Performance-Oriented Association in Large Cellular Networks with Technology Diversity

TL;DR

This paper develops a stochastic geometry framework to analyze and optimize user association policies in multi-technology cellular networks, introducing a simple yet effective Max-Ratio policy that balances performance and estimation costs.

Contribution

It introduces a generic model for analyzing association policies in multi-technology networks and proposes a simple, non-parametric Max-Ratio policy with proven optimality in certain regimes.

Findings

Performance improves with more network information.

Max-Ratio policy is near-optimal without statistical knowledge.

Simulations show trade-offs between performance and estimation costs.

Abstract

The development of mobile virtual network operators, where multiple wireless technologies (e.g. 3G and 4G) or operators with non-overlapping bandwidths are pooled and shared is expected to provide enhanced service with broader coverage, without incurring additional infrastructure cost. However, their emergence poses an unsolved question on how to harness such a technology and bandwidth diversity. This paper addresses one of the simplest questions in this class, namely, the issue of associating each mobile to one of those bandwidths. Intriguingly, this association issue is intrinsically distinct from those in traditional networks. We first propose a generic stochastic geometry model lending itself to analyzing a wide class of association policies exploiting various information on the network topology, e.g. received pilot powers and fading values. This model firstly paves the way for tailoring and designing an optimal association scheme to maximize any performance metric of interest (e.g. the probability of coverage) subject to the information known about the network. In this class of optimal association, we prove a result that the performance improves as the information known about the network increases. Secondly, this model is used to quantify the performance of any arbitrary association policy and not just the optimal association policy. We propose a simple policy called the Max-Ratio which is not-parametric, i.e. it dispenses with the statistical knowledge of base station deployments in stochastic geometry models. We also prove that this simple policy is optimal in a certain limiting regime of the wireless environment. Through simulations, we provide insights into (i) a practical compromise between performance gain and cost of estimating information and; (ii) the selection of association schemes under environments with different propagation models, i.e. path-loss exponents.