Abstract
Traditional 5G network optimization often prioritizes raw throughput, yet user satisfaction—or Quality of Experience (QoE)—is not always linearly related to data rates. This paper investigates resource allocation in Multi-Cell Multi-Carrier Non-Orthogonal Multiple Access (MC-NOMA) networks with the goal of maximizing the sum of users' Mean Opinion Scores (MOS). By decomposing the joint optimization of user association, subchannel assignment, and power allocation, the authors propose tractable solutions that balance performance and complexity. Results demonstrate that NOMA significantly outperforms traditional OMA systems while providing much better fairness among users.
Introduction
As 5G networks evolve, the focus is shifting from "sum-rate" maximization to "Quality of Experience" (QoE). While a high data rate is desirable, the subjective satisfaction of a user (MOS) often follows a logarithmic trend with diminishing returns. Consequently, a network that maximizes total throughput might leave some users with poor experiences, whereas a QoE-driven approach naturally promotes fairness.
The authors address this in the context of Multi-Cell Multi-Carrier Non-Orthogonal Multiple Access (MC-NOMA). NOMA allows multiple users to share the same frequency and time resources by multiplexing them in the power domain, as illustrated in the system model below.
Managing such a network involves three critical, interdependent decisions:
- User-BS Association: Determining which base station (BS) serves which user.
- Subchannel Assignment: Allocating frequency bands to users.
- Power Allocation: Distributing power to enable effective Successive Interference Cancellation (SIC).
Methodology
The joint optimization problem is formulated as a mixed-integer non-linear programming (MINLP) problem, which is inherently NP-hard. To solve it, the authors employ a problem transformation strategy that decomposes the global objective into more manageable subproblems.
3-D Matching for Resource Assignment
The relationship between users, BSs, and subchannels is treated as a 3-D matching problem. To avoid the computational exhaustion of a direct 3-D search, a two-step 2-D matching approach is used:
- Step I: Users are matched to BSs using an Extended Deferred Acceptance (EDA) algorithm.
- Step II: The resulting (User, BS) pairs are matched to specific subchannels using "swap-matching" to reach a stable state.
Power Allocation Optimization
With the matching fixed, the remaining continuous problem is power allocation. The authors explore two paths:
- Global Optimum: A Branch and Bound (BB) algorithm provides the best possible solution but is computationally intensive.
- Low-Complexity Alternative: A Successive Convex Approximation (SCA) approach transforms the non-convex problem into a series of convex subproblems, offering near-optimal performance with significantly lower overhead.
Key Insights and Results
Numerical simulations in a multi-cell environment (3 BSs, up to 18 users) reveal several critical findings:
- Inherent Fairness: Because the MOS function (calibrated for web browsing) is logarithmic, the optimization algorithm prioritizes users with lower data rates where the marginal MOS gain is highest. This leads to a Jain’s Fairness Index approaching 0.9, significantly higher than throughput-centric models.
- NOMA Advantage: NOMA consistently outperforms traditional Orthogonal Multiple Access (OMA) by better exploiting channel gain diversity, allowing multiple users to achieve high satisfaction on the same subchannel.
- Interference Thresholds: In multi-cell scenarios, inter-cell interference becomes a limiting factor at high transmit powers (above -10 dBm), indicating that simply increasing power does not indefinitely improve QoE.
Conclusion
This research demonstrates that QoE-driven resource allocation in NOMA networks provides a superior balance of user satisfaction and fairness compared to traditional methods. While the current study focuses on static users and web browsing traffic, future work will extend these algorithms to high-mobility scenarios and mixed traffic types, such as video streaming and VoIP, to further refine 5G network management.
Summary updated