---
title: 'Hybrid NOMA: Merging OMA and NOMA Access'
url: https://www.emergentmind.com/topics/hybrid-non-orthogonal-multiple-access-h-noma
type: topic
---

# Hybrid NOMA: Merging OMA and NOMA Access

Hybrid non-orthogonal multiple access (H-NOMA) most commonly denotes a multiple-access framework that “organically combines pure NOMA and conventional OMA,” or a design in which users can transmit in both OMA and NOMA modes through flexible resource allocation while retaining compatibility with legacy OMA-based networks [2408.14072], [2507.09458]. In a large part of the recent literature, H-NOMA is implemented as an add-on to TDMA or OFDMA, so that selected users obtain extra transmission opportunities by sharing resources that would be inaccessible under pure OMA. At the same time, other papers use closely related “hybrid” formulations to denote joint use of power-domain and code-domain NOMA, hybrid VLC-RF systems, hybrid beamforming with NOMA, or hybrid CSI-based power allocation. Collectively, this suggests that H-NOMA is best understood as a family of resource-sharing and decoding architectures rather than a single protocol [2012.08106].

## 1. Conceptual scope and historical development

A foundational formulation appears in the study of multiple-access channels with non-ideal batteries and circuit operating costs, where users may need to adopt a **hybrid NOMA-TDMA strategy** that “combines the features of NOMA and TDMA” by allocating fixed time windows for orthogonal single-user and non-orthogonal joint transmissions [1801.03794]. In that setting, pure TDMA is not always sum-rate optimal when battery internal resistance is non-zero, and the achievable rate region of the hybrid scheme contains the achievable regions of pure NOMA and pure TDMA [1801.03794]. This line of work established an early systems-level rationale for hybridization: orthogonal and non-orthogonal phases can balance different sources of inefficiency.

A broader radio-access interpretation was later articulated for future uplink massive access, where HNOMA includes both **power-domain NOMA** and **code-domain NOMA** because practical channel conditions are diverse [2012.08106]. In that view, clustered users share orthogonal resources through both power separation and codebook separation, and the base station uses a combination of **message passing algorithm (MPA)** and **successive interference cancellation (SIC)**. The same survey emphasizes imperfect CSI, polar-coded transmission, and deep-learning-based adaptive design as central design dimensions for HNOMA-based wireless networks [2012.08106].

Terminology is not entirely uniform. In cooperative relay networks, “hybrid-NOMA” is used for a **hybrid power allocation strategy** that combines statistical CSI for one hop and instantaneous CSI for another, reducing computational complexity and signaling overhead with marginal sum-rate degradation [1801.04650]. This usage differs from the OMA-plus-NOMA add-on model, but it reinforces a common theme: the term “hybrid” typically indicates a deliberate combination of otherwise separate access, signaling, or optimization mechanisms.

## 2. Common uplink H-NOMA formulation in legacy TDMA systems

A common recent formulation considers an **uplink TDMA-based legacy network** with \(M\) users, where each user \(U_n\) has its own slot and an **opportunistic user** may also transmit in the slot of a paired **legacy user** \(U_m\) [2408.14072]. The opportunistic user therefore communicates in two phases: an OMA phase in its own slot and a NOMA phase in the legacy user’s slot. This is the dominant architectural template in the recent energy-efficiency literature.

For the opportunistic user \(U_n\), the OMA benchmark rate is given by
\[
R_n=\log_2(1+\rho_n|h_n|^2),
\]
while the reduced-power OMA-phase rate in H-NOMA is
\[
R_n^{\text{OMA}}=\log_2(1+\beta\rho_n|h_n|^2),
\]
with \(0<\beta<1/2\) introduced so that the hybrid transmission consumes no more energy than conventional OMA [2507.09458]. The total H-NOMA rate is then written as
\[
R_n^{\text{H-NOMA}}=\hat R_n^{\text{NOMA}}+R_n^{\text{OMA}}.
\]
This construction is explicitly motivated by backward compatibility: NOMA is superimposed on top of a legacy slot structure rather than replacing it [2507.09458].

The basic operational constraint is transparency to the legacy user. In the NOMA phase, the opportunistic user’s transmission must not worsen the legacy user’s outage behavior relative to OMA; accordingly, its power and achievable rate are shaped by the legacy user’s target rate and by the interference threshold \(\tau_m\) [2408.14072]. User grouping is typically represented by ordered fading gains, e.g. \(|h_1|^2<\cdots<|h_M|^2\), and several papers stress that pairing and the order index of the opportunistic user strongly affect performance [2507.09458].

A closely related two-user formulation studies a QoS-user \(U_m\) and a best-effort user \(U_n\) in uplink TDMA, where the best-effort user transmits with power \(\beta_1\rho_n\) in \(U_m\)’s slot and \(\beta_2\rho_n\) in its own slot, subject to \(\beta_1+\beta_2\le \eta\) [2411.01776]. The corresponding optimization problem maximizes the best-effort user’s data rate under an energy budget and an interference constraint tied to
\[
\tau_m=\max\left\{0,\frac{\rho_m|h_m|^2}{\epsilon_0}-1\right\},\qquad \epsilon_0=2^{R_0}-1.
\]
Closed-form solutions for the optimal power split \((\beta_1^\*,\beta_2^\*)\) show that the system falls back to OMA when \(\tau_m=0\), and otherwise allocates power across both slots [2411.01776].

## 3. SIC policies, power adaptation, and asymptotic behavior

A central research theme in H-NOMA is the design of the SIC rule in the NOMA phase. Three variants recur in the recent literature: **fixed-order SIC (FSIC)**, **hybrid SIC without power adaptation (HSIC-NPA)**, and **hybrid SIC with power adaptation (HSIC-PA)** [2509.14809]. Their distinguishing feature is the degree of adaptation to instantaneous channel conditions and interference constraints.

Under FSIC, the decoding order is fixed regardless of the realized channel-gain ordering. Under HSIC-NPA, the SIC order changes with the received-power regime, but the opportunistic user’s power remains fixed. Under HSIC-PA, both the SIC order and the opportunistic user’s power are dynamically adapted; in the high-interference regime, the opportunistic user can reduce its power by a factor \(\gamma\) so that the legacy user’s rate protection constraint remains satisfied, and then select the better of two candidate decoding options [2507.09458], [2509.14809].

The principal instantaneous performance metric is the probability that H-NOMA fails to outperform OMA while using less energy. In the HSIC-PA formulation, this is
\[
\hat P_n=\Pr\!\left[T\hat R_n^{\text{NOMA}}+TR_n^{\text{OMA}}\le TR_n\right].
\]
The recent literature derives closed-form expressions for this probability and then studies its high-SNR limit. The key asymptotic distinction is sharp. For FSIC, a nonzero performance floor remains at high SNR. For HSIC-NPA, the floor disappears only under parameter-dependent conditions, such as \(\epsilon_m\le \beta/(1-\beta)\), where \(\epsilon_m=2^{R_m}-1\) [2509.14809]. By contrast, for HSIC-PA,
\[
\lim_{\rho_m,\rho_n\rightarrow\infty}\hat P_n=0,
\]
and one paper states that this holds for **all choices of target rates, channel gains, and power ratios** [2507.09458].

This is the main theoretical argument for power adaptation in the H-NOMA NOMA phase. The 2025 HSIC-PA design further shows that the probability of H-NOMA outperforming OMA approaches one in the high-SNR regime **without any constraints on either users’ target rates or transmit power ratios**, which is presented as a significant improvement over conventional H-NOMA schemes that require restrictive conditions to achieve probability one at high SNRs [2507.09458]. The same work reports that the decay rate depends strongly on the opportunistic user’s order \(n\), with \(\hat P_n\propto 1/\rho_m^n\) or \(1/\rho_n^n\), so pairing a higher-order user as opportunistic yields faster asymptotic improvement [2507.09458].

An earlier HSIC-aided scheme without power adaptation already relaxed the conditions under which H-NOMA almost surely beats OMA, but not completely. Its asymptotic analysis shows that \(\tilde P_n\rightarrow 0\) only in “relaxed” high-SNR regimes, whereas if only one user’s power increases and the other remains fixed, \(\tilde P_n\) approaches a non-zero constant [2408.14072]. This distinction underlies a common misconception: H-NOMA does not automatically eliminate the error floor; that depends on the SIC design and on whether power adaptation is available.

## 4. Resource-domain and direction-specific generalizations

H-NOMA has also been generalized beyond TDMA. Ding and Poor study **hybrid NOMA assisted OFDMA uplink transmission**, where dynamic per-subcarrier CSI enables users to exploit frequency diversity in a way that is not available in the same form in hybrid TDMA [2407.03899]. The power-minimization problem is formulated as
\[
\min_{\{P_{m,n}\ge 0\}} \sum_{m=1}^M\sum_{n=1}^m P_{m,n}
\quad\text{s.t.}\quad
\sum_{n=1}^m R_{m,n}\ge R,\ \forall m,
\]
with
\[
R_{m,n}=\log\left(1+\frac{h_{m,n}P_{m,n}}{\sum_{j=n}^{m-1}h_{j,n}P_{j,n}+1}\right).
\]
From the optimization perspective, pure OMA is “rarely optimal,” and the probability that pure OMA is optimal tends to zero as the number of users \(M\to\infty\). Unlike the hybrid TDMA case, pure NOMA can be optimal for minimizing user energy consumption in the OFDMA setting [2407.03899].

The statistical analysis of H-NOMA-OFDMA introduces the **power outage probability**
\[
\mathbb{P}_m^{\rm out}=\mathbb{P}(P_m\ge \rho)
\]
and the **power diversity gain**
\[
d=-\lim_{\rho\to\infty}\frac{\log \mathbb{P}_m^{\rm out}}{\log \rho}.
\]
For the most constrained user, the diversity gain can increase from \(1\) under OMA to \(M\) under hybrid NOMA cooperation [2407.03899]. This result attributes the gain to dynamic CSI across subcarriers, which allows opportunistic migration of power to favorable subcarriers.

Downlink H-NOMA follows a different pattern. A downlink SISO design proposes H-NOMA as an add-on to legacy TDMA and shows analytically that, when users’ channel gains are ordered and time-slot durations are equal, **downlink hybrid NOMA always outperforms TDMA**, which is explicitly contrasted with existing conclusions for uplink hybrid NOMA [2401.16965]. The same paper extends the design to a MISO near-field setting, where users in one group share beams configured for another group and the non-convex energy-minimization problem is handled by **successive convex approximation (SCA)** [2401.16965].

A different question is whether the conventional asymmetry of H-NOMA is itself suboptimal. In a simple two-user study where both users are allowed equal access to both time slots, analytical and simulation results show that **conventional hybrid NOMA is still an optimal transmission strategy**; even when equal access is permitted, the user who originally had only its own slot optimally continues pure OMA behavior [2409.09654]. This directly counters the intuition that symmetric resource access must improve performance in the basic two-user case.

## 5. Backscatter, heterogeneous media, and network-level H-NOMA

A substantial application area is **BackCom assisted H-NOMA uplink transmission** for ambient IoT. In this model, a user transmits its own information in its slot while other users transmit via backscatter using that signal as the carrier [2403.16498]. For \(M\) users, the overall power-minimization problem is
\[
\min_{\{P_m,\eta_{mi}\ge 0\}} \sum_{m=1}^{M} P_m
\quad\text{s.t.}\quad
\sum_{i=1}^m R_{mi}T\ge N,\quad 0\le \eta_{mi}\le 1.
\]
The two-user analysis shows that **pure OMA is never optimal** in BackCom-assisted H-NOMA, and that **pure NOMA can be optimal** under certain channel conditions, which is explicitly contrasted with conventional H-NOMA [2403.16498]. For the general multi-user case, the paper develops both **branch-and-bound (BB)** and **successive convex approximation (SCA)** algorithms, trading off optimality and complexity [2403.16498].

A robust extension addresses imperfect CSI through generalized uncertainty sets and worst-case robust optimization. The resulting problem is conservative, non-convex, and intractable in its original form, so the analysis uses **Lagrange duality**, **majorization-minimization**, slack variables, and a penalized objective to obtain a tractable approximation with a provably convergent polynomial-complexity algorithm [2505.07762]. Despite conservatism, the robust solution yields similar power consumption to the nominal imperfect-uncertainty-free problem, robust H-NOMA “almost always” gives more power efficiency than OMA, and its probability of feasibility is higher than that of OMA and the nominal design [2505.07762].

H-NOMA has also been studied in **hybrid VLC-RF networks with imperfect CSI**, where both the VLC and RF subsystems use NOMA [2005.03744]. The total hybrid sum rate is expressed as
\[
R_{\mathrm{SUM}}
=
B_{\mathrm{RF}}\big(\beta_{\mathrm{RF}}\hat R_{\mathrm{RF}}^{\mathrm{NOMA}}\big)
+
B_{\mathrm{VLC}}\big(\beta_{\mathrm{VLC}}\hat R_{\mathrm{VLC}}^{\mathrm{NOMA}}\big),
\]
and the energy efficiency is
\[
\hat \xi=
\frac{
B_{\mathrm{RF}}(\beta_{\mathrm{RF}}\hat R_{\mathrm{RF}}^{\mathrm{NOMA}})
+
B_{\mathrm{VLC}}(\beta_{\mathrm{VLC}}\hat R_{\mathrm{VLC}}^{\mathrm{NOMA}})
}{
Q_{\mathrm{VLC}}+Q_{\mathrm{RF}}+\sum_{i=1}^{N}P_{\mathrm{RF},i}
}.
\]
This work shows that CSI errors have a considerable impact on the average energy efficiency of both NOMA-VLC and OFDMA-VLC systems, and that the hybrid NOMA-VLC-RF system is more robust than standalone NOMA-VLC under imperfect CSI and adverse LOS conditions [2005.03744].

At the network level, one heterogeneous-network formulation combines **NOMA in small cells** with **massive MIMO in macro cells**. Coverage probability, spectrum efficiency, and energy efficiency are analyzed in a \(K\)-tier hybrid HetNet, and the results show that the proposed NOMA-enhanced HetNet outperforms OMA-based HetNets, while dense deployment of NOMA-enhanced small-cell BSs can greatly improve whole-network energy efficiency [1705.03325]. H-NOMA has also been extended to mixed semantic and bit-user communication, where slot-level switching between NOMA and OMA supports heterogeneous user types and uses **bit-to-semantic decoding order** in mixed-user slots [2505.03379].

## 6. Hybrid-domain implementations, beam-space designs, and research issues

In another strand of the literature, H-NOMA means a **hybrid-domain** design that combines code-domain and power-domain NOMA. One uplink proposal divides users into strong and weak groups, applies SCMA within each group, then uses SIC across groups and MPA within groups [2007.09179]. The spectral-efficiency maximization problem is non-convex and sparse, so it is solved by **alternating optimization**, **successive convex approximation**, and **reweighted \(\ell_1\) minimization**. The reported outcome is that HD-NOMA supports more uplink users than conventional SCMA while maintaining acceptable BER and higher sum rate [2007.09179].

A related mm-wave hybrid massive-MIMO architecture uses user grouping and hybrid beamforming so that NOMA becomes an **intra-group process**. In that framework, SCMA and MUSA receivers can be directly adapted, and a PIC-aided uplink MUSA receiver reduces complexity relative to MPA under high overloading [2007.04056]. Achievable information rate lower bounds are used as the main performance measure, and code-domain NOMA is shown to outperform conventional methods when RF chains are limited and users are spatially close [2007.04056].

Hybrid beamforming also appears in **HB-NOMA** for mmWave systems, where analog and digital beamformers serve user clusters and NOMA is applied within each cluster [1804.05444]. A lower bound derived under imperfect correlation shows that an inefficient mobile-user cluster can cause severe inter-cluster interference. This is a useful corrective to a simplified view of H-NOMA: the gain does not come from non-orthogonal superposition alone, but also from effective clustering and correlation management in beam space [1804.05444].

Across these variants, a recurring set of technical limitations is repeatedly identified. Imperfect CSI, inter-group or inter-cluster interference, SIC error propagation, increasing receiver complexity, and the difficulty of dynamic grouping under mobility are treated as central obstacles to practical deployment [2012.08106]. A plausible implication is that future H-NOMA research will remain bifurcated between two tasks: first, finding architectures in which hybrid access yields structural advantages over OMA and pure NOMA; and second, controlling the resulting complexity, CSI sensitivity, and fairness trade-offs well enough for those advantages to survive realistic channel uncertainty and heterogeneous traffic conditions.

Source: https://www.emergentmind.com/topics/hybrid-non-orthogonal-multiple-access-h-noma