Power Allocation for Iterative Multiuser Decoding Based on Large System Analysis

In this paper, we consider a power allocation problem for iterative multiuser decoding in a coded code-division multiple access (CDMA) system. Subject to given bit error rate performance targets for all users, the power allocation problem is formulated based on previously proposed large system analysis results for iterative multiuser decoding. The resulting power allocation problem is in general nonconvex and cannot be solved efficiently. However, under certain assumptions the problem can be reformulated into a quasiconvex optimization problem, which in turn can be solved efficiently by a sequence of convex feasibility problems. The approach is applied to a complexity reduced belief propagation multiuser detector as well as to popular interference cancellation based detectors. Numerical results are presented to demonstrate the performance of the proposed power allocation algorithm in finite-size systems. It is found that the belief propagation detector is more power and complexity efficient than the interference cancellation based detectors.