Multipath Clustering and Cluster Tracking for Geometry-Based Stochastic Channel Modeling

This paper presents a clustering and tracking method that exploits the geometry of the scattering points (SPs) obtained from the measurement-based ray tracer. The multipath components (MPCs) were categorized into clusters by applying the KPowerMeans (KPM) framework to those SPs. The clusters were tracked by comparing the cluster-centroid SPs of the adjacent snapshots. The clusters were estimated based on the indoor environment geometry at 11 GHz, and their physical mechanisms were interpreted. The complexity and performance of this method was assessed and compared with that of conventional KPM by comparing the number of floating point operations (FLOPS) and the channel eigenvalues obtained from the reconstructed channel matrices, which were calculated by superposing the MPCs randomly generated from intracluster parameters. The verification of this method showed that most clusters were estimated and tracked according to the physical location of the scatterers in the environment with acceptable error. Moreover, the eigenvalues reconstructed from the proposed method were closer to the measured ones with less number of FLOPS, which indicates both accuracy and complexity improvement. The results also imply that multiple-input multiple-output performance is highly dependent on the radio propagation channel; therefore, it is imperative that clusters in the channel be determined accurately.