Source-linked AI summary
Resource Allocation for Downlink NOMA Systems: Key Techniques and Open Issues
S. M. Riazul Islam, Ming Zeng, Octavia A. Dobre, Kyung-Sup Kwak
TL;DR
Downlink NOMA resource allocation must jointly address user pairing and power allocation, although exhaustive optimization is computationally complex. The article surveys these algorithms, proposes D-NLUPA for controlled cluster fairness, and compares MIMO-NOMA with MIMO-OMA. Simulations show controlled gain redistribution with unchanged aggregate throughput for D-NLUPA and lower outage probability and higher effective sum rate for NOMA under the reported settings.
Problem
NOMA resource allocation must select user pairs and allocate power, while exhaustive optimization is computationally complex and dynamic schemes add signaling overhead.
Method
The article categorizes user-pairing and power-allocation algorithms, proposes D-NLUPA for cluster fairness, and compares MIMO-NOMA with MIMO-OMA under pre-defined QoS.
Results
D-NLUPA redistributes cluster gains with unchanged aggregate throughput, while MIMO-NOMA outperforms MIMO-OMA in outage probability and effective sum rate under optimal power allocation.
Takeaways & Limitations
Resource allocation is central to realizing NOMA’s benefits, and controlled cluster fairness and comparisons with OMA provide supported directions for NOMA design.
Takeaways & Limitations
The article identifies joint UP-PA optimization, low-complexity RA for FD MC-MIMO-NOMA, SFR integration, and security-aware RA as open issues.
Abstract
from arXiv · showhide
This article presents advances in resource allocation (RA) for downlink non-orthogonal multiple access (NOMA) systems, focusing on user pairing (UP) and power allocation (PA) algorithms. The former pairs the users to obtain the high capacity gain by exploiting the channel gain difference between the users, while the later allocates power to users in each cluster to balance system throughput and user fairness. Additionally, the article introduces the concept of cluster fairness and proposes the divideand- next largest difference-based UP algorithm to distribute the capacity gain among the NOMA clusters in a controlled manner. Furthermore, performance comparison between multiple-input multiple-output NOMA (MIMO-NOMA) and MIMO-OMA is conducted when users have pre-defined quality of service. Simulation results are presented, which validate the advantages of NOMA over OMA. Finally, the article provides avenues for further research on RA for downlink NOMA.
I. INTRODUCTION
The article surveys downlink NOMA resource allocation, emphasizing user pairing and power allocation, and introduces cluster fairness alongside comparisons with MIMO-OMA.
- Research scope: Downlink NOMA resource allocation assigns users to pairs and allocates power within clusters, but exhaustive optimization is computationally complex.Dynamic pairing and power allocation can also add SIC decoding-order and power-ratio signaling overhead.
- Contributions: The article categorizes existing user-pairing and power-allocation algorithms for downlink NOMA.
- Contributions: D-NLUPA is proposed to provide cluster fairness by controlling the desired sum-rate gain across NOMA clusters.
- Contributions: The study compares MIMO-NOMA and MIMO-OMA when users have pre-defined QoS requirements.
- Contributions: The article identifies challenges and open issues for downlink NOMA resource allocation.
II. USER PARING IN NOMA
User pairing determines how users are grouped into NOMA clusters, trading channel-gain-based performance against complexity, interference handling, and pairing feasibility.
- Pairing strategies: Random pairing has the lowest complexity but suboptimal sum-rate performance because it ignores users’ channel gains.
- Pairing strategies: NLUPA pairs the highest-gain user with the lowest-gain user, then matches the next-highest and next-lowest users.Larger channel-gain differences provide higher NOMA performance gains under fixed power allocation.
- Interference and feasibility: Vertical pairing assigns adjacent sub-channels sequentially to user pairs and uses additional SIC to cancel interference from previous clusters.The additional SIC operations increase computational complexity.
- Interference and feasibility: When insufficient strong users leave weak users unpaired, hybrid pairing can access leftovers through OMA, whereas virtual pairing shares a frequency band among two weak users and one strong user.
- Complexity reduction: Pre-defined user grouping reduces pairing comparisons and signaling overhead by allowing pairs only across groups formed from channel conditions.Fairness metrics and SINR thresholds can further guide candidate selection.
- Generalized pairing: Matching games can pair users and subchannels to pursue maximum weighted sum-rate when partition-based grouping methods are unsuitable.
B. UP in MIMO-NOMA
In MIMO-NOMA, user pairing interacts with precoding and channel structure, making pairing selection a multidimensional design problem rather than a simple channel-gain ordering.
- MIMO-NOMA pairing: With zero-forcing precoding in a two-user cluster, the weak user does not influence the strong user’s rate, so the strong user can be selected first.The weak user is then chosen to optimize the performance metric.
- MIMO-NOMA pairing: Large channel-gain differences support NOMA effectiveness, while high channel correlation helps eliminate inter-cluster interference.
- MIMO-NOMA pairing: QDC precoding exploits quasi-degraded NOMA channels to formulate a low-complexity sequential pairing algorithm that reduces transmit power.The approach is not efficient when the base station has more transmit antennas than downlink users.
C. The Concept of Divide-and-NLUPA
D-NLUPA modifies NLUPA by dividing ordered users before pairing, imposing a minimum pairing range so every cluster receives a controlled minimum gain.
- Motivation: NLUPA’s gain depends on the channel-gain distance between paired strong and weak users, which increases with their ordering range.The first NLUPA cluster therefore receives the maximum gain, while later clusters may receive negligible gain.
- D-NLUPA design: D-NLUPA introduces a divide step that sets a minimum range and increases the corresponding minimum channel-gain distance.This avoids clusters with near-zero gain and guarantees a designed minimum gain.
- D-NLUPA design: For N = 16, ordered users are divided into four sets, sets 1 and 3 are merged into A, sets 2 and 4 into B, and NLUPA is applied within both merged sets.The minimum distance is evaluated across the two resulting sets.
- Simulation results: A minimum distance of around 15 dB controls D-NLUPA’s gain and guarantees a minimum sum-rate gain for each cluster.
- Simulation results: About 50% of D-NLUPA clusters achieve higher gains than NLUPA clusters, while the remaining clusters achieve lower gains.Aggregated throughput remains the same for both algorithms, so D-NLUPA redistributes gains among clusters rather than increasing total throughput.
- Simulation results: Random pairing obtains the lowest sum-rate gain in the comparison.
III. POWER ALLOCATION IN NOMA
Power allocation in NOMA determines interference management, rate distribution, user admission, fairness, and outage, motivating classification across SISO, multicarrier, and MIMO settings.
- PA directly affects interference management, rate distribution, user admission, and outage because users share the power domain.
- The article organizes PA strategies across single-carrier SISO, multicarrier, and MIMO systems.
- Fig. 3 evaluates user-pairing algorithms with fixed PA using sum-rate gain versus distance and cluster index.
- Fig. 4 classifies the PA strategies discussed in the article.
A. PA in SISO-NOMA
SISO-NOMA power allocation evolves from throughput-maximizing but unfair allocation toward schemes that incorporate channel conditions and QoS, while balancing throughput and fairness remains unresolved.
- Allocating all power to the best-channel user maximizes NOMA sum rate but causes extreme unfairness and reduces admitted users.
- F-PA cannot satisfy varied QoS requirements because it ignores users’ specific channel gains.
- FTPC allocates power inversely to channel gain with a decaying factor, but using one factor for all users is suboptimal.
- Under perfect CSI, weighted-sum-rate PA is convex, max-min fairness is quasi-convex, and energy-efficient PA uses iterative convex subproblems.
- CR-inspired PA protects the weak user’s QoS, whereas dynamic PA is proposed to address the resulting strong-user performance sacrifice.
B. PA in MC-NOMA and MIMO-NOMA
MC-NOMA and MIMO-NOMA extend resource allocation across subcarriers and spatial dimensions, but their benefits depend on CSI, interference, channel structure, and user pairing.
- MC-NOMA allows users and subcarriers to occupy multiple counterparts, with performance depending on power allocation and subcarrier assignment.
- Perfect CSI assumptions may be impractical in overloaded MC-NOMA, motivating resource allocation under statistical CSI.
- Without perfect CSI, the base station must derive an explicit SIC decoding order before performing power allocation and subcarrier assignment.
- MIMO-NOMA faces open questions about capacity for general channels and lacks a natural user-channel ordering because channels are matrices or vectors.
IV. PERFORMANCE COMPARISON BETWEEN MIMO-NOMA AND MIMO-OMA
The comparison evaluates MIMO-NOMA and MIMO-OMA under minimum-rate requirements, using admission control and alternative intra- and cross-cluster power-allocation strategies. Simulations report better outage and effective sum-rate performance for MIMO-NOMA.
- The comparison imposes minimum-rate QoS requirements and admits as many users as possible when all users cannot be served.
- For NOMA, power first satisfies the weak user’s QoS when possible, allocating remaining power to the strong user for sum-rate maximization.
- For OMA, power satisfies both users’ QoS when possible, then applies water-filling to distribute remaining power.
- The simulations present outage probability and effective sum rate for equal-power and cross-cluster power-allocation schemes.
- MIMO-NOMA outperforms MIMO-OMA in outage performance and effective sum rate; cross-cluster PA further improves both metrics at high transmit power.
V. PRACTICAL CHALLENGES AND RESEARCH DIRECTIONS
Downlink NOMA resource allocation faces practical challenges including scalability, interference, limited feedback, QoS support, and multi-hop operation. Joint UP–PA optimization remains difficult, particularly for MIMO-NOMA, where pairing-dependent precoding can reduce differences between algorithms.
- Resource allocation remains in its infancy because NOMA must address scalability, inter-cell interference, carrier aggregation, limited channel feedback, QoS guarantees, and multi-hop communications.
- Fig. 6 plots sum rate against transmit power.
- Joint optimization of user pairing and power allocation is desirable but challenging even for SISO, while MIMO-NOMA requires further effective strategies.The paper identifies joint UP–PA optimization as an open research direction.
- Under the cited MIMO-NOMA model, D-NLUPA, NLUPA, and random pairing show similar performance because pairing changes the precoding and detection matrices.These changes randomize the gain associated with path-loss-based pairing.
B. RA for FD MC-MIMO-NOMA
Integrating NOMA with full-duplex, multi-carrier, and MIMO technologies creates a highly complex resource-allocation problem. The paper highlights the need for low-complexity schemes while noting fairness and interference issues in NOMA with soft frequency reuse.
- B. RA for FD MC-MIMO-NOMA: FD MC-MIMO-NOMA combines full-duplex, multi-carrier, and MIMO technologies but makes resource allocation substantially more complicated.The combined system is described as non-trivial for resource-allocation design.
- B. RA for FD MC-MIMO-NOMA: The combinatorial optimization problem of MC-NOMA is NP-hard, and adding MIMO and full-duplex makes the problem much more complicated.
- C. RA for NOMA with Soft Frequency Reuse: NOMA over soft frequency reuse can overlap strong and weak users on the primary band, increasing strong-user rates while decreasing weak-user rates.Using the primary band at the cell edge also generates inter-cell interference.
- α-fairness uses one scalar to provide different fairness levels and efficiency–fairness tradeoffs, supporting lower-complexity throughput optimization with fairness constraints.
E. Security-Aware RA
NOMA’s SIC-based decoding creates security vulnerabilities because strong users decode weak-user signals. The paper therefore identifies security-aware resource allocation and physical-layer security measures as open research directions.
- E. Security-Aware RA: A strong-channel user must decode the weak-channel user’s signal, creating security concerns in NOMA communication.
- E. Security-Aware RA: An attacked or malicious weak user can make the decoding operations of both strong and weak users unreliable.One example is altering the channel quality indicator during feedback.
- E. Security-Aware RA: The paper identifies physical-layer security measures and security-aware resource allocation as open problems for NOMA systems.