Comparison of Automatic Tracking and Clustering Algorithms for Time-Variant Multipath Components

In mobile communications, wireless channel has been widely considered to be time-variant. Recently, the automatic clustering algorithms for dynamic multipath components (MPCs) are one of the key techniques for time-variant channel modeling. Meanwhile, in order to observe the time-variant characteristic of the dynamic MPCs, the tracking algorithm for MPCs or clusters is also required. However, how to combine clustering and tracking algorithm for time-variant channel modeling is still not well solved. In this paper, two frameworks of automatic tracking and clustering for time-variant MPCs are investigated and compared: i) The MPCs tracking based clustering algorithm (MTC), which focuses on identifying the moving paths of the time-variant MPCs between different snapshots, where the initial MPCs need to be clustered at first. For MTC, instead of using the minimum distance between each MPC/cluster which has been widely used in the past, the whole distance matrix between successive snapshots is used to accurately identify the moving paths; ii) The cluster-centroids tracking based clustering algorithm (CTC), which is designed to cluster the MPCs in each snapshot individually, and the moving paths of each cluster are identified by tracking the cluster-centroids. By using the F-Measure, the above two algorithms are compared using time-variant channel simulator.