Artificial intelligence (AI)/machine learning (ML) models in 5G and beyond networks are actively under standardization work in progress for deployment across distributed user equipment (UE) and network entities in wireless networks. Efficient deployment of ML models in wireless systems requires scalable mechanisms to manage model distribution, especially when there are multiple models embedded across UEs. This paper introduces a novel method for configuring model density grouping (MDG) to optimize model selection, signaling efficiency, and collaborative learning performance. By leveraging the density of similar models across UEs with model identification, our approach enhances the adaptability and convergence of ML operations via system information and dedicated radio protocol signaling. Our analysis shows dynamic grouping, density estimation, and signaling procedures that allow intelligent selection and activation of model sets with minimal accuracy tradeoffs and signaling overhead. Evaluation demonstrates the utility of MDGs in improving convergence efficiency and scalability in one/two-sided model environments.

We consider a novel anchorless rigid body localization (RBL) suitable for application in autonomous driving (AD), in so far as the algorithm enables a rigid body to egoistically detect the location (relative translation) and orientation (relative rotation) of another body, without knowledge of the shape of the latter, based only on a set of measurements of the distances between sensors of one vehicle to the other. A key point of the proposed method is that the translation vector between the two-bodies is modeled using the double-centering operator from multidimensional scaling (MDS) theory, enabling the method to be used between rigid bodies regardless of their shapes, in contrast to conventional approaches which require both bodies to have the same shape. Simulation results illustrate the good performance of the proposed technique in terms of root mean square error (RMSE) of the estimates in different setups.

We consider the design of low complexity and highperforming mean square error (MSE) minimization combiners for over-the-air-computing (AirComp) applications operating over the uplink of a system with one multiple-antenna access point (AP) and multiple single-antenna edge devices (EDs). Within that paradigm, we offer two contributions, namely, a simple initial combiner based on a Rayleigh quotient (RQ) design, and a low-complexity refinement stage based on a convex concave procedure (CCP). The new refinement stage algorithm is further enriched with an efficient (offline) hyper-parameter tuning mechanism via Bayesian optimization (BO) and acceleration method based on a half-space constrained least square problem reformulation solved via the adaptive moment estimation (Adam) algorithm. The low complexity and good performance of the proposed method help address typical limitations of edge devices. Numerical results demonstrate that the proposed design can achieve MSE performances equivalent to those of the best stateof-the-art (SotA) alternatives currently known, at about 200-times less complexity than the highest-performing SotA, and about 4-times less complexity than its low-complexity counterpart.

We propose a new two-stage, low-complexity, and robust beamforming (BF) method for heterogeneous MIMO rate splitting multiple access (RSMA) systems. In the proposed method, the phases and powers of the BF weights are designed separately, the first based on a tensor factorization of the channels between the base station (BS) and each user, and the second based on a fractional programming (FP) formulation of the power allocation problem, which is offered in three distinct variations, aimed as sum rate maximization (SRM), minimum rate maximization (MaxMin) and the maximization of the geometric-mean (GMean) of achievable rates, respectively. Thanks to the twostage approach, the proposed method is capable of delivering robustness to both channel state information (CSI) and successive interference cancellation (SIC) errors (incorporated in the phase design), at a low complexity compared to state-of-the-art (SotA) alternatives. Also thanks to the approach, the scheme naturally handles heterogeneity in terms of the number of antennas at each user, which can be arbitrarily distinct. Direct comparisons between SotA and the proposed schemes demonstrate that the contributed method generally outperforms the best alternative at comparable complexity, while approaching the best-performing SotA method of significantly higher complexity. In fact, the computational cost advantage of the proposed technique over the latter is quantified analytically and shown to be proportional to the cube of the number of BS antennas.

Index modulation (IM) is one of the key enabling technologies for beyond fifth generation (B5G) and sixth generation (6G) wireless systems, attracting attention for its inherent energy and spectral efficiency resulting from conveying information through the indexation of the resources utilized in during signal transmission. However, a remaining critical bottleneck for large-scale IM is the consequently infeasible detection complexity of combinatoric order. Therefore in this article, in order to maximally reap the advantages of IM in large scenarios, we propose a novel message passing (MP) decoder designed under the Gaussian belief propagation (GaBP) framework exploiting a novel unit vector decomposition (UVD) of IM signals with purpose-derived novel probability distributions. The proposed method enjoys a low decoding complexity that is independent of previously prohibitive combinatorial factors, while still approaching the performance of unfeasible state-of-the-art (SotA) search-based methods. The effectiveness of the proposed approach is demonstrated via complexity analysis and numerical results for the exemplary piloted generalized quadrature spatial modulation (GQSM) systems of truly massive sizes (up to 96 antennas).

We consider an uplink integrated sensing and communications (ISAC) scenario where the detection of data symbols from multiple user equipment (UEs) occurs simultaneously with a three-dimensional (3D) estimation of the environment, extracted from the scattering features present in the channel state information (CSI) and utilizing the same physical layer communications air interface, as opposed to radar technologies. By exploiting a discrete (voxelated) representation of the environment, two novel ISAC schemes are derived with purpose-built message passing (MP) rules for the joint estimation of data symbols and status (filled/empty) of the discretized environment. The first relies on a modular feedback structure in which the data symbols and the environment are estimated alternately, whereas the second leverages a bilinear inference framework to estimate both variables concurrently. Both contributed methods are shown via simulations to outperform the state-of-the-art (SotA) in accurately recovering the transmitted data as well as the 3D image of the environment. An analysis of the computational complexities of the proposed methods reveals distinct advantages of each scheme, namely, that the bilinear solution exhibits a superior robustness to short pilots and channel blockages, while the alternating solution offers lower complexity with large number of UEs and superior performance in ideal conditions.

In this paper, we propose a novel joint activity and channel estimation (JACE) algorithm for grant-free extra large MIMO (XL-MIMO) systems subject to spatial non-stationarity phenomena by means of a Bayesian bilinear inference framework. In XL-MIMO systems, the signal from each user is visible only by a small portion of its antenna arrays, which are typically distributed over the surface of a certain structure. The sporadic user activity due to grant-free access, as well as the spatial non-stationarity, jointly imposes a challenging JACE problem involving a nested Bernoulli-Gaussian random variable. In order to address this issue, we decompose the latter into a bilinear inference problem of two independent random quantities, deriving novel message passing rules based on Gaussian approximation and bilinear inference. Performance evaluation via software simulations is offered to demonstrate the effectiveness of the proposed algorithm, which achieves the Genie-aided ideal estimation performance.

Extra large MIMO (XL-MIMO) systems are subject to spatial non-stationarity forming visibility regions (VRs), which leads to a sub-array-wise sparse structure of the channel matrix. When XL-MIMO systems operate in grant-free access mode, in which only a fraction of the potential users are active during a given time slot, it follows that the channel matrix possesses a doubly-sparse and user-specific structure such that the activity of each user and each sub-array can be jointly modeled by a nested Bernoulli-Gaussian distribution. This article considers the joint activity and channel estimation (JACE) problem in XL-MIMO systems subject to this so-defined spatial non-stationarity, tackling this challenging inference problem. Our main contributions are 1) to introduce the novel Bernoulli-Gaussian model to simultaneously capture the aforementioned two distinct structured sparsities, and 2) a new bilinear Bayesian inference algorithm capable of jointly estimating the associated channel coefficients, user activity patterns, sub-array activity patterns ($a.k.a$. spatial non-stationarity), boosted by expectation maximization (EM)-based auto-parameterization. In addition, to shed light on a realistic modeling of VRs, we also introduce a Matérn-cluster point process (MCPP)-based approach to imitate the clustered activity pattern due to spatial non-stationarity. The efficacy of the proposed bilinear JACE algorithm is confirmed by numerical simulations, which show that the proposed method not only significantly outperforms the state-of-the-art (SotA) but also can reach the performance of a genie-aided scheme over wide signal-to-noise-ratio (SNR) ranges, in both uniformly-random and MCPP-based sub-array activity scenarios.

We consider quadrature spatial modulation (QSM) schemes, which achieve high spectral efficiency (SE) via the dispersion of a relatively small number$P$of$M$-ary modulated symbols over a large number of combinations of$n_{T}$transmit antennas and$T$transmit instances. In particular, we design a new space-time block code (STBC)-based scalable QSM scheme combining high SE with maximum diversity and optimum coding gains. Deriving a closed-form expression for the optimum SE, we show that scaling the size$T$with$n_{T}$not only is required to achieve SE optimality, but also results in further gains in bit error rate (BER) performance. Building on the latter optimal parameterization, a fully optimized scalable QSM (OS-QSM) transmitter design is then obtained by introducing a new dispersion matrix index selection algorithm that ensures even utilization of spatial-temporal resources. Finally, a new greedy boxed iterative shrinkage thresholding algorithm (GB-ISTA) QSM receiver is proposed, which exploits the inherent sparsity of QSM signals and while detecting spatially and digitally modulated bits in a greedy fashion. The resulting low complexity of the new receiver, which is linear on$n_{T}$, enables the utilization of OS-QSM in systems of previously prohibitive dimensions.

Recently, an active discussion on the feasibility of Millimeter Wave (mmWave) frequencies for the Vehicle-to-Vehicle (V2V) communication have been carried out in research community. We contribute to this discussion by providing a comparison between explicit three-dimensional ray-tracing simulations and field trial measurements on 39 GHz frequency. Three basic practical and relevant cases for V2V communications are considered covering several important scenarios of daily life traffic. A close match between the measured and simulated results is found through explicit ray tracing simulations; thus validating the feasibility of the simulation model and underlying assumptions. Moreover, these outcomes also shed light on the potential and challenges of using mmWave frequencies for V2V communication. The acquired results indicate that the Reference Signal Received Power (RSRP) levels are sufficiently above the noise level even up to 100 m distance between TX and RX in case of a single obstructing car. Results also reveal the impact of moving vehicle intersecting the LOS between the TX and RX vehicle at road intersection, and they indicate a notable blockage loss in case of short TX-RX separation.