Source-linked AI summary
Load & Backhaul Aware Decoupled Downlink/Uplink Access in 5G Systems
Hisham Elshaer, Federico Boccardi, Mischa Dohler, Ralf Irmer
TL;DR
Traditional strongest-downlink association is suboptimal in heterogeneous 5G networks, motivating independent downlink and uplink association. This paper extends DUDe with cell-load and backhaul awareness, evaluates it in a realistic Vodafone-based network, and shows that performance depends strongly on power control while load- and backhaul-aware DUDe improves key throughput and load-balancing results.
Problem
Traditional link-quality-based association omits cell load and backhaul capacity, despite their impact on DUDe performance and the growing importance of uplink capacity for cell-edge users.
Method
The paper introduces a DUDe association algorithm incorporating link quality, cell load, and backhaul capacity, and evaluates it with realistic power-control settings.
Results
DUDe-Load improves the 5th-percentile throughput by 15% with interference-aware power control and by about 40% over DUDe under conservative power control.
Takeaways & Limitations
Load- and backhaul-aware association improves uplink fairness and load balancing, while power-control choices materially affect the resulting throughput and interference.
Abstract
from arXiv · showhide
Until the 4th Generation (4G) cellular 3GPP systems, a user equipment's (UE) cell association has been based on the downlink received power from the strongest base station. Recent work has shown that - with an increasing degree of heterogeneity in emerging 5G systems - such an approach is dramatically suboptimal, advocating for an independent association of the downlink and uplink where the downlink is served by the macro cell and the uplink by the nearest small cell. In this paper, we advance prior art by explicitly considering the cell-load as well as the available backhaul capacity during the association process. We introduce a novel association algorithm and prove its superiority w.r.t. prior art by means of simulations that are based on Vodafone's small cell trial network and employing a high resolution pathloss prediction and realistic user distributions. We also study the effect that different power control settings have on the performance of our algorithm.
I. INTRODUCTION
In heterogeneous 5G networks, independent downlink/uplink association can substantially improve uplink capacity, especially for cell-edge users. Prior approaches, however, omit cell load and backhaul constraints that affect realistic performance.
- 200-300% uplink gains were reported for cell-edge users when downlink and uplink associations were decoupled.
- Growing uplink traffic makes uplink optimization, particularly for disadvantaged cell-edge users, important for consistent 5G quality of experience.
- Increasing heterogeneity in 4G and 5G networks creates more connection opportunities for independently associating uplink and downlink traffic.
- DUDe connects users between the small-cell downlink and uplink borders to the macro cell in downlink and small cell in uplink.
- Prior association methods use downlink received power and uplink pathloss but ignore cell load and backhaul capacity.
A. Related Work
Related work establishes DUDe as part of future cellular and device-centric network architectures, while backhaul-aware association has also been studied separately.
- DUDe has been discussed as a major component of future cellular networks and as part of device-centric architectures tailored to individual devices and sessions.
B. Contributions
The paper extends DUDe association beyond link quality by incorporating cell load and backhaul capacity, and evaluates the approach in a realistic Vodafone-based scenario.
- The proposed association algorithm considers link quality, cell load, and cell backhaul capacity rather than uplink pathloss alone.
- The evaluation uses a realistic cellular-network scenario based on Vodafone’s planning and optimisation tools.
- The paper presents the system model, extended association algorithm, simulation setup, results, and conclusions across Sections II–VI.
II. SYSTEM MODEL
The system model uses a dense Vodafone small-cell test network, realistic traffic distributions, uplink SINR and rate models, multiple power-control schemes, and limited backhaul.
- The simulated HetNet covers approximately one square kilometre of Vodafone’s London small-cell test network.
- Users follow peak-time field-trial traffic distributions, while flow-level traffic represents individual file or data transfers.
- Uplink link quality is modeled through pathloss, fading, interference, and UE transmit power, with SINR and Shannon-based achievable rate.
- The model compares 3GPP open-loop power control with interference-aware control that limits UE power according to interference caused to neighboring cells.
- All cells have limited backhaul capacity, with tighter constraints expected for small cells than macro cells.
III. CELL ASSOCIATION ALGORITHM
The proposed UL association extends pathloss-based DUDe by incorporating cell load and backhaul capacity, using a distributed, flow-by-flow cell-selection procedure.
- Association criterion: The association criterion extends pathloss-based DUDe by jointly considering link quality, cell load, and backhaul capacity.
- Load estimation: Because UL resource utilization poorly represents load under UE power limits, the algorithm estimates load from the stationary average number of flows.A UE with poor channel conditions may use few resource blocks despite the cell serving many UEs.
- Distributed operation: The fully distributed procedure lets each UE retain its DL anchor while selecting the highest-criterion UL cell whenever it has a flow to transmit.
- Distributed operation: Base stations periodically broadcast their load and backhaul capacity for use in UE association decisions.
- Distributed operation: After a random waiting period, each UE connects to the selected base station until its flow completes, then disconnects and returns to idle.
IV. SIMULATION SETUP
The evaluation uses a realistic Vodafone London LTE deployment, varied power-control settings, and three UL association strategies to compare conventional, DUDe, and load/backhaul-aware operation.
- Deployment: The simulation deployment contains 5 macro cells and 21 outdoor small cells in Vodafone’s London LTE test network.
- Deployment: Propagation is modeled with high-resolution 3D ray tracing that incorporates clutter, terrain, and building data, while users follow real-network traffic distributions.
- Power control: The study evaluates three power-control settings: loose full compensation, conservative partial compensation, and interference-aware power control.
- Association cases: The association comparison includes conventional DL-RSRP, pathloss-based DUDe, and DUDe-Load with cell-load and backhaul-capacity awareness.
V. RESULTS & DISCUSSIONS
The results show that load- and backhaul-aware DUDe improves lower-percentile throughput and load balancing, while power control and backhaul capacity shape the trade-offs across user percentiles. DUDe also reduces SINR variance, indicating more stable interference than DL-RSRP.
- Power-control settings: More than 100% and 150%: DUDe increases the 5th and 50th percentile throughputs, respectively, over DL-RSRP under both power settings.The gains are attributed to load balancing and improved link quality from connecting UEs to cells with the lowest pathloss.
- Power-control settings: About 20%: DUDe-Load reduces the 5th percentile throughput but increases the 50th percentile throughput by 40% versus DUDe under Setting 1.The same comparison reports about a 20% loss at the 90th percentile, reflecting a trade-off between peak and cell-edge or average throughput.
- Power-control settings: About 40%: DUDe-Load improves the 5th percentile throughput over DUDe under Setting 2, while the 50th percentile throughput is almost unchanged.Setting 2 lowers network interference, benefiting cell-edge UEs connected to a suboptimal cell in pathloss.
- Interference stability: About 10 dB and 15 dB: DUDe and DUDe-Load reduce average SINR-variance versus DL-RSRP under Setting 1, respectively.The lower variance is associated with more stable interference and improved resource utilization from load balancing.
- Interference-aware power control: DUDe-Load achieves similar or higher 5th-percentile throughput than Setting 2 and 20% higher 50th-percentile throughput than DUDe with interference-aware power control.The scheme reduces interference from high-interference UEs while allowing other UEs to transmit at higher power.
- Backhaul capacity: DUDe-Load has the highest 5th-percentile throughput as small-cell backhaul capacity increases, while DL-RSRP saturates after 10 Mbps and DUDe-Load surpasses it.At the 50th percentile, DUDe-Load also outperforms DUDe across capacities, with gains increasing as small-cell backhaul capacity rises.
- Load balancing: 470, 83 and 21: the variance of UEs per cell for DL-RSRP, DUDe and DUDe-Load, respectively, shows progressively stronger load balancing.DUDe-Load additionally balances UEs among small cells, not only between macro and small cells.
VI. CONCLUSIONS
The paper extends DUDe association by incorporating cell load and backhaul constraints, then evaluates the approach in realistic simulations. Load-aware DUDe improves throughput and stabilizes uplink interference, with gains depending strongly on power control.
- The proposed DUDe extension considers cell load and backhaul constraints rather than link quality alone.
- Load-aware DUDe further improves system throughput over baseline DUDe.
- 10-15 dB lower uplink SINR variance than baseline LTE facilitates radio resource management and self-organizing network operations.
- 15% and 20% throughput gains over baseline DUDe occur at the 5th and 50th percentiles under interference-aware power control.
- Future work targets DUDe-specific scheduling that maximizes overall uplink and downlink capacity under constrained backhaul.