In this paper, a general three-dimensional (3D) space-time-frequency non-stationary model is proposed for sixth generation (6G) channels. From the proposed model, a novel method, so-called correlated cluster based birth-death (BD) process method, is developed to efficiently and jointly mimic the 3D channel space-time-frequency non-stationarity. In this developed method, the frequency non-stationarity is properly captured by correlated clusters, which are obtained via an unsupervised learning algorithm in machine learning, i.e., K-Means clustering algorithm. Additionally, the developed method involves the cluster based space-time non-stationary modeling. Based on the correlation coefficient of clusters, the BD probabilities on the array and time axes are reasonably modified by the linear weight method and matrix iteration algorithm. Therefore, interactions among the space, time, and frequency non-stationary modeling are sufficiently considered. Important channel statistical properties are derived and thoroughly investigated. Simulation results demonstrate that the channel non-stationarity in space-time-frequency domains can be sufficiently characterized. Finally, the excellent agreement between the simulation results and measurements further verifies the accuracy of the proposed model.

中文翻译：

---

6G 信道的通用 3D 时空非平稳模型

在本文中，针对第六代 (6G) 信道提出了通用的三维 (3D) 空时频非平稳模型。根据所提出的模型，开发了一种新方法，即所谓的基于相关聚类的生死 (BD) 处理方法，以有效且联合地模拟 3D 通道时空频率非平稳性。在这种开发的方法中，频率的非平稳性被相关的簇正确捕获，这些簇是通过机器学习中的无监督学习算法，即 K-Means 聚类算法获得的。此外，所开发的方法涉及基于集群的时空非平稳建模。基于簇的相关系数，通过线性权重法和矩阵迭代算法合理修改阵列和时间轴上的BD概率。所以，充分考虑了空间、时间和频率非平稳建模之间的相互作用。导出并彻底研究了重要的通道统计特性。仿真结果表明，可以充分表征空时频域中的信道非平稳性。最后，仿真结果和测量结果之间的良好一致性进一步验证了所提出模型的准确性。