Flexible 3GPP MIMO Channel Modeling and Calibration with Spatial Consistency

As channel models are essential to the performance evaluation of wireless systems, the 3rd Generation Partnership Project (3GPP) developed different channel models to verify the effectiveness of emerging and cutting-edge technologies. Since the channel models have become more sophisticated, system-level channel simulations involve high computational complexity and memory storage requirements to generate dedicated spatially consistent random variables (SCRVs). In essence, the channel characteristics are continuous functions of the geographic location in three-dimensional (3D) space. However, it is still an open problem to generate SCRVs for both six-dimensional (6D) space (i.e., a fabric of two independent 3D spaces) and arbitrary spatial autocorrelation functions. This paper develops a two-layer perceptron network to implement a sum-of-sinusoids model that allows a smooth transition from a 3D model to a 6D model while establishing arbitrary spatial autocorrelations. Then, we elaborate on the methodology of system-level channel modeling and implement it with the developed method efficiently. Compared with numerous reference results reported in 3GPP by major industrial telecom companies, extensive system-level calibration results regarding the 3GPP channel models corroborate the effectiveness of the proposed method. Moreover, we offer new performance benchmarks regarding the state-of-the-art New Radio channel model specified in 3GPP TR 38.901 for future calibration purposes.