Deep Learning-Based Nonstationary Channel Prediction in Tactical Vehicle-to-Vehicle Communication Environments
Algorithm 1
The long short-term memory-based vehicle-to-vehicle channel prediction algorithm for tactical communication environments.
input: Complex CSI sequence, the set of hyperparameters .
1: Training stage:
2: Generate the input data of the training set by dividing the complex channel sequence into two subsequences according to the real and imaginary component.
3: for epoch =1 : do
4: Explore the temporal non-stationary characteristics of the V2V channel using the LSTM-based neural networks model.
5: Obtain the predicted CSI via the model.
6: The hyperparameters are optimized by minimizing the loss function in Equation (6).
7: end
8: Prediction stage:
9: Generate the test input data by converting the complex data to real domain.
10: Predict the target CSI via inputting the historical data into the trained LSTM-based predictor.
