Trajectory-Unaware Channel Gain Forecast in a Distributed Massive MIMO System Based on a Multivariate BiLSTM Model

Abstract

Cell-free massive MIMO networks have recently emerged as an attractive solution capable of solving the performance degradation at the cell edge of cellular networks. For scalability reasons, usercentric clusters were recently proposed to serve users via a subset of APs. In the case of dynamic mobile scenarios, this network organization requires predictive algorithms for forecasting propagation parameters to maintain performance by proactively allocating new APs to a user. However, a major scientific challenge is the accuracy of predicting the channel gain evolution in non-stationary channels with low computational complexity, considering the uncertainty caused by user mobility. The novelty of this paper is the design of a multidimensional BiLSTM-based multivariate channel gain forecasting algorithm achieving a similar accuracy to previous research at reduced computational complexity. Indeed, thanks to the combination of dual prediction by the multidimensional BiLSTM exploiting the channel diversity from multiple antennas, our model mitigates the error propagation typically faced by sequential neural networks. Our model has a lower error by at least a factor of 2.7 and lower complexity by a factor of 3.6 (for a single prediction), compared to hybrid CNN-LSTM model. Also, in contrast to parallel transformer solutions, the growth rate of the complexity of our algorithm is significantly lower.

Keywords

• Channel prediction
• channel gain forecast
• mobile networks
• massive MIMO
• machine learning
• bidirectional LSTM