Source-linked AI summary
Analysis of the Decoupled Access for Downlink and Uplink in Wireless Heterogeneous Networks
Katerina Smiljkovikj, Petar Popovski, Liljana Gavrilovska
TL;DR
In heterogeneous cellular networks, a device may use different base stations for downlink reception and uplink transmission, unlike the common same-base-station association rule. The paper uses stochastic geometry to analyze association probabilities and average throughput in a two-tier network. As femto-base-station density rises, decoupled access initially becomes more common, and the paper identifies it as a tool for improving uplink-throughput fairness.
Problem
In heterogeneous networks, devices can select among multiple base stations, and the downlink-optimal base station need not also be the uplink-optimal one.
Source: §I · p. 1Method
The paper uses stochastic geometry to jointly analyze downlink and uplink association probabilities in a two-tier macro/femto network, then evaluates average throughput.
Source: §I · p. 1, §II · p. 1, B. Average throughput · p. 4Results
The share choosing decoupled Case 2 rises rapidly with femto-base-station density and exceeds 60% in a region of λF, but tends to zero as λF/λM approaches infinity.
Source: §IV · p. 4Takeaways & Limitations
The paper concludes that decoupled downlink/uplink association can be used as a tool to improve fairness in uplink throughput.
Source: §V · p. 4
Abstract
from arXiv · showhide
Wireless cellular networks evolve towards a heterogeneous infrastructure, featuring multiple types of Base Stations (BSs), such as Femto BSs (FBSs) and Macro BSs (MBSs). A wireless device observes multiple points (BSs) through which it can access the infrastructure and it may choose to receive the downlink (DL) traffic from one BS and send uplink (UL) traffic through another BS. Such a situation is referred to as decoupled DL/UL access. Using the framework of stochastic geometry, we derive the association probability for DL/UL. In order to maximize the average received power, as the relative density of FBSs initially increases, a large fraction of devices chooses decoupled access, i.e. receive from a MBS in DL and transmit through a FBS in UL. We analyze the impact that this type of association has on the average throughput in the system.
I. INTRODUCTION
In heterogeneous networks, devices must choose among multiple access points, and the downlink and uplink need not use the same base station. The paper analyzes decoupled association in a two-tier macro/femto network using stochastic geometry and models for base-station and device locations.
I. INTRODUCTION
- Heterogeneous deployments give devices multiple base stations to choose from, while current practice applies a downlink SINR criterion to both directions even though the downlink and uplink access points can differ.The passages identify differences in signal power and interference, and in base-station congestion, as drivers of decoupled access.Source: §I · p. 1
- The paper jointly analyzes downlink and uplink association in a two-tier network to derive association probabilities and examine throughput and uplink-fairness insights.The analysis uses the same spatial realization for base stations and devices, and its results closely match simulations.Source: §I · p. 1
II. SYSTEM MODEL
- The model represents macro base stations, femto base stations, and devices as independent homogeneous Poisson point processes, with received power modeled using transmission power, fading, and path loss.It considers a typical device at the origin and assumes Rayleigh fading and constant receiver noise power.Source: §II · pp. 1–2
- For uplink interference analysis, translation invariance shifts the uplink access point to the origin while preserving the stationary point-process distribution.This addresses the distance calculations involving interfering devices and an uplink access point that is not initially at the origin.Source: §II · p. 2
- Active-device interferers are approximated by thinning the device process and treating the resulting process as an independent Poisson point process.The model selects a fraction of devices to represent one same-resource interferer per base station; the passage describes the resulting dependence as weak.Source: §II · p. 2
III. ANALYSIS
The analysis first derives downlink/uplink association probabilities and then uses them to evaluate average throughput.
III. ANALYSIS
- The analysis has two linked stages: derive downlink/uplink association probabilities, then evaluate average throughput using them.Source: §III · p. 2
A. Association Probability
Association is based on average received power, with the downlink and uplink criteria reflecting their distinct transmission directions. The two-tier model yields four possible downlink/uplink base-station combinations.
A. Association Probability
- Association uses received power averaged over fading: base-station transmission power matters in downlink, whereas uplink association selects the closest base station.The passage notes that averaging over fading avoids fading-induced ping-pong effects in association.Source: A. Association Probability · p. 2
- The nearest-base-station distance distribution follows from the probability that a Poisson point process has no point within a circle of radius x.For each tier v, the passage gives Pr(Xv > x) = e^(-πλv x^2).Source: A. Association Probability · p. 2
- The two-tier system has four possible downlink/uplink association combinations, illustrated as association regions for PM=46 dBm, PF=20 dBm, and α=4.Source: A. Association Probability · p. 2, Figure 1 · p. 2
1) Case 1: DLAP=ULAP=MBS:
Case 1 is the association to a macro base station in both directions, under the stated condition that femto transmission power is below macro transmission power.
1) Case 1: DLAP=ULAP=MBS:
- With PF < PM, Case 1 is the region where a device selects a macro base station for both downlink and uplink.Source: 1) Case 1: DLAP=ULAP=MBS · p. 2
- The remaining association cases follow the same derivation procedure, for which the paper provides final results only.Source: 1) Case 1: DLAP=ULAP=MBS · p. 2
- Case 2 is decoupled access, with different base stations serving downlink and uplink.Source: 2) Case 2: DLAP=MBS and ULAP=FBS · p. 2
2) Case 2: DLAP=MBS and ULAP=FBS:
Case 2 concerns devices receiving downlink through an MBS and transmitting uplink through an FBS. Its association probability is defined over the domain satisfying both association events.
2) Case 2: DLAP=MBS and ULAP=FBS:
- Case 2 is the joint event whose domain satisfies both association conditions.The passage identifies this domain as Region 2 on Fig. 1 and introduces the Case 2 association probability.Source: 2) Case 2: DLAP=MBS and ULAP=FBS · p. 2
- The Case 2 association probability is presented after identifying Region 2.The supplied passage introduces the probability but does not include its expression.Source: 2) Case 2: DLAP=MBS and ULAP=FBS · p. 2
- The next case is introduced by conditions for Case 3.Source: 3) Case 3: DLAP=FBS and ULAP=MBS · p. 2
3) Case 3: DLAP=FBS and ULAP=MBS:
The joint association event for downlink through an FBS and uplink through an MBS is empty, so Case 3 has zero probability.
3) Case 3: DLAP=FBS and ULAP=MBS:
- Pr(Case 3) = 0 because the intersection defining FBS downlink and MBS uplink association is empty.Source: 3) Case 3: DLAP=FBS and ULAP=MBS · p. 2
- Case 3 is defined by the joint configuration of FBS downlink and MBS uplink association.Source: 3) Case 3: DLAP=FBS and ULAP=MBS · p. 2
- The probability of associating to an FBS in both downlink and uplink is defined; separately, the empty intersection yields Pr(Case 3) = 0 when DLAP=FBS and ULAP=MBS.Source: 4) Case 4: DLAP=ULAP=FBS · p. 2, 3) Case 3: DLAP=FBS and ULAP=MBS · p. 2
4) Case 4: DLAP=ULAP=FBS:
Case 4 is the configuration in which a device associates with an FBS for both downlink and uplink.
4) Case 4: DLAP=ULAP=FBS:
- The passage denotes M as Region 4 on Fig. 1 and introduces the association probability for Case 4.Source: 4) Case 4: DLAP=ULAP=FBS · p. 3
- The Case 4 association probability is associated with Region 4 on Fig. 1.Source: 4) Case 4: DLAP=ULAP=FBS · p. 3
- The supplied passage introduces the Case 4 probability expression.Its expression is not included in the passage.Source: 4) Case 4: DLAP=ULAP=FBS · p. 3
B. Average throughput
The throughput analysis combines association probabilities, serving-distance distributions, SINR coverage, and the average number of devices per serving BS. It also gives equivalent homogeneous-model approaches for downlink and uplink calculations.
B. Average throughput
- Average throughput per serving-BS category uses SINR coverage and the average number of devices associated with that BS.The device count is defined using device density, association probability, and BS density.Source: B. Average throughput · p. 3
- The downlink and uplink throughput calculations use serving-BS distance distributions derived under their respective association rules.The downlink distance distribution is cited from prior work, while the uplink distribution is introduced separately.Source: B. Average throughput · p. 3
- The paper derives downlink throughput expressions and uses an equivalent homogeneous model for the downlink calculation.That model combines the two BS tiers into a PPP with randomized tier power weighted by tier density.Source: B. Average throughput · p. 3
- For uplink throughput, the paper distinguishes calculations with and without decoupled access and also presents an equivalent homogeneous-model approach.The uplink equivalent model follows from association based on path loss only.Source: B. Average throughput · p. 3
- Throughput gain is defined by comparing throughput with and without decoupling for each association case.The passage also introduces an average throughput-gain calculation.Source: B. Average throughput · p. 4
IV. NUMERICAL RESULTS
Decoupled access yields a positive average UL throughput gain, though gains differ between MBSs and FBSs. Its association probability first rises with FBS density, then declines as Case 4 becomes more likely.
IV. NUMERICAL RESULTS
- More than 60% of devices choose optimal Case 2 access in a region of FBS density; as λF/λM → ∞, decoupled-access probability goes to zero.Case 2 receives downlink service from an MBS and transmits uplink through an FBS; its probability rises rapidly at first, then decreases as Case 4 grows.Source: §IV · p. 4
- At γth=2 dB, UL throughput on MBSs is 40 times higher with decoupling, while throughput on FBSs is 5 times lower; average throughput gain remains positive.Fig. 3 compares the gain with accurate device simulation and an independent-active-device approximation.Source: §IV · p. 4, Figure 3 · p. 4
- Decoupling improves UL throughput for low-SINR devices in Regions 1 and 2, at the expense of a marginal decrease for devices in Region 4.The authors suggest decoupled access as a tool toward fairness among accessing devices.Source: §IV · p. 4
V. CONCLUSION
The paper concludes that increasing FBS density relative to MBS density leads a large fraction of devices to receive DL from an MBS and transmit UL through an FBS. It presents decoupling as a tool to improve fairness in UL throughput.
V. CONCLUSION
- As FBS density rises relative to MBS density, a large fraction of devices chooses MBS downlink reception and FBS uplink transmission.This decoupled access challenges associating both directions with the same BS.Source: §V · p. 4
- The authors conclude that decoupled access can be used as a tool to improve fairness in UL throughput.Source: §V · p. 4
- Future work includes the signaling and radio access protocols needed for a decoupled-access architecture.Source: §V · p. 4