[Paper Review] User Performance in Small Cells Networks with Inter-Cell Mobility

Philippe Olivier, Alain Simonian|arXiv (Cornell University)|May 26, 2016

Advanced Wireless Network Optimization1 references3 citations

TL;DR

This paper proposes an analytical framework to evaluate user performance in dense small cell networks with inter-cell mobility, modeling mobility via the residual sojourn time distribution and using a Quasi-Stationary approximation to reduce computational complexity. The key finding is that both mobile and static users benefit from a throughput gain due to opportunistic handovers, despite increased handover signaling.

ABSTRACT

We analyze the impact of intra-cell mobility on user performance in dense networks such as that enabled by LTE-A and 5G. To this end, we consider a homogeneous network of small cells and first show how to reduce the evaluation of user performance to the case of a single representative cell. We then propose simple analytical models that capture mobility through the distribution of the residual sojourn time of mobile users in the cell. An approximate model, based on Quasi-Stationary (QS) assumptions, is developed in order to speed up computation in the Markovian framework. We use these models to derive the average throughput of both mobile and static users, along with the probability of handover for mobile users. Numerical evaluation and simulation results are provided to assess the accuracy of the proposed models. We show, in particular, that both classes of users benefit from a throughput gain induced by the "opportunistic" displacement of mobile users among cells.

Motivation & Objective

• To evaluate the impact of inter-cell mobility on user performance in dense LTE-A and 5G small cell networks.
• To decouple the performance evaluation from detailed geometric user mobility models by focusing on residual sojourn time distributions.
• To derive analytical models for average throughput and handover probability in a homogeneous network of small cells.
• To develop a computationally efficient Quasi-Stationary approximation to overcome scalability issues in the exact Markovian model.
• To quantify the trade-off between throughput gains and handover signaling overhead due to mobility.

Proposed method

• Reduces network performance evaluation in a homogeneous small cell network to a single representative cell using symmetry and flow-level modeling.
• Models user mobility through the distribution of residual sojourn time in a cell, treating handover arrivals as Poisson processes.
• Constructs a Markovian flow-level queuing model to compute mean throughput and handover probability for static and mobile users.
• Introduces a Quasi-Stationary (QS) approximation to simplify the Markov model and improve computational scalability.
• Superposes fresh traffic and handover traffic using Poisson process assumptions, assuming independence across flows.
• Validates models via numerical evaluation and simulations, comparing exact Markovian results with QS approximation.

Experimental results

Research questions

• RQ1 How does inter-cell mobility affect the average throughput of mobile and static users in dense small cell networks?
• RQ2 What is the impact of mobility on the handover rate, and how does it trade off against throughput gains?
• RQ3 How accurate is the Quasi-Stationary approximation in estimating performance metrics compared to the exact Markovian model?
• RQ4 To what extent is the performance robust to different distributions of user residual sojourn times?
• RQ5 Can the performance of a large homogeneous network be accurately modeled by a single representative cell?

Key findings

• Both mobile and static users experience a throughput gain due to opportunistic channel conditions encountered during inter-cell mobility.
• The Quasi-Stationary approximation provides accurate estimates of mobile user throughput and handover probability, though with reduced accuracy for static user throughput.
• Handover probability increases with user speed, approaching 1 as speed tends to infinity, consistent with theoretical expectations.
• For a Pico cell (50 m radius) and 50% mobile users, throughput increases with speed, with mobile users achieving higher gains than static users.
• The model shows robustness to different residual sojourn time distributions, justifying the use of Markovian assumptions.
• The reduction to a single representative cell is valid and effective for homogeneous networks, enabling scalable performance evaluation.