A GAN-GRU Based Space-Time Predictive Channel Model for 6G Wireless Communications

Abstract

The advent of sixth-generation (6 G) wireless communications has posed significant challenges to channel modeling. Channel measurements cannot cover all scenarios and frequency bands for 6 G, and conventional models lack accurate predictive capabilities. To address these issues, this paper proposes a novel 6 G space-time joint predictive channel model to predict channels in the space-time domains, which can rebuild lost measurement data and correct abnormal data. The proposed model designs a space-time generative adversarial network (STGAN) framework, conditioned on channel large-scale and small-scale characteristics, to synthesize sufficient space-time channel datasets, effectively overcoming data shortages. Accompanied by path identification and characteristic classification, the coupled gated recurrent unit (GRU) framework conducts precise predictions for unknown channels in the space-time domains. Comprehensive experiments demonstrate the proposed model's superiority over other methods, including the geometry-based stochastic channel model (GBSM), GRU, long short-term memory (LSTM), and radial basis function neural network (RBF-NN). The model's effectiveness can be attributed to its architecture to capture complex space-time variations and accurately predict non-linear channel characteristics based on continuous measurements. Validation on both indoor and outdoor channel measurements further confirms the model's generality and accuracy. The proposed model provides a robust solution in the space-time joint channel prediction for advanced wireless communications.