Modelling user radio access in dense heterogeneous networks

Abstract

One of the distinctive features of today’s mobile networks is the densification of the access nodes and their heterogeneity, which lead to complex, multi-tier, multi-radio access systems. Unlike previous work, which has focussed on optimal techniques for user assignment and technology selection schemes, in this paper we present a flexible analytical model for the performance evaluation and the efficient design of the above complex systems. Leveraging a Markovian agent formalism, the model captures several essential elements, including the spatial and temporal dynamics of the user traffic demand and the availability of radio resources. Importantly, the model exhibits low complexity and an excellent match with simulation results; furthermore, it is general enough to accommodate various network architecture and radio technologies. Through an innovative mean-field solution, we derive a number of relevant performance metrics and show the ability of our framework to represent the system behaviour in large-scale, real-world scenarios, with time-varying user traffic.

Introduction

One of the essential factors driving the design of mobile networks is the need to satisfy the increasing demand for large and fast data transfers. Examples of Enhanced Mobile Broadband (eMBB) applications include on-line gaming, high-definition video, and emerging applications like augmented and virtual reality. Recent data confirm these trends: according to the Ericsson’s Mobility Report 2019 [1], in 2024 5G networks will carry 35% of the mobile data traffic, and the latter is predicted to reach 131 exabytes per month with a major contribution of the aforementioned applications.

Since most of the traditional bands (namely, frequencies in the 300 MHz-3 GHz range) are already allocated to services and capacity has almost reached Shannon’s limit [2], a relevant solution to the throughput demand by data-hungry applications is to develop heterogeneous, dense communication networks [3]. Such networks are characterized by the presence of large cells and many small cells, with the latter ones being classified as micro-cells, pico-cells, or femtocells [4]. While femtocells are mostly used indoors, micro and macro cells are already essential components of outdoor cellular networks. As a consequence, their densification is pivotal to the provisioning of high data rates, to the support of a large number of simultaneous connections, and, thanks to the shorter distance between users and network Point of Access (PoA), to a reduced power consumption.

Such multi-tier network architecture will necessarily leverage multiple radio access technologies (RAT), like LTE, WiFi, and mmwave, just to name some of the most relevant user-infrastructure communication technologies. Several studies have therefore studied, e.g., LTE access systems with two tiers as well as different RATs in 5G RAN systems [5], including the possibility to support the implementation of virtual RATs [6]. In this scenario, users are, and will ever be, equipped with more and more radio interfaces, thus enabling seamless connections with different PoAs and high-throughput performance anytime and everywhere. Nonetheless, several challenges still need to be addressed, among which, the management of radio resources – a key factor to the effective design of heterogeneous networks.

In this paper, we address the above challenge and develop a framework for the analysis of strategies to effectively handle the connectivity of mobile users. In particular, we consider that the network service area is covered by a number of PoAs, each of them hosting multiple radio access interfaces. A user can connect to any of such interfaces, provided that it is under coverage and enough radio resources are available. To efficiently allocate the radio resources, we propose an analytical model, which leverages mean-field analysis [7], [8]. Unlike other techniques, such as queueing networks, stochastic Petri nets, or process algebras, mean-field analysis permits to account for the spatial distribution of the communication nodes (PoA and users) in the system. This is clearly of fundamental importance, as a user can access a PoA only if its position is within the coverage area of that PoA. Importantly, the model exhibits a low complexity, hence it is able to deal with very large network scenarios. The model is then solved by resorting to a novel method based on the Markovian Agent formalism [9], and exploiting the results in [10].

For sake of concreteness, we focus on a specific, fully-distributed, user-centric strategy for RAT selection and study the system performance in a real-world scenario. It is worth remarking that our analytical framework can be extended to investigate other RAT selection strategies, as well as different heterogeneous, multi-layer scenarios. The main contribution of this work resides in the development of an effective and efficient application model. Furthermore, the paper provides relevant, theoretical insights through a set of equations that, by a mean field-based model, capture the dynamics of a RAT communication network.

The rest of the paper is organized as follows. In Section 2, we discuss some related work on the modelling and analysis of new generation networks. Section 3 first introduces the system scenario and the assumptions we make on both the communication network and the user traffic demand; then it provides an example of user access policy in a multi-technology network. Section 4 describes the model we developed to represent the heterogeneous network system and its model parameters, while Section 5 presents the solution techniques we adopted to solve our model. Section 6 introduces the network topology and the instance of the general model we used for validating our approach, and it shows a comparison between analytical and ns3 simulation results. Section 7 applies the proposed model and solution technique to a real-world scenario, leveraging both user mobility and traffic demand experimental traces. Finally, Section 8 draws some conclusions.