write the summary of the document

Cluster-Based Channel Models

Cluster-based channel modeling has been widely accepted by both academic and industry groups, mainly because: Many high resolution algorithms are used for MPC extraction so that clustered MPCs can be more frequently observed. Multiple-input-multiple-output (MIMO) and massive MIMO in 4G and 5G systems require a double-directionally characterized channel, which again helps to see more clearly the clustering structure of the MPCs. Clustered channel modeling has been demonstrated to reduce complexity while maintaining accuracy. From the Saleh-Valenzuela model for single-input-single-output (SISO) channels to MIMO channel models such as COST 259, COST 2100, etc. [1], cluster-based channel modeling has been widely adopted. In general, the following four concepts of clusters are used in channel models. Delay-Domain Cluster: Mainly observed in the delay domain (e.g., from SISO CIRs), especially when the measurement bandwidth is large. The power delay profile (PDP) of a single delay-domain cluster is often modeled as a one-sided exponential function, as shown in Fig. 3a, and different clusters have different initial delays, and may have the same or different decay time constants. A typical application of delay-domain clustering is the Saleh-Valenzuela model. Angle-Domain (or Angle-Delay-Domain) Cluster: When angles of MPCs, e.g., azimuth of arrival (AOA) and azimuth of departure (AOD), are extracted from directional (or MIMO) channel sounding, the discrete MPCs are usually found to be clustered in the angular domain or the joint angle-delay domain, as shown in Fig. 3b. To characterize such clusters, angles of MPCs need to be modeled in terms of angular spread and intra-cluster angular distribution. This kind of cluster is widely used in double-directional channel models, such as COST 259. Twin Clusters: Widely used in geometry-based stochastic channel models when characterizing the impact of multiple interactions of an MPC with the environment. As shown in Fig. 3c, the twin-clusters are two coupled clusters of equivalent interacting objects and are used to represent multiply reflected or diffracted MPCs. The location of the twin-clusters can be chosen independently, allowing an independent adjustment of angles at the transmitter and receiver. The concept of the twin-cluster has been adopted as a fundamental modeling approach for the COST 273 channel model. Common Cluster: Used to simulate the interlink correlation properties of multi-link scenarios. The main idea is to control the correlation between different links by allowing a certain proportion of the energy in different links to propagate through the same clusters. As shown in Fig. 3d, if two propagation links are influenced by the same cluster (i.e., the common cluster), the corresponding two links are correlated. The common cluster is used in the COST 2100 channel model for multi-link simulations. It is noteworthy that the above concepts of clusters overlap with each other, e.g., the twin clusters or common cluster may exist in terms of angle-delay-domain clusters, or they can be described in a geometrical (x-y-z) plane. Note that clustering can also be conducted in the geometric dimension of a propagation environment, especially when it is based on ray tracing. However, due to space limitations, the remainder of the article will focus on clustering in the delay/angle domain.