Distributed optimization in heterogeneous networks with proportional fairness

In this paper we develop a discrete and distributed optimization framework that maximizes the overall spectral efficiency of heterogeneous networks while ensuring a minimum fairness among users. We split the problem into two subproblems: the first one is handled by the users and consists in a transmission by each user of a minimum rate requirement to the base stations and the second is the resource allocation performed by the base stations with respect to the rate constraints generated by the users. The first subproblem is solved using a process where each user tries to adjust its rate requirements to the network capacity. This process is inspired from the TCP Vegas algorithm, where the network usage is improved by reducing congestion. Based on the users' rate requirements, the base stations perform a discrete resource optimization in order to allocate power and rate to each user. The rates are selected among a finite set of discrete rates since Adaptive Modulation and Coding (AMC) technique is assumed to be used in each network. The proposed algorithm is compared to an extension of the Proportional Fair algorithm, developed in this paper to allow joint rate and power allocation. Simulation results show that the proposed algorithm provides better performance than the simple PF while ensuring an acceptable fairness among the users.