HAL API platform

In this dissertation, we study distributed resource allocation techniques in full reuse multicell networks. Throughout this work, we consider a system model in which simultaneous transmissions mutually interfere, and thus it is applicable to a number of wireless access schemes. On the basis of this model, we define the specific resource allocation problem addressed in this work: \textit{joint power allocation and user scheduling} in view of maximizing network capacity, defined as the sum of individual link rates. We initially investigate the behavior of interference in large random wireless networks, where analytical expressions are derived for the average interference as a function of distance between transmitter and receiver in cellular networks. Intuition from this study allows us to propose the interference-ideal network model, which enables us to approximate the instantaneous interference by its average value. This model is applied to the resource allocation problems considered later in the dissertation. We then proceed to study the user scheduling sub-problem in the multicell context under a standard power allocation policy and a resource fairness constraint. We derive the network capacity optimal scheduling policy, based on which a distributed algorithm for the user scheduling problem is proposed. Next, we investigate the optimal power allocation problem considering a weighted sum-rate objective function. Though this is a non-convex optimization problem, for two interfering links we are able to characterize the optimal power allocation solution. Interestingly, when the weights are equal, the optimal power allocation turns the links either on or off, and we term this binary power allocation. Having looked at scheduling and power allocation individually, we proceed to propose algorithms for joint power allocation and scheduling to maximize the sum network capacity. In the first approach, we employ the interference-ideal network model and binary power allocation to derive a distributed iterative algorithm for power allocation and scheduling. The key idea in this approach is to switch off cells which do not contribute enough capacity to outweigh the interference caused to the network. The previous approach relies on a large network assumption, and as such can not be employed for any number of cells. Thus, we propose a framework for distributed optimization of transmit powers based upon partitioning network parameters into local and non-local information. By assuming instantaneous knowledge of local, and statistical knowledge of non-local information, a distributed algorithm is derived which requires no information exchange between links. We also propose an algorithm which uses minimal information message passing (in this case one bit) to further improve the performance gain. User scheduling is shown to be easily incorporated into the power allocation algorithms. In the end, we briefly touch upon an alternative approach called multicell access schemes, inspired by the classical multiple access problem in ad-hoc networks.