On optimal minimum-rate discrete loading in OFDMA systems

This paper considers the discrete bit loading problem in OFDMA downlink systems where each user has a minimum-rate constaint. Whereas OFDMA loading problems are generally non-convex and combinatorial, their optimal solutions can be approximated with a dual decomposition. The dual approach in the per-user minimum-rate case has a K-dimensional Lagrange search space for K users, and as a result is a very difficult problem. We consider two cases of the minimum-rate problem: instantaneous and ergodic allocation, and compare approaches and complexity in obtaining their solutions. We propose a subgradient ergodic allocation approach that is both practical and yields near optimal solutions.