In cellular networks, users are grouped into different cells and served by different access points (base stations) that provide wireless access to services and applications. In general, the service demand is very heterogeneous, non-uniformly distributed, and dynamic. Consequently, radio access networks create very irregular topologies with more access points where service demand is concentrated. While this dynamism requires networks with the ability to adapt to time-varying conditions, the non-uniformity of the service demand makes the planning, analysis, and optimization difficult. In order to help with these tasks, a framework based on canonical domains and spatial mappings (e.g., conformal mapping) have recently been proposed. The idea is to carry out part of the planning in a canonical (perfectly symmetric) domain that is connected to the physical one (real-scenario) by means of a spatial transformation designed to map the access points consistently with the service demand. This paper continues the research in that direction by introducing additional tools and possibilities to that framework, namely the use of centroidal Voronoi algorithms and non-conformal composite mappings. Moreover, power optimization is also introduced to the framework. The results show the usability and effectiveness of the proposed method and its promising research perspectives.
I. INTRODUCTION
A. Context and motivationRadio access planning and optimization are fundamental tasks in cellular networks. Broadly speaking, planning refers to the tasks of determining the number, location, and configuration of access points to provide wireless access to users (and things) to services and applications, with a certain targeted Quality of Service (QoS). In particular, the problem of finding the number of access points is also referred to as dimensioning [1], and this initial step aims at providing the required capacity for the service demand volume that is expected. However, in practice, both dimensioning and sites positioning are very difficult problems because the service demand is not uniformly distributed and it is quite diverse and dynamic. Nowadays, taking into account the continuous evolution of radio access technologies, and the new concepts and paradigms that are expected for the fifth generation (5G) of cellular networks, the boundary between planning and optimization tasks becomes blurred. Indeed, according to the excellent work presented in [1], planning and optimization are iterative tasks that go hand-in-hand. In this line of thinking, the authors of [2] also pose the need for re-thinking planning. They emphasize the importance of distributing the service demand as evenly as possible among cells as a key criterion to achieve effective planning; a goal that in the opinion of the authors of [2] (and in our's) is a very valid way to enhance system performance.In our previous work [3], also motivated by the aforementioned ideas, a novel framework for planning and optimization based on the use of canonical domains and spatial transforma...