SPC10-2: Iterative Water-filling for Optimal Resource Allocation in OFDM Multiple-Access and Broadcast Channels

A class of optimal resource allocation problems in linear Gaussian multiple-access and broadcast channels (MAC and BC) can be summarized as weighted sum power minimization problem. In this paper an iterative water-filling algorithm is proposed to solve this problem efficiently. It is shown that by formulating an explicit rate expression for MAC, though non-convex of power spectral densities, the optimality conditions demonstrate a strong water-filling flavor. By iteratively solving the optimality conditions, whereas in each iteration a slightly modified single-user margin adaptive water-filling(MAWF) algorithm is applied to update the dual variable in a greedy manner, the power spectral density of each user converges to the optimal solution very fast. Simulations verify convergence and optimality. The problem in BC can be solved in its dual MAC.