Computation capacity constrained joint transmission design for C-RANs

This paper considers the joint processing design for the cloud radio access network (C-RAN) with limited cloud computation capacity. This amounts to determine the set of remote radio heads (RRHs) serving each user and the corresponding precoding vectors whose corresponding computation effort (CE) is a non-linear function of the number of antennas pooled from all serving RRHs and the modulation bits. Toward this end, we propose a novel three-step approach to solve the underlying mixed non-linear integer program. First, we transform this problem into a group association problem (GAP) with additional association constraints where each user must be associated with exactly one particular group of RRHs. Second, we study the relaxed power minimization problem (PMP) where the group association integer variables are relaxed and the computational constraint functions are approximated by weighted linear functions of transmission powers. We prove that this relaxed PMP can be solved optimally and the obtained optimal solution satisfies all association constraints of the original GAP problem. Third, we develop an iterative procedure to update the weight parameters of the approximated computational constraint functions to drive the achieved solution to an efficient and feasible solution of the original problem. Finally, we present numerical results to demonstrate the significant gains of our proposed design compared to that due to a fast greedy algorithm.