Constructive Multiple Access (CoMA) Overview

• Constructive Multiple Access (CoMA) is a paradigm where transmitter-side waveform and code design forces signal alignment, mitigating interference and enabling direct function computation.
• It employs unique steering vectors and subspace-based decoding (e.g., root-MUSIC) to resolve collisions without the need for successive interference cancellation.
• The approach achieves near-optimal throughput and reduced detection complexity in multi-antenna settings, with proven gains in SER and power efficiency over traditional NOMA/OMA.

Constructive Multiple Access (CoMA) is a paradigm within multiuser communications wherein transmitter-side waveform and code design force the sum of signals over a multiple access channel to align constructively at the receiver, enabling interference-free recovery of transmitted messages or direct computation of desired functions. Unlike classical schemes that treat cross-user interference as noise or require successive interference cancellation (SIC), CoMA leverages deterministic structure—often via orthogonal subspaces, unique steering vectors, or joint code and constellation design—so that signal superposition enables, rather than hinders, throughput, reliability, and computation.

1. Mathematical Model and Signal Design

CoMA encompasses both many-to-one and many-to-many topologies, with each transmitter assigned structured parameters to guarantee unique recoverability. For a time-slotted system with $R$ receivers, let $A_j$ be the (possibly unknown) active transmitter set at receiver $j$. The essential construct is a transmitter-specific steering vector, assigned permanently as $r_i = e^{j\theta_i}$, $0 \leq \theta_i < \pi$ and known to each receiver. Over $N$ time slots, transmitter $i$ applies powers of $r_i$ to its packet, producing the steering vector $w_{i,N} = [r_i^0,\ldots, r_i^{N-1}]^T$ of dimension $N$.

The receiver observes a stacked $N \times P$ matrix $Y_j$:

$$Y_j = \sum_{i\in A_j} h_{ji} w_{i,N} s_i^{(0) T} + N_j = W_{j,N} S_j + N_j$$

where $h_{ji}$ is the local complex channel coefficient and $s_i^{(0)}$ is the $P$-symbol payload. The Vandermonde structure renders columns of $W_{j,N}$ linearly independent for $N \geq |A_j|$, confining each packet to a unique subspace.

Beyond slot-synchronous packetized settings, CoMA supports digital functional computation in multiaccess (MAC) channels with arbitrary constellation and code assignment. Consider the ChannelComp/CoMA signal model for multiple transmitters computing $f(x_1,\ldots,x_K)$ at a fusion center, with quantized symbols $\tilde{x}_k$, nonlinear digital constellation encoders $\mathcal{E}_k(\cdot)$, and repetition-code slots:

$$y_\ell = \sum_{k=1}^K h_{k,\ell} p_{k,\ell} x_k c_{k,\ell} + z_\ell$$

with $c_{k,\ell}\in\{0,1\}$ performing symbol repetition and selection, and channel inversion enabling additive superposition.

In multi-antenna NOMA-type settings, CoMA realizes constructive superposition by precoder design: the transmit symbol $x = w_1 s_1 + w_2 s_2$ is phase-rotated such that, for the strong user, the combined signal lies within the desired decision region for M-PSK, eliminating the need for SIC.

2. Collision Resolution and Subspace-Based Decoding

CoMA's signature collision resolution is subspace-based. Upon reception, the receiver performs:

1. Rank detection: The receiver collects slots until $\operatorname{rank}(Y_j)$ first drops to $K_j$, thus inferring the number of colliding packets.
2. Steering vector recovery via root-MUSIC: The MUSIC pseudospectrum

$J_j(z) = w'(z)^H U_{j,\perp} U_{j,\perp}^H w'(z)$

is evaluated for $w'(z) = [1, z, z^2, \ldots, z^{N-1}]^T$; zeros on the unit circle pinpoint active steering vectors $r_i$.

1. Packet recovery by subspace projection: Signals are projected onto the estimated signal subspace, and payloads $\hat{S}_j$ are recovered by least squares.

The superposition is resolved without retransmissions or receiver-side SIC, provided slot-level synchronization and channel coherence.

3. Constructive Interference Exploitation and Practical CoMA Precoding

In MISO Downlink settings, CoMA alters the paradigm of NOMA by exploiting constructive interference via precoding. The signal is rotated so that interference from an unintended user is mapped into the correct detection region of the strong user; this is formalized by imposing constructive interference constraints (CI) on the real and imaginary components of $h_1^T x$ for user 1:

$$|\operatorname{Im}\{h_1^T \tilde{x}\}| \leq \left( \operatorname{Re}\{h_1^T \tilde{x}\} - \sqrt{\gamma_1} \sigma_1 \right) \tan \theta$$

The transmit optimization targets either minimization of total power under QoS, or minimization of symbol error rate (SER) under power constraints, with resulting convex or iterative optimization steps.

This constructive alignment allows both users' symbols to be detected directly, removing the SIC stage and substantially reducing computational load and power expenditure when compared to standard NOMA.

4. Throughput, Efficiency, and Performance Analysis

The theoretical capacity of CoMA under idealized conditions exhibits full multiplexing efficiency. When synchronization and channel coherence are perfect:

$\lim_{K_j \to \infty} \frac{K_j}{N} \to 1$

As each transmitter's packet resides in a unique subspace, each additional transmitter increases subspace dimension, approaching 100% throughput. For slot-misaligned transmitters, each occupies two subspaces. In worst-case, as $K_j \to \infty$, throughput remains at least 50%, exceeding classical tree-splitting and interference alignment bounds.

In practical digital compute-MAC scenarios, jointly optimizing the digital constellation and repetition-based channel code achieves up to 30 dB reduction in computation error (normalized mean squared error) for nonlinear functions such as products, compared to standard approaches without joint design. This gain is realized at moderate SNR and with manageable codeword and lookup-table complexity (Yan et al., 2024).

SER performance in precoded MISO CoMA exceeds that of NOMA/OMA at moderate to high SNR, with 1e-3 SER achieved at 8 dB for QPSK with $N=4$ antennas in CoMA versus $>10$ dB in competing methods. The complexity of detection is halved relative to NOMA at typical constellation sizes (Salem et al., 2022).

5. Implementation, Scalability, and Limitations

Key requirements for operational CoMA include:

• Slot-level synchronization: Required for subspace uniqueness; achievable with beaconing or GPS-based timing for small clusters.
• High SNR regime: Necessary for sharp separation between signal and noise subspaces; at lower SNR, the slot overhead ($N-K_j$) must grow, incurring latency.
• Local CSI: Needed only at the receiver for weighting the subspace decoding, not at the transmitter or across receivers.
• Computational complexity: Per receiver, complexity is dominated by an SVD of size $N \times P$ (with $P \gtrsim K_j$) and root-finding for root-MUSIC, scaling linearly in the number of colliding users.
• Scalability: The receiver evaluates up to an $N$-degree polynomial for root-MUSIC; empirical results indicate only a constant slot overhead is typically required.
• Extensions: CoMA is compatible with many-to-many topologies; each receiver performs subspace decoding independently without explicit coordination.

Repetition-code optimized CoMA ("ReChCompCode") for function computation introduces tractable alternating optimization (SDP for constellation, MIP for timing code), with convergence to stationarity under mild conditions (Yan et al., 2024).

6. Functional Computation via Constructive Multi-Access and Future Directions

CoMA enables over-the-air computation by jointly shaping waveform, constellation, and codebook so that the MAC channel output directly computes a (possibly nonlinear) function of user data. Recent advances such as ChannelComp/CoMA with repetition-code channel coding facilitate reliable digital computation of functions such as sum, max, and products with a lookup-table decoder. Channel coding and constellation points are optimized to maximize pairwise separation of target function outputs, subject to power/distance constraints, yielding robust compute performance under moderate SNR (Yan et al., 2024).

A notable trajectory is integrating digital modulation/coding design and structured functional mapping, bridging classical digital communication and in-the-air computation. Plausible implications are the expansion of CoMA architectures to MIMO, adoption of algebraically-structured codebooks for complexity reduction, and analytical characterization of trade-offs between spectral efficiency, reliability, and computational complexity.

7. Comparative Summary and Impact

CoMA provides a structured, generalizable principle for managing multiaccess interference and collision by embedding transmitter identities or data in orthogonal or constructive base spaces, resolvable at the receiver either via subspace algorithms (root-MUSIC) or via constructively-aligned multiuser precoding. Its defining properties are:

Feature	Classical MAC/NOMA	Constructive Multiple Access (CoMA)
Interference Management	SIC, Collision Avoidance	Subspace/CI alignment, direct decoding
Throughput (large K)	≤ 50%	50–100% (worst case–ideal)
Receiver Complexity	High (SIC, ML search)	No SIC, ML or Table-Lookup only
Functional Computation	Separate After Detection	Over-the-air during transmission

CoMA thereby extends the achievable rate and functional flexibility of MAC systems, improving power, latency, and complexity. Future research directions include full integration with federated learning, non-orthogonal compute paradigms, extension to fading/mobility, and deployment in dense device networks (Akl et al., 2017, Salem et al., 2022, Yan et al., 2024).