Semantic communication has gained significant attention with the advances in machine learning. Most semantic communication works focus on either task execution or data reconstruction, with some recent works combining the two. In this work, we propose a semantic communication system for concurrent task execution and data reconstruction for a multi-view scenario, which we formulate as the maximization of mutual information. To investigate the trade-off between the two objectives, we formulate a joint objective as a convex combination of task execution and data reconstruction. We show that under specific assumptions, the SSIM loss can be obtained from the mutual information maximization objective for data reconstruction, which takes human visual perception into account. Furthermore, for constant resource use, we show that by increasing the weight of the reconstruction objective up to a certain point, the task execution performance can be kept nearly constant, while the data reconstruction can be significantly improved.

Reliable inspection of nanosurfaces is essential to ensure the quality of nanostructure manufacturing. Angle-resolved scatterometry provides a non-invasive inspection method that can be used in-line but often suffers from long acquisition times due to dense angular sampling. This paper addresses the data acquisition challenge by proposing an end-to-end compressed learning framework for 5-level vacancy deficiency detection in zinc oxide nanograss using ARS images. The proposed framework integrates a learnable latitude-based sampling layer with a convolutional neural network, allowing sampling and classification to be jointly optimized during training. The sampling layer exploits the physical structure of ARS patterns and learns informative latitudinal regions, which reduces the sampling search space and improves convergence. Evaluation results show that the proposed approach achieves high and stable deficiency-level classification performance under different noise conditions. Using full ARS images, the model achieves 94.2

Accurate channel state information (CSI) is essential for spectral efficiency (SE) in 6G multiple-input multiple-output orthogonal frequency-division multiplexing (MIMO-OFDM) systems, yet conventional methods incur substantial pilot overhead. This letter introduces KronFormer, a pilot-to-prediction (P2P) Transformer with factorized spatial-temporal attention aligned to the Kronecker structure of MIMO channel correlations. Unlike unfactorized Transformers that flatten spatial dimensions, KronFormer preserves the 4-D tensor structure throughout processing and decouples spatial and temporal attention, avoiding the quadratic complexity of joint spatial-temporal processing. Simulations on COST 259 channels demonstrate that KronFormer reduces attention complexity by 20 & times; and mean squared error (MSE) by up to 87 & times; over unfactorized Transformers in 8 & times;8 MIMO. KronFormer limits SE degradation to 10-13% from 50 to 200 km/h, a 3 & times; reduction compared to the 34-39% loss of conventional Wiener filters. The architecture enables direct multi-step inference for scalable 6G channel prediction.

This paper highlights key areas in which AI/ML can play a transformative role in 6G, with an emphasis on energy-efficient solutions. Contextualized within Germany’s national 6G research initiative, "6G Access, Network of Networks, Automation and Simplification" (6G-ANNA) emerges as the lighthouse project, providing a holistic vision that sets the direction for numerous specialized research efforts. Within this context, this paper aims to present ideas relevant to both academia and industry by identifying key research directions for AI/ML. We begin by reviewing the latest developments in using AI/ML for the Fifth Generation (5G) New Radio (NR) air interface as discussed in 3rd Generation Partnership Project (3GPP) Releases 18 and 19, and examine how these advancements pave the way for a native and energy-efficient AI/ML air interface in 6G. Key results are presented on AI/ML-driven optimization of radio frequency (RF) frontends, along with a strong focus on the role of AI/ML in diverse signal processing tasks and energy saving mechanisms, which demonstrate the potential of AI/ML in improving spectral efficiency and reducing energy consumption. The discussion further introduces methodologies for testing AI/ML-based signal processing tailored for the 6G physical layer, addressing practical challenges relevant to industry stakeholders and standard development organizations. Finally, we discuss the standardization aspects critical for realizing a future AI-native air interface in 6G, aligning our findings with ongoing and upcoming global standardization activities.

Industrial networks often demand hyper-reliable, low-latency communication (HRLLC) to support closed-loop control and automation. However, performance is severely undermined by dynamic fluctuations caused by unknown interferers, making it difficult to uphold reliability requirements. Channel statistics and mobility of interferers lead to significant variations in perceived signal-to-interference-plus-noise ratio (SINR). To cope with such interference dynamics, predictive resource allocation is required, whereby resource selection must be determined before transmission, making SINR prediction essential. This paper introduces Reliability-Aware SINR Prediction (RASP), a probabilistic framework that predicts the SINR distribution under explicit block error rate (BLER) constraints. RASP employs a Mixture Density Network within a Conditional Value-at-Risk (CVaR)-based primal–dual optimization scheme to adapt streaming SINR samples while satisfying reliability constraints. Based on real-world industrial measurements with co-subband interferers, we show that RASP achieves the target BLER of 10−6 at the 95th percentile across three distinct scenarios, outperforming the baselines by a significant margin.

5G-Beyond (5G-B)/6G require higher spectral efficiency (SE) and data throughput (DT), motivating reduced pilot overhead in orthogonal frequency-division multiplexing (OFDM) systems. Traditional methods target temporal prediction and address frequency selectivity via pilot-tone interpolation, leaving time-frequency prediction unexplored. To address this gap, we propose a Transformer-based pilot-to-prediction (P2P) neural network (NN) for joint channel estimation and prediction. Exploiting the Transformer's self-attention, our method captures temporal dynamics and inter-subcarrier correlations to predict the frequency-selective channel across all subcarriers using only past pilot observations, requiring no pilots in the symbols being predicted. Simulation results show that our approach closely matches the accuracy of a genie-aided receiver with perfect channel knowledge and outperforms Wiener filtering and long short-term memory (LSTM) baselines. Consequently, pilot overhead is substantially reduced, enhancing SE by 33.3%. Achievable data rates are thus improved, underscoring the suitability of Transformer-based channel prediction for future wireless standards.

Nanoscale manufacturing requires high-precision surface inspection to guarantee the quality of the produced nanostructures. For production environments, angle-resolved scatterometry offers a non-invasive and in-line compatible alternative to traditional surface inspection methods, such as scanning electron microscopy. However, angle-resolved scatterometry currently suffers from long data acquisition time. Our study addresses the issue of slow data acquisition by proposing a compressed learning framework for the accurate recognition of nanosurface deficiencies using angle-resolved scatterometry data. The framework uses the particle swarm optimization algorithm with a sampling scheme customized for scattering patterns. This combination allows the identification of optimal sampling points in scatterometry data that maximize the detection accuracy of five different levels of deficiency in ZnO nanosurfaces. The proposed method significantly reduces the amount of sampled data while maintaining a high accuracy in deficiency detection, even in noisy environments. Notably, by sampling only 1% of the data, the method achieves an accuracy of over 88%, which further improves to more than 95%—approaching the 96.4% accuracy achieved with full image classification—when the sampling rate is increased to 7%. These results demonstrate a favorable balance between data reduction and classification performance. The obtained results also show that the compressed learning framework effectively identifies critical sampling areas. The framework was further validated on GaN nanosurfaces, which confirmed that the sampling strategy generalizes effectively across different materials and achieves classification accuracies close to full-image performance.

The sixth generation (6G) of wireless networks is envisioned to achieve far beyond the capabilities of fifth generation (5G), necessitating significant innovations at the physical layer (PHY). These include exploration of several fundamental trade-offs between spectral efficiency, reliability, and energy consumption, and enhancing the performance of key enablers for 6G PHY. This paper synthesizes key insights from the 6G-ANNA research initiative on emerging PHY technologies for 6G to provide a holistic exploration of the ongoing trends in the 6G research. The investigations span novel waveform and channel coding techniques for improved energy efficiency, the “Gearbox PHY” concept for adaptive transceiver operations, and optimized radio transceiver designs that balance complexity and power consumption. The study also examines advanced multiple access schemes and cell-free massive multiple-input multiple-output (MIMO) architectures to enhance spectral efficiency and uniform coverage. Integrated artificial intelligence (AI) solutions at the PHY layer and insights to security and trustworthiness challenges in 6G networks are also provided. The findings offer insights into the fundamental trade-offs and provide several key PHY innovations that address sustainability, capacity, and resiliency challenges of future 6G wireless systems.

While learning-based channel predictors show promise for reducing pilot overhead in next-generation orthogonal frequency-division multiplexing (OFDM) systems, the lack of principled performance benchmarks prevents rigorous assessment of their proximity to theoretical limits. We derive a closed-form linear minimum mean-square error (LMMSE) benchmark for strictly causal pilot-to-prediction (P2P) in frequency-selective OFDM systems. For Gaussian wide-sense stationary uncorrelated scattering (WSSUS) channels, this benchmark equals the Bayesian Cramér–Rao bound (BCRB), decomposing irreducible error into temporal prediction and pilot-induced estimation components with closed-form eigenvalue expressions. Evaluating five neural architectures, we find that Transformer and state-space model (SSM) closely track the LMMSE benchmark on COST 259 channels, while long short-term memory (LSTM) and Autoformer exhibit larger gaps. Crossformer operates below the benchmark at most tested speeds, consistent with nonlinear exploitation of non-Gaussian structure in finite sum-of-sinusoids (SoS) channel models, though it exhibits elevated gaps at low Doppler; a reduced-capacity ablation indicates a capacity–architecture interaction. On 3rd Generation Partnership Project (3GPP) TR 38.901 Tapped Delay Line-A (TDL-A) channels, both Transformer and Crossformer exhibit strictly positive gaps across all tested speeds. This demonstrates that while nonlinear methods can surpass the linear benchmark, Transformer and SSM have closely approached it, implying that further gains depend primarily on pilot density, signal-to-noise ratio (SNR), or prediction horizons.

This paper addresses the challenge of pilot overhead reduction in Multi-Input Multi-Output (MIMO) systems employing Orthogonal Frequency Division Multiplexing (OFDM), a promising candidate waveform for future 6G networks. As networks scale with increasing antennas and subcarriers, the pilot overhead also rises, reducing spectral efficiency. Moreover, the stringent reliability requirements of 6G demand a robust channel estimation scheme with minimal pilot overhead. To meet this need, we introduce a Semi-Blind Channel Estimation framework with Adaptive Pilot Design (SBCE-APD) that aims to achieve a target estimation accuracy with minimal pilot overhead. Traditional techniques primarily aim to improve estimation accuracy for a fixed pilot budget. In contrast, the proposed SBCE-APD operates as a closed-loop framework that minimizes pilot overhead subject to a target accuracy constraint. Its main novelty lies in the system-level integration of adaptive pilot-count selection, pilot placement, and semi-blind channel estimation. The method combines two complementary estimation branches: a pilot-based estimator and a non-pilot-based estimator. Their integration forms a semi-blind estimation structure, reducing dependence on pilot symbols. Furthermore, the pilot pattern is adaptively adjusted over the air, optimizing the number of pilots under varying network conditions to maintain the desired accuracy efficiently. Simulation results demonstrate substantial gains in pilot efficiency while preserving reliable channel estimation, establishing SBCE-APD as an effective and scalable solution for 6G MIMO-OFDM networks.

High-accuracy positioning is critical for emerging applications such as autonomous driving, industrial automation, augmented reality, and smart cities. 3GPP Release 18 introduced carrier-phase (CP) positioning for 5G that offers superior accuracy compared to conventional time-based methods such as time of arrival (ToA). However, CP-based positioning requires resolving the integer phase ambiguity, which refers to the unknown number of full-wavelength cycles completed during signal propagation. Joint processing of ToA and CP can mitigate this integer ambiguity by narrowing down the search space of possible integers, particularly for short wavelengths. This paper investigates the performance of a positioning method that integrates ToA and CP measurements. As a main contribution, the analysis explicitly accounts for the error correlation between ToA and CP measurements. Furthermore, the study analyzes the impact of key 5G system parameters on positioning accuracy using this correlation-aware joint method in both factory and urban environments, where many 5G positioning applications are expected to emerge. The results highlight that exploiting this correlation can further improve positioning performance by approximately 7 percent. Moreover, the findings of this study provide insight into how 5G system parameters can be tuned to achieve centimeter-level accuracy under favorable conditions.

We propose a novel interference prediction scheme to improve link adaptation (LA) in densely deployed industrial sub-networks (SNs) with high-reliability and low-latency communication (HRLLC) requirements. The proposed method aims to improve the LA framework by predicting and leveraging the heavy-tailed interference probability density function (pdf). Interference is modeled as a latent vector of available channel quality indicator (CQI), using a vector discrete-time state-space model (vDSSM) at the SN controller, where the CQI is subjected to compression, quantization, and delay-induced errors. To robustly estimate interference power values under these impairments, we employ a low-complexity, outlier-robust, sparse Student-t process regression (SPTPR) method. This is integrated into a modified unscented Kalman filter, which recursively refines predicted interference using CQI, enabling accurate estimation and compensating protocol feedback delays-crucial for accurate LA. Numerical results show that the proposed method achieves over 10x lower complexity compared to a similar non-parametric baseline. It also maintains a BLER below the 90th percentile target of10(-6) while delivering performance comparable to a state-of-the-art supervised technique using only CQI reports.

In industrial settings, reliable and efficient data transmission is crucial. However, traditional Orthogonal Multiple Access (OMA) techniques are limited by their inability to manage mixed-criticality systems and varying data importance. We propose a novel transmission scheme that combines Non-Orthogonal Multiple Access (NOMA) with Unequal Error Protection (UEP) and leverages Multi-Service Modulation and Coding Scheme (MCS) (MS-MCS) tables. This approach enables simultaneous transmission of multiple services over a single channel, each with tailored reliability levels, improving spectral efficiency and reliability in dense wireless environments.

Satellite-based communications are expected to be a substantial future market in 6G networks. As satellite constellations grow denser and transmission resources remain limited, frequency reuse plays an increasingly important role in managing inter-user interference. In the multi-user downlink, precoding enables the reuse of frequencies across spatially separated users, greatly improving spectral efficiency. The analytical calculation of suitable precodings for perfect channel information is well studied, however, their performance can quickly deteriorate when faced with, e.g., outdated channel state information or, as is particularly relevant for satellite channels, when position estimates are erroneous. Deriving robust precoders under imperfect channel state information is not only analytically intractable in general but often requires substantial relaxations of the optimization problem or heuristic constraints to obtain feasible solutions. Instead, in this paper we flexibly derive robust precoding algorithms from given data using reinforcement learning. We describe how we adapt the applied Soft Actor-Critic learning algorithm to the problem of downlink satellite beamforming and show numerically that the resulting precoding algorithm adjusts to all investigated scenarios. The considered scenarios cover both single satellite and cooperative multi-satellite beamforming, using either global or local channel state information, and two error models that represent increasing levels of uncertainty. We show that the learned algorithms match or markedly outperform two analytical baselines in sum rate performance, adapting to the required level of robustness. We also analyze the mechanisms that the learned algorithms leverage to achieve robustness. The implementation is publicly available for use and reproduction of the results.

In wireless communication systems, accurate Channel State Information (CSI) is essential for base stations to perform downlink precoding. While much of the existing research primarily focus on compressing the CSI matrix, they often neglect the impact of subsequent pre-transmission processes such as quantization, channel coding, and modulation. This paper investigates two distinct approaches for CSI dimensionality reduction: TransNet, a transformer-based neural network, and Dynamic Mode Decomposition (DMD), a mathematical decomposition technique for dynamical systems. We analyze how quantization, channel coding, and modulation affect CSI feedback for both methods. Unlike TransNet, DMD can decompose the channel matrix into components (called modes) with varying significance. This decomposition allows for an effective application of Unequal Error Protection (UEP) techniques to DMD modes, which is not feasible with TransNet-based CSI. Simulation results reveal that while the compression performance of TransNet and DMD varies based on factors like target CSI size and channel estimation error, integrating UEP techniques for DMD-based CSI yields superior CSI transmission performance compared to TransNet-based CSI.

Non-Terrestrial Networks (NTNs) are critical enablers of ubiquitous connectivity in future 6 G systems, but they also face challenges such as long propagation delays, significant Doppler effects, and diverse channel conditions. Developing and testing communication technologies for NTNs requires realistic and flexible simulation tools. In this work, OpenNTN is presented as an open-source software framework for simulating NTN channel models based on the 3GPP TR38.811 standard. OpenNTN is designed as an extension to the Python-based Sionna ${ }^{\text {TM }}$ framework, providing channel models compatible with existing interfaces, enabling the use of the powerful tools found in Sionna ${ }^{\text{TM}}$. The framework offers a flexible user interface, enabling researchers to investigate diverse scenarios. As an open-source implementation, OpenNTN empowers the research community to fully use and adapt the framework, offering a solution for both fast and easy NTN research using realistic channel models, while also providing the opportunity to extend the model for custom research interests beyond the current state of the standards.

In wireless communication, base stations rely on downlink Channel State Information (CSI) to perform precoding. Without channel reciprocity, the mobile station must transmit the estimated CSI back to the base station. Due to the time-varying nature of the environment, channel characteristics change constantly, requiring regular CSI feedback updates at intervals that depend on the rate of change. Thus, increasing the interval between the CSI updates, can reduce the average CSI feedback overhead. Additionally, in Multiple-Input Multiple-Output (MIMO) systems, the CSI feedback overhead grows with the number of antennas and bandwidth, leading to a potential performance bottleneck. To reduce the CSI feedback overhead and increase the intervals between CSI updates, we propose a novel method that integrates Dynamic Mode Decomposition (DMD) and Convolutional AutoEncoders (CAE) to model and compress channel dynamics. DMD decomposes the channel matrix into modes that can predict the future state of the channel, thereby extending CSI feedback intervals, while CAE captures the most relevant features of these modes for further compression. Simulation results demonstrate that this method effectively reduces feedback overhead and prolongs the intervals between CSI updates.

Interference prediction that accounts for extreme and rare events remains a key challenge for ultra-densely deployed sub-networks (SNs) requiring hyper-reliable low-latency communication (HRLLC), particularly under dynamic mobility, rapidly varying channel statistics, and sporadic traffic. This paper proposes a novel calibrated interference tail prediction framework—a hybrid statistical and machine learning (ML) approach that integrates an inverted quantile patch transformer (iQPTransformer) within extreme value theory (EVT). It captures interference dynamics and tail behavior while quantifying uncertainty to provide statistical coverage guarantees. In resource-constrained SN scenarios, we introduce the split-iQPTransformer, enabling collaborative training by distributing neural network components between sensor-actuator (SA) pairs and the SN controller, while maintaining minimal performance disparity compared to the centralized iQPTransformer. The framework effectively handles deep fading, random traffic, and time-division duplexing (TDD) misalignments and is resilient to rare and extreme interference events. Extensive evaluations are performed under two mobility models and two realistic SN traffic patterns, using a spatially consistent 3GPP channel model across all scenarios. Its effectiveness is demonstrated by leveraging the predicted interference tail distribution, which inherently captures risk, within an existing predictive resource allocation framework evaluated under stringent block error rate (BLER) requirements for short-packet transmissions. Experimental results show consistent achievement of BLER targets beyond the 95th percentile in the hyper-reliable regime, significantly outperforming baseline approaches.

The rapid evolution of cellular networks, driven by the proliferation of mobile devices and the exponential growth of the Internet of Things (IoT), has significantly advanced wireless communication technologies. Fifth generation of wireless communications technology (5G) enhanced data rates, latency, and network capacity, resulting in the emergence of new applications. However, the sixth generation (6G) is foreseen to support a new set of use cases with diverse requirements. This paper explores the critical role of artificial intelligence (AI) in shaping the trajectory from 5G to 6G. We discuss AI applications in 5G for network planning, resource allocation, traffic management, and security, as well as propose infrastructure upgrades, like edge servers and enhanced network topologies, to support AI in 6G. Additionally, we outline a visionary perspective on AI’s potential contributions to 6G, highlighting its role in enabling innovative services and applications. By providing this forward-looking perspective, this paper aims to stimulate discussion and guide the development of intelligent and autonomous 6G networks.

The rapid growth of non-terrestrial communication necessitates its integration with existing terrestrial networks, as highlighted in 3GPP Releases 16 and 17. This paper analyses the concept of functional splits in 3D-Networks. To manage this complex structure effectively, the adoption of a Radio Access Network (RAN) architecture with Functional Split (FS) offers advantages in flexibility, scalability, and cost-efficiency. RAN achieves this by disaggregating functionalities into three separate units. Analogous to the terrestrial network approach, 3GPP is extending this concept to non-terrestrial platforms as well. This work presents a general analysis of the requested Fronthaul (FH) data rate on feeder link between a non-terrestrial platform and the ground-station. Each split option is a trade-of between FH data rate and the respected complexity. Since flying nodes face more limitations regarding power consumption and complexity on board in comparison to terrestrial ones, we are investigating the split options between lower and higher physical layer.