Summary:
As wireless networks become nowadays increasingly heterogeneous and necessary,
many
difficulties in their planning and management arise. As a consequence, modern
wireless
systems have difficulties in Quality of Service (QoS), because stable
distribution of resources
fails to consider possible interference. Therefore, it is crucial to optimize
the allocation
of resources so as to maximize the utility of the network, while successfully
managing
interference. In this thesis, this problem is extensively studied so as to
investigte the
process by which it can be solved. Although this problem has been studied in
the past
and other solutions have been proposed, this work is not only an overview of
these older
methods, but also a reassessment of the problem. In our work, we apply new and
modern
optimization techniques based on a generalization of Geometric Programming:
Signomial
Programming. Furthermore, using new techniques of non-linear convex regression,
we
show that the problem can be solved optimally with this analysis. Concluding,
the proposed
method is simple to implement, has an improvement compared to past solutions and
potentially yields a significant contribution to the solution of the original
problem: network
utility maximization.
Keywords:
network utility maximization, wireless networks, non-convex optimization, geometric programming extensions, signomial programming