Multi-User Efficient Ergodic Loading for OFDMA Systems

We consider the multi-user discrete bit loading problem in OFDMA systems with objective and constraints in terms of ergodic rates and powers. This differs from most existing techniques, which allocate rate and power according to instantaneous constraints and channel conditions. The maximum weighted ergodic sum-rate optimization problem is non-convex in general, but a dual decomposition results in an convex optimization problem where the duality gap vanishes with an increasingly large number of subchannels. We propose an approach tailored for the discrete case based on our previous work that is a flexible, near optimal, and results in an efficient algorithm. Simulation results showing convergence behavior are presented.