Integrating sensing, enabling network intelligence, and improving spectral efficiency are key drivers of the vision for sixth-generation (6G) wireless systems. Conventional pilot-based channel characterization methods incur significant radio-resource overhead, particularly in indoor traffic scenarios, whereas sensing-aided channel characterization offers a viable alternative. Digital twins (DTs) are seen as a promising enabler providing high-fidelity virtual representations of physical environments, which can enhance sensing-aided channel characterization. This paper proposes a DT framework for channel characterization in radio access networks (RANs), leveraging multimodal sensing. Ray tracing is performed on a fused and segmented multimodal point cloud that supports arbitrary placements of both transmitters and receivers, and the DT is validated using a bistatic radio link. In this work, we use received signal strength (RSS) and power delay profile (PDP) as the KPIs. Our results show that multimodal sensing significantly improves environmental awareness, achieving a relative mean absolute error of 2.5 % for RSS and below 15 % for PDP. Therefore, the DT facilitates accurate sensing-aided channel characterization in indoor environments, enabling closed-loop optimization in RAN.

Spiking neural networks (SNNs) achieve significantly lower energy consumption than traditional artificial neural networks (ANNs) through event-driven and sparse computation. This property is highly desirable in wireless systems with growing demands for energy-efficient intelligence, particularly on power-constrained user equipments. Therefore, this paper investigates the feasibility of SNN-based signal processing by examining an SNN-based soft demodulation architecture, termed spkLLR. We introduce two neuron activity regularized architectures; IM-spkLLR maximizes the information and RR-spkLLR regularizes the spiking rate. These models are simulated under an additive white Gaussian noise channel, and simulation results demonstrate that spkLLR achieves comparable bit error rate performance to ANN-based approach while reducing energy consumption by more than two times. The RR-spkLLR model consistently provides the highest energy efficiency by constraining spiking rates and minimizing synaptic operations. Analysis across 16-quadrature amplitude modulation (QAM), 64-QAM, and 256-QAM schemes further shows that spiking demappers maintain favorable scalability. Overall, the proposed spkLLR framework highlights the potential of SNN-based demapping for low-power, high-efficiency communication systems and establishes a foundation for future neuromorphic receiver designs.

Neural channel estimation has demonstrated promising link-level performance gains in simulation, yet its viability for real-time deployment in fifth generation (5G) base-station receivers under practical hardware constraints remains insufficiently explored. In particular, most existing work evaluates channel estimation offline using recorded datasets rather than under realistic over-the-air (OTA) conditions. This paper investigates the real-time OTA performance of neural channel estimation for 5G new radio physical uplink shared channel within a fully integrated hardware testbed. We develop two neural estimators: CNN-est, a 3D convolutional model, and DSCNN-est, a computationally efficient depthwise-separable variant. Both estimators employ a fused estimator–denoiser architecture and operate with a single-symbol Type-2 demodulation reference signal pilot pattern, achieving up to 5.1% uplink throughput gain. Simulation results show accuracy comparable to linear minimum mean square error across diverse clustered delay line channels, while real-time OTA measurements demonstrate up to 1 dB signal-to-noise-ratio reduction at a target uplink block error rate of 0.1 compared to the OpenAirInterface least-squares baseline under practical radio-frequency impairments. These findings provide concrete evidence of the feasibility and performance trade-offs of neural channel estimation in live 5G systems.

An end-to-end learning method for constellation shaping with a shaping-encoder assisted transceiver architecture is presented. The shaping encoder, which produces shaping bits with a higher probability of zeros, is used to produce an efficient symbol probability distribution. Both the probability distribution and the constellation geometry are jointly optimized, using end-to-end learning. Optimized constellations are evaluated using two iterative receiver architectures. Bit error rate (BER) performance gain is quantified against standard amplitude phase-shift keying (APSK) and quadrature amplitude modulation (QAM) constellations. A maximum BER gain of 0.3 dB and 0.15 dB are observed under two receivers for the learned constellations compared to standard APSK or QAM. The basic approach is extended to incorporate the full iterative detection and decoding loop, using the deep unfolding technique. A bit error rate gain of 0.1 dB is observed for the iterative scheme with learned constellations under block fading channel conditions, when compared to standard APSK.

This paper introduces the concept of seamlessly embedding AI-native Radio Access Networks (RANs) and integrated device-network approaches, aiming to extend AI capabilities across diverse devices - e.g., smartphones, Internet of Things (IoT) sensors, autonomous vehicles - and network infrastructures. By embedding AI deeply within RANs and fostering tight integration between devices and network architectures, we unlock unprecedented levels of intelligence, efficiency, flexibility, and performance in wireless communication systems. This integration enables devices and networks to collaboratively process, adapt, and optimize data in real-time, creating a synergistic AI-enabled ecosystem that forms the backbone of AI-native 6G. Unlike traditional wireless communication systems, which rely on static configurations and manual interventions, AI-native networks dynamically adapt to changing conditions, autonomously optimize resource allocation, and respond to diverse user demands and environmental factors. The 6GARROW project explores these transformative approaches, demonstrating how AI-native RANs and integrated device-network solutions deliver higher throughput, lower latency, and improved reliability. Beyond enhanced performance, this paradigm shift supports a diverse range of innovative services, including spatio-temporal communications, critical and compute-AI applications, omnipresent IoT, immersive user experiences, and global broadband connectivity. This paper outlines the vision, methodologies, and potential impact of the project, highlighting its role in advancing toward a more connected, intelligent, and adaptive wireless ecosystem for the future.

The large untapped spectrum in sub-THz allows for ultra-high throughput communication to realize many seemingly impossible applications in 6G. Phase noise (PN) is one key hardware impairment, which is accentuated as we increase the frequency and bandwidth. Furthermore, the modest output power of the power amplifier demands limits on peak to average power ratio (PAPR) signal design. In this work, we design a PN-robust, low PAPR single-carrier (SC) waveform by geometrically shaping the constellation and adapting the pulse shaping filter pair under practical PN modelling and adjacent channel leakage ratio (ACLR) constraints for a given excess bandwidth. We optimize the waveforms under conventional and state-of-the-art PN-aware demappers. Moreover, we introduce a neural-network (NN) demapper enhancing transceiver adaptability.We formulate the waveform optimization problem in its augmented Lagrangian form and use a back-propagation-inspired technique to obtain a design that is numerically robust to PN, while adhering to PAPR and ACLR constraints. The results substantiate the efficacy of the method, yielding up to 2.5 dB in the required Eb / N 0 under stronger PN along with a PAPR reduction of 0.5 dB. Moreover, PAPR reductions up to 1.2 dB are possible with competitive BLER and SE performance in both low and high PN conditions.

Future wireless networks will need to support diverse applications (such as extended reality), scenarios (such as fully automated industries), and technological advances (such as terahertz communications). Current wireless networks are designed to perform adequately across multiple scenarios so they lack the adaptability needed for specific use cases. Therefore, meeting the stringent requirements of next-generation applications incorporating technology advances and operating in novel scenarios will necessitate wireless specialized networks which we refer to as SpecNets. These networks, equipped with cognitive capabilities, dynamically adapt to the unique demands of each application, e.g., by automatically selecting and configuring network mechanisms. An enabler of SpecNets are the recent advances in artificial intelligence and machine learning (AI/ML), which allow to continuously learn and react to changing requirements and scenarios. By integrating AI/ML functionalities, SpecNets will fully leverage the concept of AI/ML-defined radios (MLDRs) that are able to autonomously establish their own communication protocols by acquiring contextual information and dynamically adapting to it. In this paper, we introduce SpecNets and explain how MLDR interfaces enable this concept. We present three illustrative use cases for wireless local area networks (WLANs): bespoke industrial networks, traffic-aware robust THz links, and coexisting networks. Finally, we showcase SpecNets' benefits in the industrial use case by introducing a lightweight, fast-converging ML agent based on multi-armed bandits (MABs). This agent dynamically optimizes channel access to meet varying performance needs: high throughput, low delay, or fair access. Results demonstrate significant gains over IEEE 802.11, highlighting the system's autonomous adaptability across diverse scenarios.

Sub-THz communications are envisioned to unlock ultra-high data rates, ultra-low latencies, and massive connectivity in future wireless systems by utilizing bandwidths of up to tens of GHz. This paper presents the holistic perspective on sub-THz communications developed within the European sixthgeneration (6G) flagship project Hexa-X-II. According to this perspective, successfully deploying sub-THz systems will require addressing several physical-layer (PHY) challenges associated with the harsh signal propagation and hardware limitations. In addition, sub-THz systems will drive advancements in essential wireless applications beyond communications, such as joint communications and sensing (JCAS). In this context, we present a collection of technical contributions and key findings from the Hexa-X-II project that have shaped this perspective. In contrast to existing works, these efforts collectively tackle the fundamental PHY challenges of sub-THz communications, i.e., understanding and modeling of radio propagation, radio-frequency (RF) power consumption and complexity, and hardware impairments. To achieve this, we present several results across three core areas, i.e., signal propagation and channel modeling, RF transceiver design, and PHY enablers, offering useful insights into the development of sub-THz systems. Lastly, we provide an overview of sub-THz JCAS, emphasizing it as one of the most promising applications in the sub-THz range.

Phase noise (PN) poses a significant challenge in sub-terahertz (sub-THz) communications, alongside the necessity for low peak-to-average power ratio (PAPR) transmissions. This letter introduces an end-to-end learned single-carrier (SC) neural transceiver, which consists of a PN-resilient and PAPR-constrained transmitter utilizing a trainable pilot scheme combined with a deep neural receiver tailored for sub-THz. The learned transceiver effectively compensates for both correlated and uncorrelated PN and the flat-fading line-of-sight (LOS) channel while maintaining lower PAPR. The results show a substantial reduction in pilot overhead while delivering superior spectral efficiency and up to 1.2 dB PAPR gains over the conventional baselines.

Reconfigurable intelligent surfaces (RIS)-based communications with reflection modulation (RM) is a novel area of research that opens up a range of unconventional modulation techniques. Existing literature primarily focuses on specific applications where the RIS encodes its own information onto its reflection pattern. Quadrature reflection modulation (QRM) and reflection pattern modulation (RPM) are two promising reflection pattern designs that effectively deliver local data available at the RIS. This paper explores a more general application of RIS-based information transfer for a single-user downlink system via jointly mapped RM (JRM), where the RIS and the access point (AP) jointly deliver the information available at the AP. The data symbols are mapped to a constellation of tuples, each tuple containing a transmit signal and a reflection pattern. Two JRM constellation designs are proposed, namely jointly-mapped QRM (JQRM) and jointly-mapped RPM (JRPM). The proposed constellation design employs a smaller transmit signal set size compared to a generic modulation scheme, increasing the separation among adjacent constellation points. A jointly active and passive beamforming design is adopted for a multiple-input-single-output (MISO) downlink system. The simulation results analyze and compare the bit-error-rate (BER) performance of the proposed JQRM and JRPM schemes, with their respective separately-mapped counterparts and theoretical upper bounds as benchmarks.

In this letter, we investigate the capacity of delay schemes in delayed bit-interleaved coded modulation (DBICM) at high signal-to-noise ratios (SNR) using alternative methods to numerical integration. A closed-form formula is derived to approximate the DBICM capacity at high SNR. For Gray codes, DBICM capacity is proven as asymptotically optimal irrespective of the delay scheme and we further characterize the minute variations in finite SNR values. Compared to BICM using the binary reflected Gray Code, an improvement of 0.1 dB in the BER performance is achieved using a non-Gray bit-labeling combined with its optimized DBICM delay scheme.

The objective of this study is to evaluate different waveform designs under typical hardware impacts and constraints for realizing sub-terahertz (sub-THz) wireless transmission. The assessment of the waveforms is performed using a proof-of-concept (PoC) demonstrator system based on real practical sub-THz hardware operating in the D-band at 144 GHz, with a transmission bandwidth of up to 5 GHz. The setup consists of radio-frequency integrated circuit (RFIC) components specifically designed for the D-band. We examine both conventional waveforms and those with low peak-to-average power ratio (PAPR) and enhanced robustness against phase noise, learned through data-driven techniques. The results showcase the feasibility of transmitting the low-PAPR waveforms without EVM degradation, suggesting that high-rate sub-THz wireless transmissions could be a viable option for future communication systems.

Ongoing discussions within 3 rd generation partnership project (3GPP) on Ambient IoT (A-IoT) studies are paving way towards more sustainable and energy-efficient communications for A-IoT devices. One device type under consideration is an energy harvesting device with active transmission capabilities that can harvest energy from an ambient sources. In this work, we embark on characterizing the operation of such A-IoT devices through system-level simulations, serving as an initial step in realizing the operation of these devices within the framework of 3GPP networks. Results demonstrate the importance of considering sleep state power consumption and the potential benefits of aligning energy harvesting cycle with an appropriate Discontinuous Reception (DRX) cycle for operation of A-IoT devices within the 3GPP framework.

The large untapped spectrum in the sub-THz allows for ultra-high throughput communication to realize many seemingly impossible applications in 6G. One of the challenges in radio communications in sub-THz is the hardware impairments. Specifically, phase noise is one key hardware impairment, which is accentuated as we increase the frequency and bandwidth. Furthermore, the moderate output power of the sub-THz power amplifier demands limits on peak to average power ratio (PAPR) signal design. Single carrier frequency domain equalization (SC-FDE) has been identified as a suitable candidate for sub-THz, although some challenges such as phase noise and PAPR still remain to be tackled. In this work, we design a phase noise robust, modest PAPR SC waveform by geometrically shaping the constellation under practical conditions. We formulate the waveform optimization problem in its augmented Lagrangian form and use a back-propagation-inspired technique to obtain a constellation design that is numerically robust to phase noise, while maintaining a relatively low PAPR compared to the conventional waveforms.

Integrated Sensing and Communication (ISAC) is one new technology under development for Sixth Generation (6G) systems. This paper focuses on creating a simulation pipeline for dynamic vehicular traffic scenarios and a novel approach to reducing wireless communication overhead with a LiDAR-based system. The simulation pipeline can be used to generate data sets for numerous research. Additionally, developed error model for vehicle detection algorithms can be used to identify LiDAR performance with respect to different parameters like LiDAR height, range, and laser point density. A periodic beam index map is developed by capturing antenna azimuth and elevation angles which denote maximum Reference Signal Receive Power (RSRP) for a simulated receiver grid on the road and classifying areas using Support Vector Machine (SVM) algorithm to reduce the number of Synchronization Signal Blocks (SSBs) that needed to send in Vehicle to Infrastructure (V2I) communication.

The mmWave communication up to 71 GHz is already specified in 3rd generation partnership project (3GPP)5G New Radio (NR), and communication in sub-THz bands is being studied for 6G widely in the academia and industry. Operation with very narrow beamwidths and much higher bandwidths in contrast to Frequency Range 1 (sub-6 GHz) can cater to the high data rate requirements at the expense of extra signal processing burden to overcome the unfavourable conditions such as high attenuation and scattering in the presence of obstacles. Such severe signal power attenuation caused by an obstacle may degrade the network performance due to link failures occurring as a result of line-of-sight (LoS) to non-LoS (NLoS) transitions. These limitations raise the necessity of a sensing system to collect situational awareness data to assist the wireless communication network. This work proposes a method to improve the LoS detection and user localization accuracy using multiple light detection and ranging (LiDAR) sensors co-located in access points (APs). We also propose an approach to predict the LoS transitions based on static LiDAR maps and the proposed method detected the LoS transition 400ms before its occurrence.

Automating processes with the increased use of robots is one of the key vertical applications enabled by 6G. Sensing the surrounding environment, localization and communication become crucial factors for these robots to operate. Light detection and ranging (LiDAR) has emerged as an appropriate method for sensing due to its capability generating detail-rich positional information with high accuracy. However, LiDARs are power-hungry devices that generate bulk amounts of data, limiting their use as on-board sensors in robots. In this paper, we present a novel approach to the methodology of generating an enhanced 3D map with improved field-of-view using multiple LiDAR sensors. This offloads the sensing burden from robots to the infrastructure where a centralized communication network will establish localization. We utilize an inherent property of LiDAR point clouds; point rings with Inertial Measurement Unit (IMU) data embedded in the sensor for point cloud registration. The generated 3D point cloud map has an accuracy of 10 cm compared to the real-world measurements. We also carry out a proof of concept design of the proposed method using two LiDAR sensors fixed in the infrastructure at elevated positions. This extends to an application where a robot is navigated through the mapped environment using a wireless link with minimal support from the on-board sensors. Our results further validate the idea of using multiple elevated LiDARs as a part of the infrastructure for various localization applications.

5G New Radio (NR) mmWave operates with narrow beams. Beam-based connections require careful management of beams to ensure a reliable connection, specially when the user has mobility. 5G NR defines beam management procedures to achieve this, at the expense of periodic reporting with increased overheads and resource usage. Concurrently, recent interest in sensing for assisting wireless systems provides an opportunity to extract situational awareness information which can aid in proactive decisions for the network. In this work, we utilize an infrastructure-mounted light detection and ranging (LiDAR) sensor system simultaneously operating with the wireless system to predict future beam decisions. A recurrent neural network (RNN) based learning model is proposed for the beam prediction, employing tracking information of users facilitated by the LiDARs and beam sequence information from the wireless system. Furthermore, a method for predictive beam management with increased periodicity of the reporting mechanism and aperiodic reporting is analyzed. The results for the considered scenario reveal 86.8% of the resources can be saved compared to the conventional beam reporting procedure, while achieving an 88.7% accuracy for optimal beam decisions.

Leveraging higher frequencies up to THz band paves the way towards a faster network in the next generation of wireless communications. However, such shorter wavelengths are susceptible to higher scattering and path loss forcing the link to depend predominantly on the line-of-sight (LOS) path. Dynamic movement of humans has been identified as a major source of blockages to such LOS links. In this work, we aim to overcome this challenge by predicting human blockages to the LOS link enabling the transmitter to anticipate the blockage and act intelligently. We propose an end-to-end system of infrastructure-mounted LiDAR sensors to capture the dynamics of the communication environment visually, process the data with deep learning and ray casting techniques to predict future blockages. Experiments indicate that the system achieves an accuracy of 87% predicting the upcoming blockages while maintaining a precision of 78% and a recall of 79% for a window of 300 ms.

Three-dimensional space-time velocity filters may be used to enhance dynamic passband objects of interest in videos while attenuating moving interfering objects based on their velocities. In this paper, we show that the attenuation of interfering stopband objects may be significantly improved using recently proposed shifted-velocity filters. It is shown that an improvement of approximately 20 dB in signal-to-interference ratio may be achieved for stopband to passband velocity differences of only 1 pixels/frame. More importantly, this improvement is achieved without increasing the computational complexity.