Despite the advantages of multi-agent reinforcement learning (MARL) for wireless use case such as medium access control (MAC), their real-world deployment in Internet of Things (IoT) is hindered by their sample inefficiency. To alleviate this challenge, one can leverage model-based reinforcement learning (MBRL) solutions, however, conventional MBRL approaches rely on black-box models that are not interpretable and cannot reason. In contrast, in this paper, a novel causal model-based MARL framework is developed by leveraging tools from causal learn- ing. In particular, the proposed model can explicitly represent causal dependencies between network variables using structural causal models (SCMs) and attention-based inference networks. Interpretable causal models are then developed to capture how MAC control messages influence observations, how transmission actions determine outcomes, and how channel observations affect rewards. Data augmentation techniques are then used to generate synthetic rollouts using the learned causal model for policy optimization via proximal policy optimization (PPO). Analytical results demonstrate exponential sample complexity gains of causal MBRL over black-box approaches. Extensive simulations demonstrate that, on average, the proposed approach can reduce environment interactions by 58%, and yield faster convergence compared to model-free baselines. The proposed approach inherently is also shown to provide interpretable scheduling decisions via attention-based causal attribution, revealing which network conditions drive the policy. The resulting combination of sample efficiency and interpretability establishes causal MBRL as a practical approach for resource-constrained wireless systems.

Traditional joint source-channel coding employs static learned semantic representations that cannot dynamically adapt to evolving source distributions. Shared semantic memories between transmitter and receiver can potentially enable bandwidth savings by reusing previously transmitted concepts as context to reconstruct data, but require effective mechanisms to determine when current content is similar enough to stored patterns. However, existing hard quantization approaches based on variational autoencoders are limited by frequent memory updates even under small changes in data dynamics, which leads to inefficient usage of bandwidth.To address this challenge, in this paper, a memory-augmented semantic communication framework is proposed where both transmitter and receiver maintain a shared memory of semantic concepts using modern Hopfield networks (MHNs). The proposed framework employs soft attention-based retrieval that smoothly adjusts stored semantic prototype weights as data evolves that enables stable matching decisions under gradual data dynamics. A joint optimization of encoder, decoder, and memory retrieval mechanism is performed with the objective of maximizing a reasoning capacity metric that quantifies semantic efficiency as the product of memory reuse rate and compression ratio. Theoretical analysis establishes the fundamental rate-distortion-reuse tradeoff and proves that soft retrieval reduces unnecessary transmissions compared to hard quantization under bounded semantic drift. Extensive simulations over diverse video scenarios demonstrate that the proposed MHN-based approach achieves substantial bit reductions around 14% on average and up to 70% in scenarios with gradual content changes compared to baseline.

This paper pioneers the domain of multi-user holographic unmanned aerial vehicle (UAV) communications, establishing a robust foundation for future advancements in next-generation aerial wireless networks. It investigates the joint design of hybrid holographic beamforming and three-dimensional (3D) positioning for a UAV equipped with a reconfigurable holographic surface (RHS), with the objective of maximizing the network’s sum rate. To tackle this inherently complex and non-convex optimization problem, a novel alternating optimization framework is proposed. The solution leverages zero-forcing (ZF) digital beamforming and a gradient ascent strategy to iteratively update the holographic beamforming weights and the UAV’s 3D location, while satisfying key system constraints. This framework is tailored to efficiently navigate the trade-offs between hybrid transceiver design and UAV mobility limitations, ensuring both adaptability and performance scalability. Simulation results confirm that the proposed approach achieves substantial gains in sum rate and system robustness compared to conventional methods, validating its effectiveness under diverse channel and deployment conditions.

The use of terahertz (THz) communications with massive multiple input multiple output (MIMO) systems in 6G can potentially provide high data rates and low latency communications. However, accurate channel estimation in THz frequencies presents significant challenges due to factors such as high propagation losses, sensitivity to environmental obstructions, and strong atmospheric absorption. These challenges are particularly pronounced in urban environments, where traditional channel estimation methods often fail to deliver reliable results, particularly in complex non-line-of-sight (NLoS) scenarios. This paper introduces a novel vision-based channel estimation technique that integrates causal reasoning into urban THz communication systems. The proposed method combines computer vision algorithms with variational causal dynamics (VCD) to analyze real-time images of the urban environment, allowing for a deeper understanding of the physical factors that influence THz signal propagation. By capturing the complex, dynamic interactions between physical objects (such as buildings, trees, and vehicles) and the transmitted signals, the model can predict the channel with up to twice the accuracy of conventional methods. This model improves estimation accuracy and demonstrates superior generalization performance. Hence, it can provide reliable predictions even in previously unseen urban environments. The effectiveness of the proposed method is particularly evident in NLoS conditions, where it significantly outperforms traditional methods such as by accounting for indirect signal paths, such as reflections and diffractions. Simulation results confirm that the proposed vision-based approach surpasses conventional artificial intelligence (AI)-based estimation techniques in accuracy and robustness, showing a substantial improvement across various dynamic urban scenarios. This framework provides a promising solution for enabling resilient THz communication, offering scalability and practicality for future 6G deployments in diverse urban landscapes.

Aligning with the global mandates pushing towards advanced technologies with reduced resource consumption and environmental impacts, the sustainability of wireless networks becomes a significant concern in 6G systems. To address this concern, a native integration of sustainability into the operations of next-generation networks through novel designs and metrics is necessary. Nevertheless, existing wireless sustainability efforts remain limited to energy-efficient network designs which fail to capture the environmental impact of such systems. In this paper, a novel sustainability metric is proposed that captures emissions per bit, providing a rigorous measure of the environmental footprint associated with energy consumption in 6G networks. This metric also captures how energy, computing, and communication resource parameters influence the reduction of emissions per bit. Then, the problem of allocating the energy, computing and communication resources is posed as a multi-objective (MO) optimization problem. To solve the resulting non-convex problem, our framework leverages MO reinforcement learning (MORL) to maximize the novel sustainability metric alongside minimizing energy consumption and average delays in successfully delivering the data, all while adhering to constraints on energy resource capacity. The proposed MORL methodology computes a global policy that achieves a Pareto-optimal tradeoff among multiple objectives, thereby balancing environmental sustainability with network performance. Simulation results show that the proposed approach reduces the average emissions per bit by around 26% compared to state-of-the-art methods that do not explicitly integrate carbon emissions into their control objectives.

When derived from the Bethe Free Energy (BFE) of the Generalized Linear Model (GLM), Approximate Message Passing (AMP) algorithms combine two asymptotic Large System Limit (LSL) simplifications which are asymptotic Gaussianity of extrinsics and large random matrix theory based asymptotic variance computations. In the provably convergent AMBGAMP algorithm, a LSL version of the BFE is derived. In Expectation Propagation (EP) style minimization, the LSL BFE cost function is augmented with Lagrangian terms for mean and variance consistency constraints, augmented with a quadratic version of the mean constraints as in the Method of Multipliers (MM). The mean Lagrange multipliers then get updated ADMM-style (Alternating Direction of MM). In this approach, the weights of the MM terms need to be carefully chosen, which is not part of the MM philosophy, and the Lagrange multipliers have no particular meaning. On the other hand, AMP can be derived by directly introducing LSL simplifications in the Belief Propagtion (BP) algorithm that minimizes the original GLM BFE. This allows to relate extrinsic messages to posterior pdfs by first-order Taylor series expansion based perturbations. We also apply LSL approximations to the variances of the various Gaussians involved, which in fact leads to a rederivation of a fundamental LSL theorem describing the deterministic limit of posterior variances. We show that this LSL version of BP leads to BFE modifications that correspond to the augmented Lagrangian of the LSL BFE, explaining its weights and Lagrange Multipliers. These insights should facilitate the extension of AMP to more complex settings such as bilinear models.

Semantic communication (SC) is a promising approach for enabling reliable communication with minimal data transfer while maintaining seamless connectivity for wireless users. Unlocking the advantages of multi-user SC systems requires revisiting the communication and computation resource allocation problem focusing on the users' reasoning abilities. Reasoning in SC allows end-users to infer missing information or anticipate future events more effectively. Yet, state-of-the-art SC systems primarily focus on resource allocation through compression based on semantic relevance, while overlooking the underlying data generation mechanisms and the tradeoff between communications and computing. Thus, they cannot help prevent a disruption in connectivity. In contrast, in this paper, a novel framework for computing and communication resource allocation is proposed that seeks to demonstrate how SC systems with reasoning capabilities at the users can improve reliability in an end-to-end multi-user wireless system with intermittent communication links. Towards this end, a novel reasoning-aware SC system is proposed for enabling users to utilize their local computing resources to reason the representations when the communication links are unavailable. To optimize communication and computing resource allocation in this system, a noncooperative game is formulated to maximize the effective semantic information (computed as a product of reliability and semantic information) while controlling the number of semantically relevant links that are disrupted. To find a Nash equilibrium of the game, an algorithm based on best response is proposed. Simulation results show that the proposed reasoning-aware SC system results in at least a 16.6% enhancement in throughput and a significant improvement in reliability compared to classical communications systems that do not incorporate reasoning.

Building the next-generation wireless systems that could support services such as the metaverse, digital twins (DTs), and holographic teleportation is challenging to achieve exclusively through incremental advances to conventional wireless technologies like metasurfaces or holographic antennas. While the 6G concept of artificial intelligence (AI)-native networks promises to overcome some of the limitations of existing wireless technologies, current developments of AI-native wireless systems rely mostly on conventional AI tools such as auto-encoders and off-the-shelf artificial neural networks. However, those tools struggle to manage and cope with the complex, nontrivial scenarios faced in real-world wireless environments and the growing quality-of-experience (QoE) requirements of the aforementioned, emerging wireless use cases. In contrast, in this article, we propose to fundamentally revisit the concept of AI-native wireless systems, equipping them with the common sense necessary to transform them into artificial general intelligence (AGI)-native systems. Our envisioned AGI-native wireless systems acquire common sense by exploiting different cognitive abilities such as reasoning and analogy. These abilities in our proposed AGI-native wireless system are mainly founded on three fundamental components: a perception module, a world model, and an action-planning component. Collectively, these three fundamental components enable the four pillars of common sense that include dealing with unforeseen scenarios through horizontal generalizability, capturing intuitive physics, performing analogical reasoning, and filling in the blanks. Toward developing these components, we start by showing how the perception module can be built through abstracting real-world elements into generalizable representations. These representations are then used to create a world model, founded on principles of causality and hyperdimensional (HD) computing. Specifically, we propose a concrete definition of a world model, viewing it as an HD causal vector space that aligns with the intuitive physics of the real world—a cornerstone of common sense. In addition,we discuss how this proposed world model can enable analogical reasoning and manipulation of the abstract representations. Then, we show how the world model can drive an action-planning feature of the AGI-native network. In particular, we propose an intent-driven and objective-driven planning method that can maneuver the AGI-native network to plan its actions. These planning methods are based on brain-inspired frameworks such as integrated information theory and hierarchical abstractions that play a crucial role in enabling human-like decision-making. Next, we explain how an AGI-native network can be further exploited to enable three use cases related to human users and autonomous agent applications: 1) analogical reasoning for the next-generation DTs; 2) synchronized and resilient experiences for cognitive avatars; and 3) brain-level metaverse experiences exemplified by holographic teleportation. Finally, we conclude with a set of recommendations to ignite the quest for AGI-native systems. Ultimately, we envision this article as a roadmap for the next generation of wireless systems beyond 6G.

In this paper, a pragmatic semantic communication framework that enables effective goal-oriented information sharing between two-intelligent agents is proposed. In particular, semantics is defined as the causal state that encapsulates the fundamental causal relationships and dependencies among different features extracted from data. The proposed framework leverages the emerging concept in machine learning (ML) called theory of mind (ToM). It employs a dynamic two-level (wireless and semantic) feedback mechanism to continuously fine-tune neural network components at the transmitter. Thanks to the ToM, the transmitter mimics the actual mental state of the receiver's reasoning neural network operating semantic interpretation. Then, the estimated mental state at the receiver is dynamically updated thanks to the proposed dynamic two-level feedback mechanism. At the lower level, conventional channel quality metrics are used to optimize the channel encoding process based on the wireless communication channel's quality, ensuring an efficient mapping of semantic representations to a finite constellation. Additionally, a semantic feedback level is introduced, providing information on the receiver's perceived semantic effectiveness with minimal overhead. Numerical evaluations demonstrate the framework's ability to achieve efficient communication with a reduced amount of bits while maintaining the same semantics, outperforming conventional systems that do not exploit the ToM-based reasoning.

Semantic communication (SC) is an effective approach to communicate reliably with minimal data transfer while simultaneously providing seamless connectivity. In this paper, a novel emergent SC (ESC) framework is proposed. This ESC system is composed of two key components: A signaling game for emergent language design and a neuro-symbolic (NeSy) artificial intelligence (AI) approach for causal reasoning. In order to design the language, the signaling game is solved using an alternating maximization between the transmit and receive nodes utilities. The generalized Nash equilibrium is characterized, and it is shown that the resulting transmit and receive signaling strategies lead to a local equilibrium solution. As such, the emergent language not only creates an efficient (in physical bits transmitted) transmit vocabulary dependent on communication contexts but it also aids the reasoning process (and enables generalization to unseen scenarios) by splitting complex received messages into simpler reasoning tasks for the receiver. The causal description (symbolic component) at the transmitter is then modeled using the emerging AI framework of generative flow networks (GFlowNets), whose parameters are optimized for higher semantic reliability. Using the reconstructed causal state, the receiver evaluates a set of logical formulas (symbolic part) to execute its task. This evaluation of logical formulas is done by combining GFlowNet, with the logical expressiveness of the symbolic structure, inspired from logical neural networks. The ESC system is also designed to enhance the novel semantic metrics of information, reliability, distortion and similarity that are designed using rigorous algebraic properties from category theory thereby generalizing the metrics beyond Shannon’s notion of uncertainty. Simulation results confirm that the ESC system effectively communicates with reduced bits and achieves superior semantic reliability compared to conventional wireless systems and state-of-the-art SC systems lacking causal reasoning capabilities. Additionally, the overhead involved in language creation diminishes over time, validating the system’s ability to generalize across multiple tasks.

Generalized Approximate Message Passing (GAMP) allows for Bayesian inference in linear models with non-identically independently distributed (n.i.i.d.) priors and n.i.i.d. measurements of the linear mixture outputs. It represents an efficient technique for approximate inference, which becomes accurate when both rows and columns of the measurement matrix can be treated as sets of independent vectors and both dimensions become large. It has been shown that the fixed points of GAMP correspond to the extrema of a large system limit of the Bethe Free Energy (LSL-BFE), which represents a meaningful approximation optimization criterion regardless of whether the measurement matrix exhibits the independence properties. However, the convergence of (G)AMP can be notoriously problematic for certain measurement matrices, and the only sure fixes so far are damping (by a difficult-to-determine amount) or perform a double ADMM. In this paper, we revisit the GAMP algorithm (as e.g. for sparse Bayesian learning (SBL)) by more rigorously applying an alternating constrained minimization strategy to an appropriately reparameterized LSL BFE. This guarantees convergence, at least to a local optimum. We furthermore introduce a natural extension of the BFE to integrate the estimation of (the SBL) hyperparameters via Variational Bayes, leading to Variational AMBGAMP or VAM-BGAMP.

Future wireless services, such as the metaverse require high information rate, reliability, and low latency. Multi-user wireless systems can meet such requirements by utilizing the abundant terahertz bandwidth with a massive number of antennas, creating narrow beamforming solutions. However, existing solutions lack proper modeling of channel dynamics, resulting in inaccurate beamforming solutions in high-mobility scenarios. Herein, a dynamic, semantically aware beamforming solution is proposed for the first time, utilizing novel artificial intelligence algorithms in variational causal inference to compute the time-varying dynamics of the causal representation of multi-modal data and the beamforming. Simulations show that the proposed causality-guided approach for Terahertz (THz) beamforming outperforms classical MIMO beamforming techniques.

Generalized Approximate Message Passing (GAMP) allows for Bayesian inference in linear models with non-identically independently distributed (n.i.i.d.) priors and n.i.i.d. measurements of the linear mixture outputs. It represents an efficient technique for approximate inference, which becomes accurate when both rows and columns of the measurement matrix can be treated as sets of independent vectors and both dimensions become large. It has been shown that the fixed points of GAMP correspond to the extrema of a large system limit of the Bethe Free Energy (LSL-BFE), which represents a meaningful approximation optimization criterion regardless of whether the measurement matrix exhibits the independence properties. However, the convergence of (G)AMP can be problematic for certain measurement matrices. In this paper, we revisit the GAMP algorithm by applying a simplified version of the Alternating Direction Method of Multipliers (ADMM) to minimizing the LSL-BFE. We show convergence of the mean and variance subsystems in AMBGAMP and in the Gaussian case, convergence of mean and LSL variance to the Minimum Mean Squared Error (MMSE) quantities.

Massive multiple-input multiple-output (MaMIMO) has become an integral part of the fifth-generation (5G) standard, and is envisioned to be further developed in beyond 5G (B5G) networks. With a massive number of antennas at the base station (BS), MaMIMO is best equipped to cater prominent use cases of B5G networks such as enhanced mobile broadband (eMBB), ultra-reliable low-latency communications (URLLC) and massive machine-type communications (mMTC) or combinations thereof. However, one of the critical challenges to this pursuit is the sporadic access behaviour of a massive number of devices in practical networks that inevitably leads to the conspicuous pilot contamination problem. Conventional linearly precoded physical layer strategies employed for downlink transmission in time division duplex (TDD) MaMIMO would incur a noticeable spectral efficiency (SE) loss in the presence of this pilot contamination. In this paper, we aim to integrate a robust multiple access and interference management strategy named rate-splitting multiple access (RSMA) with TDD MaMIMO for downlink transmission and investigate its SE performance. We propose a novel downlink transmission framework of RSMA in TDD MaMIMO, devise a precoder design strategy and power allocation schemes to maximize different network utility functions. Numerical results reveal that RSMA is significantly more robust to pilot contamination and always achieves a SE performance that is equal to or better than the conventional linearly precoded MaMIMO transmission strategy.

Intent-based networks that integrate sophisticated machine reasoning technologies will be a cornerstone of future wireless 6G systems. Intent-based communication requires the network to consider the semantics (meanings) and effectiveness (at end-user) of the data transmission. This is essential if 6G systems are to communicate reliably with fewer bits while simultaneously providing connectivity to heterogeneous users. In this paper, contrary to state of the art, which lacks explainability of data, the framework of neuro-symbolic artificial intelligence (NeSy AI) is proposed as a pillar for learning causal structure behind the observed data. In particular, the emerging concept of generative flow networks (GFlowNet) is leveraged for the first time in a wireless system to learn the probabilistic structure which generates the data. Further, a novel optimization problem for learning the optimal encoding and decoding functions is rigorously formulated with the intent of achieving higher semantic reliability. Novel analytical formulations are developed to define key metrics for semantic message transmission, including semantic distortion, semantic similarity, and semantic reliability. These semantic measure functions rely on the proposed definition of semantic content of the knowledge base and this information measure is reflective of the nodes' reasoning capabilities. Simulation results validate the ability to communicate efficiently (with less bits but same semantics) and significantly better compared to a conventional system which does not exploit the reasoning capabilities.

In many applications such as massive multi-input multi-output (MIMO) radar, massive MIMO channel estimation, speech processing, image and video processing, the received signals are tensors. In such applications, utilizing techniques from tensor algebra can be beneficial since it retains the tensorial structure in the received signal compared to processing on the matricized version of the same signal. Furthermore, the underlying parameters or states to be estimated are sparse in many of the above-said applications compared to the large system dimensions. In this paper, we propose techniques which allow handling the extension of sparse Bayesian learning (SBL) to time-varying states. Adding the parameters of the autoregressive process which is used to the model the time-varyings of the state leads to a non-linear (at least bilinear) state-space model. Belief propagation (BP) is a promising method to compute the minimum mean squared error (MMSE) or maximum a posteriori (MAP) estimates, but at the the expense of a high computational burden. However, inspired by a previous work on a combined BP and variational Bayes (VB) technique, we noted that using a combination of BP, VB, and expectation propagation (EP) can help to alleviate the computational complexity.

This paper considers a bidirectional full-duplex Multi-Input Multi-Output (MIMO) OFDM system. The limited dynamic range (LDR) noise model takes into account the hardware impairments in the radio frequency (RF) chain and is thus more practical. Hence we propose a beamforming (BF) design which takes into account the LDR noise characteristics and also is robust to imperfections in the estimated channel. At the transmit side, we introduce a two stage beamformer (BF) with an inner BF of lower dimension and an outer BF of higher dimension, both BFs being at the digital (baseband) side. The inner BF in OFDM domain handles directive transmission, while the outer BF in time domain handles self interference (SI). At the receive side, we propose a hybrid combiner which involves an analog phase shifter based BF, with fewer RF chains compared to the number of receive antennas and a digital (baseband) BF in OFDM domain. The analog BF helps reduce SI before analog-to-digital conversion (ADC). All the BFs are optimized using maximization of the expected weighted sum rate (WSR) which is solved using an alternating minorization approach. The proposed multi-stage BF architecture has multiple advantages including SI reduction during OFDM cyclic prefixes, with uplink (UL) or downlink (DL) possibly using different numerology or being asynchronous, allowing proper ADC operation.

In this paper, we address the fundamental problem of Sparse Bayesian Learning (SBL), where the received signal is a high-order tensor. We furthermore consider the problem of dictionary learning (DL), where the tensor observations are assumed to be generated from a Kronecker structured (KS) dictionary matrix multiplied by the sparse coefficients. Exploiting the tensorial structure results in a reduction in the number of degrees of freedom in the learning problem, since the dimensions of each of the factor matrices are significantly smaller than the matricized dictionary if we vectorize the observations. We propose a novel fast algorithm called space alternating variational estimation with dictionary learning (SAVED-KS), which is a version of variational Bayes (VB)-SBL pushed to the scalar level. Similarly, as for SAGE (space-alternating generalized expectation maximization) compared to EM, the component-wise approach of SAVED-KS compared to SBL renders it less likely to get stuck in bad local optima and its inherent damping (more cautious progression) also leads to typically faster convergence of the non-convex optimization process. Simulation results show that the proposed algorithm has a faster convergence rate and lower mean squared error (MSE) compared to the alternating least squares (ALS) based method for tensor decomposition.

We consider coordinated beamforming (BF) for the Multi-Input Single-Output (MISO) Interfering Broadcast Channel (IBC). The beamformers are optimized for the Ergodic Weighted Sum Rate (EWSR) or various approximations and bounds thereof, for the case of Partial Channel State Information at the Transmitters (CSIT). Gaussian (posterior) partial CSIT can optimally combine channel estimate and channel covariance information. With Gaussian partial CSIT, the beamformers only depend on the means (estimates) and (error) covariances of the channels. We extend a recently introduced large system analysis for optimized beamformers with partial CSIT, by a stochastic geometry inspired randomization of the channel covariance eigen spaces, leading to much simpler analytical results, which depend only on some essential channel characteristics. In the Massive MISO (MaMISO) limit, we obtain deterministic approximations of the signal and interference plus noise powers at the receivers for various BFs, which are tight as the number of BS antennas and the total user subspace dimension tend to infinity at fixed ratio. Simulation results exhibit the correctness of the large system results and the performance superiority of optimal BF designs based on both the MaMISO limit of the EWSR and using Linear Minimum Mean Squared Error (LMMSE) channel estimates.

This work deals with hybrid beamforming for the MIMO Interfering Broadcast Channel (IBC), i.e. the Multi-Input Multi-Output (MIMO) Multi-User (MU) Multi-Cell (MC) downlink (DL) channel. Hybrid beamforming (HBF) is a low complexity alternative for fully digital precoding in Massive MIMO systems. Hybrid architectures involve a combination of digital and analog processing that enables both beamforming and multiplexing gains. We consider BF design by maximizing the Weighted Sum Rate (WSR) for the case of Perfect Channel State Information at the Transmitter (CSIT). We optimize the WSR using minorization and alternating optimization, the result of which is observed to converge fast. The design is proposed for both fully and partially connected analog BF architectures. Moreover, we consider the BF design under realistic scenarios with per-RF (radio frequency) chain or per-antenna power constraints, leading to novel interference leakage aware water filling procedures. Simulation results illustrate the good WSR performance of the designs and the gains over naive constraint satisfaction approaches.