Radio remote head clustering with affinity propagation algorithm in C-RAN

The optimal number of clusters (K) differs depending on the radio remote head (RRH) density. This paper verifies that the K values cannot be met by the conventional affinity propagation (AP) clustering algorithm. In an ultra-dense network (UDN) environment, the density of RRH is a very important factor for the bender because it is directly related to the cost of configuring the wireless communication network. Likewise, in order to provide the optimal communication environment to the user in the UDN environment, it is necessary to enable flexible clustering according to changing channel environment by utilizing semi-dynamic clustering technology. As a result, we propose an AP algorithm that finds a better K value than the conventional method. To this end, the proposed algorithm additionally utilizes a non-coordinated multi-point (CoMP) interference power that varies depending on the RRH density, user position, and the variations in propagation channel. The simulation results show that the proposed algorithm shows a better average capacity than the conventional algorithm.