Capacity-achieving resource allocation for OFDMA fading channels

This paper considers the jointly optimal rate and power allocation that maximizes a utility function of average rates in OFDMA multiple access and broadcast channels. Derived under the assumption that superposition coding is allowed per subcarrier, our low-complexity (yet capacity-achieving) resource allocation scheme can yield higher achievable rates than the existing approaches based on exclusive subcarrier assignment. In addition, relying on the stochastic optimization tools, we develop a class of stochastic gradient iterations that are capable of converging to the optimal benchmarks without the knowledge of fading channel cdf a priori.