Abstract-Joint subcarrier, power and rate allocation in orthogonal frequency division multiple access (OFDMA) scheduling is investigated for both downlink and uplink wireless transmissions. Using a time-sharing argument, a convex formulation is obtained avoiding the NP-hardness of the usual 0-1 integer program solution. It is rigourously established that the optimal allocation can be obtained almost surely through a greedy waterfilling approach with linear complexity in the number of users and subcarriers. Stochastic approximation is further employed to develop on-line algorithms which are capable of dynamically learning the underlying channel distribution and asymptotically converges to the off-line optimal solution from arbitrary initial values.
Xin Wang, Georgios B. Giannakis, Yingqun Yu