Hybrid NOMA Slicing

• Hybrid Non-Orthogonal Slicing (H-NOMA) is a communication paradigm that combines NOMA and OMA to enable flexible and efficient spectrum allocation for heterogeneous 5G/6G services.
• It leverages advanced SIC strategies, including dynamic decoding order and power adaptation (HSIC-PA), to significantly improve throughput, energy efficiency, and reliability over traditional methods.
• Analytical and simulation results demonstrate that H-NOMA can lower error probabilities and achieve 20–60% energy savings, highlighting its practical benefits in network slicing.

Hybrid Non-Orthogonal Slicing (H-NOMA) represents a resource allocation paradigm that merges the principles of non-orthogonal multiple access (NOMA) with classical orthogonal multiple access (OMA) to enable more efficient spectrum utilization and flexible service multiplexing in wireless networks. The H-NOMA concept is especially relevant for 5G/6G slicing scenarios that must simultaneously satisfy heterogeneous requirements—e.g., enhanced mobile broadband (eMBB), ultra-reliable low-latency communications (URLLC), and massive machine-type communication (mMTC)—while coping with the limitations of pure OMA and NOMA schemes. H-NOMA leverages successive interference cancellation (SIC) in various dynamic and hybrid forms, including advanced hybrid SIC (HSIC) and power adaptation strategies, to ensure that non-orthogonal spectrum reuse yields strict or almost-sure improvements over OMA in throughput, energy efficiency, and reliability.

1. Hybrid NOMA System Architecture and Slicing Models

In H-NOMA systems, network slicing is realized by flexibly combining OMA (e.g., TDMA/FDMA) resource assignments with NOMA-based spectrum sharing phases. Consider an uplink frame of $T$ symbol durations with $M$ users transmitting to a base station. In the legacy OMA (TDMA) scenario, each user $U_n$ is allocated an exclusive time slot, achieving per-frame rate

$$R_{\text{OMA}}(n) = T \log_2\big(1 + \rho_n |h_n|^2\big),$$

with energy $E_{\text{OMA}}(n) = T \rho_n$, where $|h_n|^2$ is the Rayleigh fading channel gain and $\rho_n$ the transmit power.

In H-NOMA, certain users (“opportunistic users” $U_n$) obtain an additional transmission opportunity by being allowed to transmit at reduced power ($\beta \rho_n$ with $\beta \in (0, \frac12)$) in another user’s (“legacy user” $U_m$) slot, as well as in their own slot, but under strict constraints to ensure that total energy expenditure does not exceed OMA: $$E_{\text{H-NOMA}}(n) = 2T \beta \rho_n < T \rho_n = E_{\text{OMA}}(n).$$ Pairing—the mapping between each $U_n$ and its paired $U_m$—and the value of $\beta$ directly shape the achievable rate–energy region (Sun et al., 2024, Wang et al., 18 Sep 2025).

2. Hybrid SIC Strategies: Decoding Order and Power Adaptation

SIC within the context of H-NOMA determines the achievable rates of co-scheduled users. Three central SIC variants are established:

• Fixed-order SIC (FSIC): The decoding order per NOMA slot is static, e.g., always decode $U_m$ first, then $U_n$ (Wang et al., 18 Sep 2025). This rigid approach leads to persistent error floors at high SNR unless the power ratios are extremely favorable.
• Hybrid SIC without Power Adaptation (HSIC-NPA): The base station dynamically selects which user to decode first, based on the instantaneous received powers compared to an interference-tolerance threshold $\tau_m$, but does not further adapt the transmit powers (Sun et al., 2024, Wang et al., 18 Sep 2025).
• Hybrid SIC with Power Adaptation (HSIC-PA): On top of dynamic order selection, opportunistic users decrease their in-slot power as needed to precisely meet legacy users’ maximum tolerable interference threshold. This approach provably removes the error floor at high SNR and guarantees almost sure rate and energy-efficiency superiority over OMA without restrictive conditions on power ratios or rate requirements (Ning et al., 13 Jul 2025, Wang et al., 18 Sep 2025).

The core HSIC-PA mechanism operates as follows: For a paired ($U_m$, $U_n$), the base station calculates the interference-tolerance threshold for $U_m$ as

$$\tau_m = \max\left\{0, \frac{\rho_m |h_m|^2}{2^{R_m}-1} - 1\right\},$$

where $R_m$ is the rate target for $U_m$. If $\beta \rho_n |h_n|^2 \le \tau_m$, $U_m$ is decoded first; otherwise, power adaptation is applied so that the received interference equals $\tau_m$, and the better rate of two possible SIC orderings is selected (Sun et al., 2024, Ning et al., 13 Jul 2025).

3. Achievable Rate and Energy Efficiency Analysis

The total H-NOMA rate for an opportunistic user is the sum of achievable rates in the NOMA slot (with its pair) and in its dedicated slot, given by

$R_{\text{H-NOMA}}(n) = R_{n,1}(h_m,h_n) + R_{n,2},$

with $R_{n,1}$ set by the HSIC scheme and $R_{n,2} = T \log_2(1 + \beta \rho_n |h_n|^2)$. The performance of H-NOMA is benchmarked by the “underperformance” probability: $$P_n = \Pr\{ R_{\text{H-NOMA}}(n) \le R_{\text{OMA}}(n) \},$$ whose exact closed-form is derived for each SIC variant. Tables of analytical expressions are presented in (Sun et al., 2024, Wang et al., 18 Sep 2025), with joint fading PDFs handled via order-statistics methods.

Crucially, under HSIC-PA, asymptotic analysis shows that $P_n \to 0$ for all user pairings, SNR ratios, and target rates as SNR increases. Thus,

$$\Pr\big\{ R_{\text{H-NOMA}}(n) > R_{\text{OMA}}(n) \,\text{and}\, E_{\text{H-NOMA}}(n) < E_{\text{OMA}}(n) \big\} \longrightarrow 1,$$

whereas FSIC and non-adaptive HSIC schemes generally retain nonzero error floors unless their parameters are tightly restricted (Wang et al., 18 Sep 2025, Ning et al., 13 Jul 2025).

Energy efficiency is rigorously defined as throughput per unit energy: $$\text{EE}_{\text{H-NOMA}} = \frac{R_{\text{H-NOMA}}(n)}{2 T \beta \rho_n}, \quad \text{EE}_{\text{OMA}} = \frac{R_{\text{OMA}}(n)}{T \rho_n}.$$ HSIC-PA ensures that, almost surely at high SNR, $\text{EE}_{\text{H-NOMA}} > \text{EE}_{\text{OMA}}$ (Sun et al., 2024, Ning et al., 13 Jul 2025).

4. Heterogeneous Service Slicing via H-NOMA: Advanced Scenarios

H-NOMA has been extended from pair-wise user setup to the multidimensional slicing of 5G services such as eMBB, URLLC, and mMTC. Advanced H-NOMA frameworks utilize message splitting (as in RSMA), space–frequency diversity, and resource grouping to build slices with custom reliability, latency, or massive connectivity constraints (Liu et al., 2022, Popovski et al., 2018). In such settings:

• eMBB–URLLC slicing: URLLC users are prioritized in SIC, leveraging ultra-reliability by decoding and canceling before eMBB. The optimal rate region is achieved by tuning power split factors and SIC order for the required reliability targets (Liu et al., 2022, Tominaga et al., 2021, Popovski et al., 2018).
• eMBB–mMTC slicing: mMTC users with strong channels can be decoded first, facilitating almost linear scaling of mMTC connectivity with the number of spatial degrees of freedom at the BS, particularly under multi-antenna reception (Tominaga et al., 2021). BS-side Maximum Ratio Combining combined with H-NOMA enables significantly higher mMTC loads at moderate eMBB rates.
• Multi-layer slicing: Rate-splitting and message splitting extend H-NOMA to strictly enlarge the MAC achievable rate/outage regions compared to OMA/NOMA, with the ability to trace the convex hull of capacity boundaries via adaptive $\beta$ and splitting (Liu et al., 2022).

5. Downlink H-NOMA and Extension to Multi-Antenna/Multi-Beam Systems

Downlink H-NOMA applies the same joint OMA/NOMA principles, but with key structural differences:

• In SISO settings with ordered channel gains and equal slot durations, downlink H-NOMA is analytically shown to always outperform TDMA in terms of sum throughput for the same energy budget. This is established via convex optimization and closed-form solutions for multi-slot power allocation, guaranteeing that each user transmits in every slot up to its own slot, with strictly less total energy than in OMA (Ding et al., 2024).
• Practical extension to MISO and near-field domains demonstrates that, by superposing additional streams on preconfigured SDMA beams, H-NOMA outperforms space-division OMA especially at high rate targets or sharp angular resolution, preserving performance when OMA fails due to beanforming singularities (Ding et al., 2024).

6. Numerical Validation and Practical Design Insights

Comprehensive simulation studies confirm analytical findings:

• Error probabilities for H-NOMA under HSIC-PA drop at high SNR, converging to zero regardless of user pairings or SNR ratios, while FSIC and HSIC-NPA may retain error floors (Sun et al., 2024, Wang et al., 18 Sep 2025).
• Energy savings over OMA and non-adaptive hybrid NOMA reach 20–60% under relevant SNR/rate configurations, depending on network size, number of antennas, and pairing policies (Ding et al., 2024, Saggese et al., 2021).
• Sensitivity analysis shows that larger $\beta$ and smaller legacy-user rates $R_m$ generally benefit the opportunistic user's gain probability, revealing a trade-off between reliability guarantees for OMA users and H-NOMA exploitation. Multiuser and multi-slice optimization of resource allocation and SIC-order appears essential for practical deployments (Wang et al., 18 Sep 2025, Liu et al., 2022).

For network slicing design, best practices include:

• Enforcing stringent reliability and latency for high-priority slices (URLLC), always decoding such slices first in SIC.
• Applying resource and power adaptation (e.g., per-slot or per-user $\beta$ setting) to optimize aggregate rate and energy efficiency under per-slice constraints.
• In multi-antenna settings, leveraging space diversity to expand feasible slicing regions.
• For small or moderate slice cardinality, explicit message splitting and scheduling can be optimized offline; at larger scale, adaptive online scheduling is favored (Liu et al., 2022, Popovski et al., 2018).

7. Limitations, Assumptions, and Outlook

The existing analyses largely rely on idealized assumptions: perfect SIC, perfect CSI (at least at the BS or for scheduled users), and asymptotically large blocklengths. Finite blocklengths, imperfect SIC, or channel estimation errors may degrade performance, but the fundamental error floor removal property of HSIC-PA is robust to such impairments in high SNR (Wang et al., 18 Sep 2025). The compatibility of H-NOMA as a backward-compatible add-on to legacy OMA networks is established, with only minor overhead in control signaling for power-split factors and SIC order (Ding et al., 2024). Implementation complexity is dominated by BS-side SIC and power control adaptation, both tractable with state-of-the-art hardware.

Open questions include extending H-NOMA beyond two-user slices to higher-cardinality or fully generalized multi-slice networks, incorporating random access dynamics, and developing low-complexity optimization algorithms for large-scale 6G systems.