Source-linked AI summary
Adaptive Spatial Intercell Interference Cancellation in Multicell Wireless Networks
Jun Zhang, Jeffrey G. Andrews
TL;DR
The paper asks when downlink ICIC is preferable to standard single-cell beamforming and studies this trade-off analytically. It proposes adaptive coordination based on user locations and evaluates limited-feedback effects, finding gains from adaptive and carefully designed ICIC strategies.
Problem
Other-cell interference limits multicell wireless-network performance, creating a need to determine when ICIC is preferable to standard beamforming.
Method
The paper proposes adaptive ICIC, in which multiple base stations jointly select transmission strategies using user locations, and investigates limited-feedback designs.
Results
At medium edge SNR, adaptive ICIC outperforms systems without ICIC and with static ICIC, while ICIC retains significant throughput gain with carefully designed limited-feedback strategies.
Takeaways & Limitations
Adaptive coordination provides throughput gains with low overhead because base stations exchange user locations, while CSI is needed only when ICIC is applied.
Abstract
from arXiv · showhide
Downlink spatial intercell interference cancellation (ICIC) is considered for mitigating other-cell interference using multiple transmit antennas. A principle question we explore is whether it is better to do ICIC or simply standard single-cell beamforming. We explore this question analytically and show that beamforming is preferred for all users when the edge SNR (signal-to-noise ratio) is low ($<0$ dB), and ICIC is preferred when the edge SNR is high ($>10$ dB), for example in an urban setting. At medium SNR, a proposed adaptive strategy, where multiple base stations jointly select transmission strategies based on the user location, outperforms both while requiring a lower feedback rate than the pure ICIC approach. The employed metric is sum rate, which is normally a dubious metric for cellular systems, but surprisingly we show that even with this reward function the adaptive strategy also improves fairness. When the channel information is provided by limited feedback, the impact of the induced quantization error is also investigated. It is shown that ICIC with well-designed feedback strategies still provides significant throughput gain.
I. INTRODUCTION
The paper examines downlink multicell interference mitigation, contrasting coordinated single-cell ICIC with coordinated multicell transmission and standard beamforming. It proposes adaptive ICIC to balance interference suppression against the spatial, information-exchange, CSI, synchronization, and complexity costs of coordination.
- Motivation: Other-cell interference limits the performance of contemporary multicell wireless networks, especially on the downlink.The downlink is often the capacity-limiting link because mobile users generally cannot cooperate.
- Proposed strategy: The proposed adaptive ICIC strategy lets multiple base stations jointly select transmission techniques from user locations, requiring location exchange and CSI only when ICIC is used.This design targets lower coordination overhead while adapting transmission to network conditions.
- Coordination approaches: Coordinated multicell transmission can ideally eliminate other-cell interference with full CSI and centralized data availability.This approach uses joint processing across base stations but requires substantial coordination infrastructure.
- Coordination approaches: Coordinated multicell transmission imposes high overhead and complexity through backhaul exchange, synchronization, and CSI requirements at coordinated base stations.These demands challenge practical implementation and make CSI estimation and feedback difficult.
- Coordination approaches: Coordinated single-cell transmission avoids inter-base-station data exchange and generally requires CSI from only some neighboring base stations.It therefore has lower overhead and complexity than coordinated multicell transmission.
- Proposed strategy: The paper applies coordinated single-cell transmission as zero-forcing ICIC, but canceling neighboring-cell interference consumes spatial degrees of freedom and reduces home-user signal power.Because of this trade-off, ICIC is not necessarily optimal at every base station.
B. Contributions
The paper develops spatial ICIC and an adaptive strategy that selects transmission strategies using user locations, then evaluates throughput, fairness, and limited-feedback requirements in multicell networks.
- The paper investigates downlink spatial ICIC using ZF precoding to suppress downlink OCI.
- Closed-form ergodic sum-rate expressions support joint selection of each BS's transmission strategy based on user locations.
- Strategy selection: High edge SNR favors ICIC, low edge SNR favors beamforming, and medium edge SNR favors the adaptive strategy.
- Numerical results: In a 3-cell network at 15 dB average edge SNR, average throughput increases by about half and edge throughput increases three-fold.
- Fairness: Using sum throughput as the metric, the adaptive strategy also encourages fairness by helping neighboring-cell edge users.
- Limited feedback: To maintain constant rate loss relative to perfect CSI, feedback bits to a neighboring helper BS must grow linearly with transmit antennas and edge SNR in dB.
A. Adaptive Coordination
Adaptive coordination chooses among selfish beamforming and interference cancellation at multiple BSs by maximizing sum throughput from location-dependent channel conditions.
- Transmission choices: Each BS can choose selfish beamforming or interference cancellation for neighboring cells.
- Transmission choices: In a 2-cell network, the strategy set is S1 = {BF, IC} and S2 = {BF, IC}.
- Optimization: Adaptive ICIC selects the strategy pair maximizing R1(s1, s2) + R2(s1, s2).
- Information exchange: The adaptation uses active-user locations, while instantaneous neighboring-user CSI is needed only when ICIC is applied.
- Outcome: Maximizing sum throughput also increases both sum throughput and edge throughput, providing fairness in the reported analysis.
B. Transmission Strategies
The transmission strategies trade own-user signal power against interference suppression, using eigen-beamforming for selfish transmission and ZF precoding for ICIC.
- Selfish beamforming: Eigen-beamforming uses the normalized home-user channel direction to maximize received signal power in the single-cell MISO setting.
- ICIC through ZF precoding: ZF ICIC chooses a precoder in the nullspace of neighboring users' channels while maximizing the desired signal component.
- ICIC through ZF precoding: With Nt antennas, a BS can maximally precancel interference for up to Nt − 1 neighboring cells.
- Strategy trade-off: ZF is used despite MMSE's low-SNR advantage because the analysis finds negligible loss where low edge SNR does not require ICIC.
C. Signal Power and Interference Power
The paper derives closed-form throughput expressions for strategy pairs in a 2-cell model and explains their selection through noise- and interference-limited regimes.
- Throughput analysis: Closed-form ergodic achievable throughputs for strategy pairs can be used to select strategies maximizing sum throughput.
- Regime analysis: Noise-limited users favor (BF, BF) because ICIC offers marginal gain while beamforming increases own-user received signal power.
- Regime analysis: Interference-limited users favor (IC, IC) because higher OCI makes neighboring-cell interference cancellation beneficial for sum throughput.
- Regime analysis: When one user is interior and the other is at the edge, the predicted strategy pair is (IC, BF).The interior user's throughput is less sensitive to signal-power reduction, while the edge user is OCI-limited and power-limited.
- Scope: The analysis notes that the heuristic strategy pair may differ from the actual selection, which also depends on noise level and edge SNR.
- Location adaptation: At 5 dB, strategy selection depends on locations: interiors favor (BF, BF), edges favor (IC, IC), and mixed locations trigger ICIC for the edge user.
IV. FROM 3-CELL TO MULTICELL NETWORKS
The paper extends adaptive coordination from a simplified 2-cell setting to 3-cell networks, derives closed-form throughput expressions, and proposes approaches for general multicell networks.
- The simplified 2-cell investigation motivated adaptive coordination, but its result cannot be readily implemented in a general multicell network.
- The paper extends its adaptive strategy to a 3-cell network.
- The authors derive closed-form expressions for achievable throughput under strategy selection in 3-cell networks.
- The paper proposes approaches to extend adaptive coordination from 3-cell networks to general multicell networks.
A. The Strategy Set
For a 3-cell network, each base station selects among beamforming and interference-cancellation strategies, with coordinated selection maximizing sum throughput under shared strategy knowledge.
- Each base station has four initial transmission strategies when three cells coordinate.
- The strategy set combines beamforming with ICIC for either neighboring cell or both neighboring cells.
- ICIC for both neighboring cells requires N_t ≥3, while combining single-neighbor choices reduces strategy-set size and selection complexity.
- The reduced strategy set yields 3^3 = 27 strategy combinations for three users.
- Theorem 2 gives ergodic achievable throughput for user 1 with given user locations and perfect CSI, enabling strategy selection to maximize sum throughput.
- The three base stations jointly determine the strategy tuple because they share a common objective and each knows the others’ strategies.
C. Extension to Multicell Networks
The paper extends adaptive ICIC to general multicell networks through local 3-cell coordination, sectoring, or distributed decisions, while incorporating limited-feedback CSI design.
- The proposed adaptive ICIC strategy is extended to general multicell settings by estimating local sum-throughput gain and selecting ICIC or beamforming accordingly.
- Cell sectoring can organize coordination among every three neighboring cells, whose base stations jointly select transmission strategies.
- A distributed implementation lets each base station independently determine its transmission strategy.
- Local 3-cell subnetwork estimates can be used in arbitrary-sized networks by treating interference from outer cells as white Gaussian noise without cluster structure.
- Limited-feedback CSI introduces channel-direction quantization error, requiring feedback design for adaptive ICIC transmission.
- Users quantize channel directions with unit-norm codebooks of size 2^B and feed back B-bit codeword indices.
- Channel directions from multiple neighboring base stations are independently quantized, and feedback codebook sizes may differ across users and base stations.
B. Throughput Analysis
The paper analyzes throughput under limited-feedback CSI and shows that quantization affects desired signal and ICIC interference differently, producing residual interference that can limit high-SNR throughput.
- Limited feedback changes the mean of the desired signal term but not its distribution.
- With ICIC, limited feedback produces residual interference, whereas without ICIC the interference distribution remains unchanged.
- At high edge SNR, signal-term quantization causes a constant rate loss of log ξ_i,i, while residual OCI increases with edge SNR and limits throughput.
- The CDI need not have the same accuracy for the home and helper base stations, allowing flexible feedback design.
- Theorem 3 approximates achievable throughput for a 3-cell network with given user locations and limited feedback.
- With limited feedback, strategy selection depends on user locations, average edge SNR, and the number of feedback bits.
- With B = 10 feedback bits per channel direction, approximations are very accurate, ICIC gains are reduced, and the (IC, IC) strategy remains preferred.
- Limited feedback shrinks the operating regions associated with ICIC strategies.
C. Limited Feedback Design
The paper develops limited-feedback and adaptive ICIC designs for multicell networks, addressing feedback allocation and transmission-strategy selection across SNR regimes. Adaptive ICIC outperforms fixed alternatives at medium SNR while reducing CSI requirements relative to static ICIC.
- Limited-feedback design: With fixed feedback, the design problem allocates bits between home-BS and helper-BS CSI rather than enforcing equal allocation.
- Performance comparison: At low P0, static ICIC provides lower throughput than the other two systems, indicating that ICIC is unnecessary in this regime.
- Performance comparison: At high P0, static and adaptive ICIC provide significant throughput gains over no ICIC, including 53% average and 210% edge gains at P0 = 15 dB.
- Performance comparison: Adaptive ICIC selects transmission strategies jointly across base stations based on user locations.
- Performance comparison: At medium P0 (0 ∼5), adaptive ICIC outperforms both no ICIC and static ICIC.
- CSI requirements: Adaptive ICIC reduces required CSI relative to static ICIC because neighboring-BS CSI is needed only when that BS performs ICIC for the user.
B. Impact of limited feedback
The limited-feedback analysis studies how quantized CSI affects ICIC and how feedback should scale or be allocated. Proper helper-BS feedback preserves near-constant rate loss and retains significant throughput gains over no ICIC.
- Feedback scaling: Scaling helper-BS feedback can maintain a rate loss below 1 bps/Hz relative to perfect CSI for average and edge throughput.
- Feedback scaling: At high edge SNR, home-BS feedback need not scale because a fixed number of bits causes only a fixed rate loss, whereas helper-BS feedback must increase.
- Feedback allocation: With 30 total feedback bits, adaptive allocation improves high-P0 performance over uniform allocation by approximately 6% in average throughput and 11% at the edge.
- Feedback allocation: The gains from adaptive allocation are marginal because the available feedback budget cannot meet the requirement BI = 18 at P0 = 15 for a 1 bps/Hz loss target.
- Overall impact: Even with limited feedback, adaptive ICIC systems provide significant average and edge throughput gains over systems without ICIC.
- Conclusions: The conclusions report average and edge throughput gains, with adaptive ICIC outperforming both no ICIC and static ICIC at medium edge SNR.
- Scope: The reported results are based on two- and three-cell networks, while extension to larger networks is conjectured rather than established.
APPENDIX
The appendix derives achievable-throughput expressions for limited-feedback beamforming and ICIC in three-cell networks. It uses distributional calculations, approximations, and lemmas to characterize signal, interference, and rate-loss terms.
- Analytical derivation: The appendix derives the cumulative distribution function of a random variable and then evaluates the expectation of ln(1 + x).
- Analytical derivation: Integration by parts and integral expressions are used to obtain a closed-form representation involving I1 and I2.
- Analytical derivation: The exponential-integral function of the first order, E1(x), appears in the expression for I2.
- Limited-feedback model: For limited feedback, the signal and interference terms are approximated using quantized channel directions and the quantization-cell approximation.
- ICIC construction: Under ICIC, the beamforming vector is the projection of the quantized home-channel direction onto the nullspace of neighboring quantized channel directions.
- Three-cell rate analysis: The three-cell achievable-rate expression combines desired signal power with interference contributions from the two neighboring base stations.
- Rate-loss analysis: At high edge SNR, ICIC throughput is limited by residual OCI, and the rate-loss analysis determines helper-BS feedback requirements.