-BiLSTM-SA, which integrates the Crested Porcupine Optimizer (CPO), bidirectional temporal convolution (BiTCN), and bidirectional long short-term memory (BiLSTM) networks. BiTCN extends the traditional TCN by incorporating bidirectional data information, while BiLSTM enhances the network’s capability to learn from long sequences. Moreover, self-attention (SA) mechanisms are utilized to emphasize the crucial segments in the data. Subsequently, the BiTCN-BiLSTM-SA model is optimized by CPO to obtain the best network hyperparameters, and model training prediction is performed to achieve multi-step predictions based on single-step prediction. To evaluate the model’s generalization ability, two distinct datasets are employed for traffic prediction. Experimental findings demonstrate that the proposed model surpasses existing models in terms of the root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination ($R^2$). In comparison with the traditional XGBoost model, the proposed model has an average reduction of 29.50%, 25.43%, and 25.00% in RMSE, MAE, and MAPE, respectively, with a 6.70% improvement in $R^2$.

In this paper, an improved end-to-end autoencoder based on reinforcement learning by using Decision Tree for optical transceivers is proposed and experimentally demonstrated. Transmitters and receivers are considered as an asymmetrical autoencoder combining a deep neural network and the Adaboost algorithm. Experimental results show that 48 Gb/s with 7% hard-decision forward error correction (HD-FEC) threshold under 65 km standard single mode fiber (SSMF) is achieved with proposed scheme. Moreover, we further experimentally study the Tree depth and the number of Decision Tree, which are the two main factors affecting the bit error rate performance. Experimental research afterwards showed that the effect from the number of Decision Tree as 30 on bit error rate (BER) flattens out under 48 Gb/s for the fiber range from 25 km and 75 km SSMF, and the influence of Tree depth on BER appears to be a gentle point when Tree Depth is 5, which is defined as the optimal depth point for aforementioned fiber range. Compared to the autoencoder based on a Fully-Connected Neural Network, our algorithm uses addition operations instead of multiplication operations, which can reduce computational complexity from 10⁸ to 10⁷ in multiplication and 10⁶ to 10⁸ in addition on the training phase.

Nonlinear effects have been restricting the development of high-speed visible light communication (VLC) systems. Neural network (NN) has become an effective means to alleviate the nonlinearity of a VLC system due to its powerful ability to fit complicated functions. However, the complex training process of traditional NN limits its application in high-speed VLC. Without performance penalty, reservoir computing (RC) simplifies the training process of NN by training only part of the network connection weights, and has become an alternative scheme to NN. For the indoor visible light orthogonal frequency division multiplexing (VLC-OFDM) system, this paper studies the signal recovery effect of the pilot-assisted reservoir computing (PA-RC) frequency domain equalization algorithm. The pilot information is added to the feature engineering of RC to improve the accuracy of channel estimation by traditional least squares (LS) algorithm. The performance of 64 quadrature amplitude modulation (QAM) signal under different transmission rates and peak to peak voltage (Vpp) conditions is demonstrated in the experiments. Compared with the traditional frequency domain equalization algorithms, PA-RC can further expand the Vpp range that meets the 7% hard-decision forward error correction (FEC) limit of 3.8 × 10${}^{-3}$. At the rate of 240 Mbps, the BER of the system is reduced by about 90%, and the utilization rate of the available frequency band of the system reaches 100%. The results show that PA-RC can effectively improve the transmission performance of VLC system well, and has strong generalization ability.

In this paper, a simple sampling frequency offset (SFO) estimation and compensation scheme based on two phase-conjugated pilots is proposed and experimentally demonstrated in an OFDM-based underwater wireless optical communication (UWOC) system. The phase shift is obtained by simple multiplication for phase-conjugated pilots, and the results are averaged to perform more accurate phase estimation. The experimental results show that the estimation offset is limited within ±3 ppm when the SFO ranges from −1000 ppm to +1000 ppm over a 1 m tap water channel. Moreover, with the help of the proposed scheme, up to ±300 ppm SFO can be well-compensated with error vector magnitude (EVM) penalties below 1 dB after 1 m underwater transmission. In addition, the results demonstrate that, compared with the ideal case without SFO, our proposed SFO compensation scheme can provide nearly negligible bit error rate (BER) penalties in saltwater with the highly scattering property.

The development of low-orbit satellite communication networks marks the beginning of a new era in global communication. However, in the context of large-scale LEO satellite communication scenarios, the traditional adjacent connection transmission method limits the advantages of low latency in optical communication. Multi-hop transmission increases the number of hops and propagation distance, thereby affecting time-sensitive business transmissions. Therefore, based on the design of optical interconnect parallel subnetworks, this paper proposes a scheduling strategy for time-sensitive business transmissions between LEO satellites. Firstly, this strategy integrates the gate control scheduling mechanism from Time-Sensitive Networking (TSN) transmission in the interconnect parallel subnetwork scenario. Secondly, considering issues like queuing after subnetwork division, excessive burden, and algorithm complexity, mathematical problem abstraction modeling is applied to subsequent route scheduling, with reinforcement learning used to solve the problem. Through simulation experiments, it has been observed that compared to SPF (Shortest Path First) and ELB (Equal Load Balance), this approach can effectively enhance the control capability of end-to-end latency for TSN services in long-distance transmissions within Low Earth Orbit mega-constellations. The integration of reinforcement learning decision algorithms also reduces the complexity compared to traditional constraint-solving algorithms, ensuring a certain level of practicality. Overall, this solution can enhance the communication efficiency and performance of time-sensitive services between satellite constellations. By integrating time-sensitive network transmission technologies into optically interconnected subnets, further exploration and realization of low-latency and controllable latency satellite communication networks can be pursued.