Approximate Projection Based Global Proportional Fairness Scheduling

Nowadays proportional fairness (PF) scheduling has attracted much attention in various wireless systems. But most previous work just considers the systems with only one base station (or data center), which just achieves local PF. In this paper we consider the problem of achieving global PF for the multiple base station multiple user scenario. Compared with previous works in the literature, the main contributions of this paper are threefold: (1) The PF rule is employed in the multiple base station multiple user case. Here we propose an approximate gradient projection based PF scheduling scheme, GP-PF, to approach the global PF optimality. And the convergence of the proposed algorithm is proved. (2) We study the communication and computation complexity of GP-PF and show that the developed GP-PF algorithm can be implemented either in a user selection mode, or in a random accessing way. And GP-PF applies to distributed systems in the sense that it does not need any inter base station cooperation at all. (3) By simulation, it is shown that global PF leads to higher throughput and greater fairness for users than local PF.