Penalized iterative waterfilling algorithm for multi-cell and multi-user OFDMA systems

We propose a penalized iterative waterfilling to improve the performance of a distributed resource allocation method in terms of quality of service (QoS) requirements in multi-cell and multi-user OFDMA systems. In the distributed resource allocation method, since each selfish user tries to maximize its own throughput (i.e., sum rate) without any information on other users, it is less likely to satisfy QoS requirements from the system point of view. To overcome this problem, we introduce a self-status prediction period (SPP) and user-based penalty function (PF). Each user can measure its own status during SPP. Using this measurement, each user computes its own PF and then reallocates its resource so that the number of users being supported can increase. Simulation results show that the proposed algorithm can greatly increase the number of users satisfying QoS requirements, also providing tradeoff between total power consumption and achieved sum rate.