Energy Efficient Joint Resource Allocation for Multi-cell C-RAN System

In this paper, we propose a joint resource allocation algorithm for energy efficiency (EE) of a cloud radio access network (C-RAN). To tackle the joint optimization problem, we utilize a successive convex approximation (SCA) based joint antenna selection and power allocation (JASPA) algorithm. In addition, we propose an improved multi-candidate remote radio heads (RRHs) user association (MCRUA) algorithm. The simulation results reveal that for a C-RAN, we can achieve a significant gain in the EE of the proposed joint resource allocation. INTRODUCTION Cloud radio access network (C-RAN) has been considered for fifthgeneration (5G) mobile networks due to high spectral efficiency and energy efficiency (EE) [1]. In C-RAN, the baseband unit (BBU) is located in the center of a cell, and is connected to remote radio heads (RRHs). The basic idea of the C-RAN is that the transmission distance between antennas and mobile users (MUs) is reduced, so that significant improvement to the EE are achieved. From the viewpoint of green information technology, the EE has recently attracted more attention. Thus, the resource allocation for EE maximization has attracted considerable attention. However, most existing works on resource allocation for the EE of C-RAN consider the separated optimization problems. In this paper, we maximize the downlink EE in C-RAN by solving a joint resource allocation optimization problem. To tackle the complex joint optimization problem, we utilize a successive convex approximation (SCA) based joint antenna selection and power allocation (JASPA) algorithm [2]. Furthermore, we propose an improved multi-candidate RRHs user association (MCRUA) algorithm. This algorithm utilizes the tabu search technique [3-4] to obtain the appropriate large-scale fading threshold (LSFT) for EE improvement. Finally, a joint resource allocation algorithm with fast convergence is proposed, which is based on the Dinkelbach method [5] for solving nonlinear fractional problems. SYSTEM MODEL We consider a C-RAN system with L hexagonal cells, where each cell has N RRHs. In each cell, K single antenna MUs are uniformly distributed. We also assume that the n -th RRH in the j -th cell is equipped with jn M antennas and the total number of system antennas in the j -th cell is given by max 1 N jn n M M    . The channel vector between the RRHs in the l -th cell and the k -th MU in the j -th cell is denoted as 1/2 jkl jkl jkl  g Λ h where   1/2 T 1 diag [ , , ] jkl jkl jklN N     Λ I , 1/2 jkln jkln jkln ad s    , and T T T 1 [ , , ] jkl jkl jklN   h h h . Here, jkln  represents the long-term path loss between the k -th MU in the j -th cell and the n th RRH in the l -th cell.  represents the Kronecker product, a is the path loss gain, and  is the path loss exponent. jkln h denotes an 1 ln M  small-scale fading channel vector which contains independent and identically distributed circularly symmetric complex Gaussian random variables with zero mean and unit variance. The minimum mean-square error channel estimation for the channel vector is given by 1 P, ˆ jkl jkl jk jk   g Λ Q y where   max 1 2 P 1 L jk jkl M l       Q Λ I and P, 1 L jk jkl lk l    y g z is max 1 M  received pilot signal. Here, the large-scale fading between the n -th RRH in the l -th cell and the k -th MU in the j -th cell is given by   1/2 2 P 1 L jkln jkln jkln l     