Non-orthogonal multiple access (NOMA) promises to fulfill the fast-growing connectivities in future Internet of Things (IoT) using abundant multiple-access signatures. While explicitly notifying the utilized NOMA signatures causes large signaling cost, blind signature classification naturally becomes a low-cost option. To accomplish signature classification for NOMA, we study both likelihood- and feature-based methods. A likelihood-based method is firstly proposed and showed to be optimal in the asymptotic limit of the observations, despite high computational complexity. While feature-based classification methods promise low complexity, efficient features are non-trivial to be manually designed. To this end, we resort to artificial intelligence (AI) for deep learning-based automatic feature extraction. Specifically, our proposed deep neural network for signature classification, namely DeepClassifier, establishes on the insights gained from the likelihood-based method, which contains two stages to respectively deal with a single observation and aggregate the classification results of an observation sequence. The first stage utilizes an iterative structure where each layer employs a memory-extended network to explicitly exploit the knowledge of signature pool. The second stage incorporates the straight-through channels within a deep recurrent structure to avoid information loss of previous observations. Experiments show that DeepClassifier approaches the optimal likelihood-based method with a reduction of 90% complexity.

The integration of non-terrestrial networks (NTNs) in 6G enables seamless connectivity and unleashes the potential of stringent Internet of Things (IoT) services worldwide. Among these, safety-critical IoT aims to provide ubiquitous service along with comprehensive protection against diversified over-the-air threats, which requires high-security technology under the unique constraints of NTNs. As a promising secure transmission technology rooted in information theory, wireless covert communication can conceal the presence of transmission so as to eliminate possible attack threats. In this article, we point out that the continuous evolution of covert communication paves a comprehensive avenue to tackle the complicated security challenges of safety-critical IoT in 6G NTN. We elucidate several typical security threats that may be encountered in NTN IoT and analyze how covert communication can resist these security vulnerabilities through its endogenous potential. Further, we propose an artificial intelligence-based three-participant collaborative-adversarial network to strike a balance between communication efficiency and covertness performance. Finally, some future research directions and challenges of covert communication are discussed.

Artificial intelligence (AI) has become a promising solution for meeting the stringent performance requirements on wireless physical layer in sixth-generation (6G) communication systems, due to its strong ability to learn complex model, achieve end-to-end optimization and adapt to dynamic environments. This article provides a comprehensive review with respect to artificial intelligence for wireless physical-layer technologies (AI4PHY). Specifically, we first analyze the characteristics of the classic AI techniques and their potential applications for physical-layer technologies. Then we study the AI-enhanced designs from the point of view of the basic physical-layer modules, including coding, modulation, multiple access, multiple-input-multiple-output (MIMO), channel estimation, as well as relay transmission. The standardization progress of AI4PHY in 3GPP is also discussed. Based on the current AI4PHY researches, we propose some potential future research directions to inspire and encourage the further exploration.

Mega constellation, as an extremely large-scale radio access network, faces severe multi-user interference when accommodating ubiquitous access. Distributed multi-user detection (MUD) can utilize the multi-satellite spatial diversities and processing capabilities to alleviate inter-user interference. However, the spaceborne nature makes it seriously chained by inter-satellite link (ISL) constraints including the limited number of ISL ports per satellite and the constrained bandwidth per ISL port. To this end, this paper proposes an efficient message passing (MP) based distributed MUD framework under stringent ISL constraints. First, the overheads introduced by the optimal fully-connected distributed MUD, including required ISL ports and bandwidth, are quantitatively characterized by employing distributed factor graph (FG) model and MP mechanism. By analyzing the FG representation, we propose two ISL-compatible design principles for distributed MUD, i.e., orchestrating message flow (MF) hierarchically among satellites to save ISL ports, and propagating messages selectively to save ISL bandwidth. Specifically, a novel multi-branch tree-like MF orchestration is proposed to forward and aggregate the locally generated detection messages in a partially-connected manner under limited ISL port number. The relationship between MF structure and overall performance is revealed via EXIT chart and a fairness-aware orchestration algorithm is developed. Further, we introduce a novel squeeze node into the distributed FG, compressing messages and facilitating selective MP under bandwidth constraint. Three criteria are correspondingly proposed to identify the most effective messages for distributed MUD. The proposed framework is then evaluated under practical simulation settings with satellite orbits and propagation environments generated via STK. Simulation results demonstrate that our framework reduces ISL costs by 50% with less than 1 dB loss in terms of BER performance compared to the fully-connected MUD, and achieves up to 5 dB gain in BER over the state-of-the-art distributed reception methods under fair comparisons.

Terminal positioning using a single low earth orbit (LEO) communication satellite is an important means to compensate for the loss of positioning capability under global navigation satellite system (GNSS) denial. Traditional single-satellite positioning relies on the accumulative measurement over a segment of the trajectory, resulting in a long response time. In this paper, we propose a novel single-satellite positioning scheme, achieving instant positioning through the joint use of time-frequency Doppler analysis of signals and communication beam switching. First, through wide beam reception, range and velocity information are extracted using interpolation enhanced parameter estimation to determine candidate coordinates. To eliminate ambiguity in this process, further disambiguation is achieved through the high directivity of the narrow beam. The proposed scheme is proved to reach the Cramér-Rao lower bound (CRLB) while achieving real-time positioning with the central limit theorem (CLT). Simulation results indicate that at a signal-to-noise ratio (SNR) of 17 dB, the probability of positioning error being lower than 50 meters can reach 97.5%, demonstrating performance comparable to existing schemes.

Millimeter wave (mmWave) multiple-input-multiple-output (MIMO) with hybrid beamforming enables high data rate broadband services for 5G. To further improve the spectral efficiency and connectivity, non-orthogonal multiple access (NOMA) has been considered to be integrated with mmWave MIMO. Nonetheless, the existing power-domain mmWave MIMO-NOMA with narrow analog beams suffers from beam misalignments which deteriorate the data rates of the misaligned users as well as the system fairness. Thereby, we propose a beam aggregation-based mmWave MIMO-NOMA scheme to loosen the requirement of beam alignment and improve the system fairness. The proposed scheme generates virtual beams with wider beamwidth by aggregating adjacent analog beams. NOMA transmissions are utilized within each aggregated virtual beam. Then, we propose a non-orthogonal multiuser precoding scheme to ensure the fairness within the aggregated beam by maximizing the minimum achievable rates of the grouped users. To address this issue, a max-min problem is proposed and artificial intelligence (AI) technology is exploited to solve this non-trivial problem. The problem is firstly converted to an equivalent penalized minimization problem and then an unsupervised deep neural network (DNN) is trained to map the instantaneous channel coefficients to the precoders in a data-driven fashion. Specifically, we explicitly introduce the transmit power as the DNN input to achieve better generalization in different signal-to-noise ratio regions. Performance evaluations reveal that the proposed scheme can achieve significant gain on the max-min data rate compared with the conventional scheme.