Dynamic RRH Clustering using Affinity Propagation Algorithm in Ultra-Dense C-RAN

Joint transmission across the cluster of radio remote heads (RRHs) by exploiting the centralized processing of the cloud-radio access network (C-RAN) is a promising technique to overcome the severe interference problems in ultra-dense small cell networks. In practical considerations, local clustering of networks is preferred over global clustering as the number of RRHs that can be jointly transmitted on the network is finite. In this study, we attempt to revisit the principles of the well-known affinity propagation (AP) clustering algorithms, especially for the naive method of determining the exemplar with a fixed threshold. To further explain the decision method, we propose a method to easily determine the threshold using the network map of converged value of messages generated by AP algorithm, by combining Otsu's threshold and density peak searching method with the existing AP clustering algorithm. The proposed algorithm provides higher spectral efficiency than the conventional AP algorithms as well as statically coordinated multi-point (CoMP) techniques, although it has similar execution time as the traditional AP algorithms.