On user association and multiple access optimisation in 5G massive MIMO empowered ultra dense networks,

Abstract

Ultra network densification and massive multiple-input and multiple-output are considered major 5th generation technology enablers, promising huge capacity gains as they offer proximity, spectral and spatial reuse benefits. Towards succeeding in harnessing the anticipated gains, a systematic treatment of radio resources management challenges encountered in such network deployments is of utmost importance. In this respect, we introduce and study in detail the user-association and multiple-access optimisation aspects of massive multiple-input and multiple-output ultra dense networks. We derive for the first time to the best of our knowledge a series of mathematical programming models for computing the optimal user-association decisions and assessing the impact of two multiple access candidate approaches, namely, frequency domain and spatial domain multiple access. We support the work with an extensive performance evaluation study, which reveals the rate scaling performance trends of the different approaches, considering various network operation conditions, as well as obtain useful design insights for future setups employing the particular technologies.

Near‐optimal user assignment in LTE mobile networks with evolutionary computing

“…The review suggests that, to the best of our knowledge, there is no work tackling the user‐cell problem in LTE (with the goal of minimizing the DTS) driven by an EA with the LTE‐domain specific initial population that we have introduced in Section 1. We have also shown that there are some similar (although different) approaches within the general context of wireless communications. While our proposal aims at finding a near‐optimal user‐cell association LTE by minimizing the DTS metric using an EA, focuses on finding a near‐optimal association of users to access points in WiFi networks by minimizing all the user's satisfaction ratio (the reciprocal of the load) using a greedy algorithm.…”

“…It exploits the aforementioned fact that, when the network operates in the massive MIMO regime, the achievable rates do not depend on the other associations, which simplifies the problem at great extent. This key formulation allows for reaching the exact global optimum to the user‐cell association problem with reasonable complexity, using integer linear programming tools . The obtained exact model, although not feasible for on‐line implementation (because of its combinatorial nature and centralized approach), is a very useful tool for the design/dimensioning phase (the stage our proposal also focuses on).…”

“…However, as shown in the state‐of‐the‐art section review in Section 2, there are some very recent, key studies that propose mathematical methods able to overcome this limit by applying mathematical methods that transform the original intractable problem into another that can be solved exactly with lower computational complexity (although still high). Although there are some works in a similar line of research (see the LB and convex optimization part in Section 2) we highlight here, which exactly transform the original intractable problem into a more feasible and tractable formulation by applying linearization‐reformulation methods. These very recent and novel formulations allow for reaching the exact global optimum solution to the user‐cell association problem with reasonable complexity, using integer linear programming tools .…”

“…Other metrics to study LTE performance, essentially based on maximizing the overall system throughput or on delay minimizing, are basically focused on system operation , while our DTS metric addresses the dimensioning phase . We formulate the LTE user assignment problem as a combinatorial problem, well suited for being approximately solved by an EA. The optimal user association combinatorial problem is computationally intractable in this class of conventional macrocell networks (involving several hundred of users) and even in emerging heterogeneous networks (HetNets),34 and ultra‐dense networks (UDN) (in which user density and node density can become similar36). However there are some very recent key studies that propose mathematical methods able to overcome this intractability by applying mathematical methods that transform the original intractable problem in another that can be solved exactly with lower computational complexity (although still high) ,24,…”

“…The optimal user association combinatorial problem is computationally intractable in this class of conventional macrocell networks (involving several hundred of users) and even in emerging heterogeneous networks (HetNets),34 and ultra‐dense networks (UDN) (in which user density and node density can become similar36). However there are some very recent key studies that propose mathematical methods able to overcome this intractability by applying mathematical methods that transform the original intractable problem in another that can be solved exactly with lower computational complexity (although still high) ,24, We tackle this highly complex problem by proposing a particular implementation of EA that always maintains the best solution in the evolving population.…”