Delay-Optimal Biased User Association in Heterogeneous Networks

TL;DR

This paper develops a delay-optimized user association scheme in heterogeneous networks, balancing load and delay by tuning biasing factors, with explicit formulas and tradeoffs between delay and coverage.

Contribution

It introduces a convex optimization framework for delay minimization through bias tuning, providing explicit and closed-form solutions under certain conditions.

Findings

Proper bias tuning significantly reduces queuing delay.

Tradeoff identified between delay performance and SIR coverage.

Explicit formulas enable practical bias optimization.

Abstract

In heterogeneous networks (HetNets), load balancing among different tiers can be effectively achieved by a biased user association scheme with which each user chooses to associate with one base station (BS) based on the biased received power. In contrast to previous studies where a BS always has packets to transmit, we assume in this paper that incoming packets intended for all the associated users form a queue in the BS. In order to find the delay limit of the network to support real-time service, we focus on the delay optimization problem by properly tuning the biasing factor of each tier. By adopting a thinned Poisson point process (PPP) model to characterize the locations of BSs in the busy state, an explicit expression of the average traffic intensity of each tier is obtained. On that basis, an optimization problem is formulated to minimize a lower bound of the network mean queuing delay. By showing that the optimization problem is convex, the optimal biasing factor of each tier can be obtained numerically. When the mean packet arrival rate of each user is small, a closed-form solution is derived. The simulation results demonstrate that the network queuing performance can be significantly improved by properly tuning the biasing factor. It is further shown that the network mean queuing delay might be improved at the cost of a deterioration of the network signal-to-interference ratio (SIR) coverage, which indicates a performance tradeoff between real-time and non-real-time traffic in HetNets.