In this paper, we propose a hybrid precoding/combining framework for communication-centric integrated sensing and full-duplex (FD) communication operating at mmWave bands. The designed precoders and combiners enable multiuser (MU) FD communication while simultaneously supporting monostatic sensing in a frequency-selective setting. The joint design of precoders and combiners involves the mitigation of self-interference (SI) caused by simultaneous transmission and reception at the FD base station (BS). Additionally, MU interference needs to be handled by the precoder/combiner design. The resulting optimization problem involves non-convex constraints since hybrid analog/digital architectures utilize networks of phase shifters. To solve the proposed problem, we separate the optimization of each precoder/combiner, and design each one of them while fixing the others. The precoders at the FD BS are designed by reformulating the communication and sensing constraints as signal-to-leakage-plus-noise ratio (SLNR) maximization problems that consider SI and MU interference as leakage. Furthermore, we design the frequency-flat analog combiner such that the residual SI at the FD BS is minimized under communication and sensing gain constraints. Finally, we design an interference-aware digital combining stage that separates MU signals and target reflections. The communication performance and sensing results show that the proposed framework efficiently supports both functionalities simultaneously.

Convolutional neural networks (CNNs) have demonstrated strong potential for improving channel estimation (CE) accuracy; however, their high computational complexity poses a challenge for real-world deployment. This paper presents DdUnet, a lightweight dual-domain U-shaped convolutional neural network tailored for CE in the physical uplink shared channel (PUSCH). DdUnet employs a multi-stage U-shaped architecture with PUSCH-specific downsampling and upsampling schemes to reduce computational complexity while enhancing model capacity. For scenarios involving large numbers of resource blocks (RB), a lightweight sparsity-aware CE solution that estimates the channels in the delay domain is proposed to further reduce complexity with minimal performance degradation. Additionally, a mixed RB training strategy is proposed to improve model generalization across varying RB numbers, a common scenario in PUSCH. Extensive evaluations across different signal-to-noise-ratios levels, RB numbers, and channel profiles show that DdUnet consistently outperforms other CNN-based CE solutions in both estimation performance and computational efficiency.

We propose an integrated sensing and communications (ISAC) framework that supports chirp signal transmission in CP-OFDM-based multiple access communication systems, enabling efficient coexistence of communication and sensing capabilities. Our framework employs the discrete Fourier transform phase rotated and permuted frequency division multiple access (DFT-p-FDMA) waveform to transmit chirp signals using a portion of the frequency resources, while ensuring interference-free concurrent CP-OFDM data transmissions on other bands. We analyze the effective channel behavior under the DFT-p-FDMA waveform, characterizing how delays and Doppler shifts impact radar target echoes. We also show how processing multiple received symbols improves Doppler resolution in practical scenarios. Our framework allows flexible adjustment of range-Doppler resolution through optimized time-frequency resource allocation, offering a versatile solution for ISAC applications. Simulation results validate the framework's performance in delay and Doppler estimation, highlighting its potential to support ISAC in next-generation wireless networks.

Artificial intelligence (AI)-based channel state information (CSI) feedback has emerged as a promising alternative to conventional codebook-based methods in multiple-input multiple-output system. However, its performance deteriorates under practical channel estimation errors caused by interference, noise, and hardware impairments, resulting in inaccurate eigenvector reconstruction and degraded digital beamforming. To address this issue, we propose a novel 3rd Generation Partnership Project (3GPP)-compliant noise-robust denoising encoder (D-Enc) framework that trains the encoder using paired noisy and ideal data (eigenvectors and latent representations) to learn a denoising mapping while maintaining compatibility with inter-vendor training collaboration options agreed in 3GPP Release 19 without modifying the network-side decoder. System-level evaluations demonstrate that D-Enc significantly improves robustness, eigenvector reconstruction accuracy and 5th percentile downlink user throughput over legacy AI methods.

User Equipment (UE) power saving is critical to extending battery life and improving user experience. Traditionally, most configurations including power saving are left to the network (NW). This design is well-suited for early-generation devices with limited capabilities, as NW possesses higher compute power to make better decisions. Modern devices such as smartphones, however, have plenty of compute and memory resources, which could learn the usage pattern and environmental constraints. This article introduces on-device learning that can scale better than NW-side solutions in certain scenarios and with less privacy concerns. We demonstrate the feasibility of the approach with two examples based on real-world data. We also share several research directions. We hope this will help pave the way for more research to exploit UE's intelligence to save power and optimize UE performance in general.

Accurate downlink channel state information (CSI) in the next-generation NodeB (gNB) is essential to enable high-performance precoding in massive multiple input multiple output (MIMO) systems. Many existing works have utilized deep learning for efficient downlink CSI feedback from user equipment (UE) in massive MIMO. However, in practice, UEs can only estimate partial downlink CSI from sparsely distributed CSI-RS. To address the deleterious effects of sparse CSI-RS and UE feedback delay on downlink CSI estimation, we leverage the strong correlation between magnitudes of uplink-downlink CSIs and the dense CSI available from common downlink signals of PBCH/PSS/SSS. We generalize a classic phase retrieval framework to propose a deep learning-enhanced Gerchberg–Saxton (GS) algorithm that reconstructs dense downlink CSI from sparse RB-domain feedback. We develop a neural network architecture that unfolds the iterative GS process to effectively improve phase estimation accuracy and robustness. Simulation results demonstrate that the proposed method significantly outperforms conventional and baseline learning-based models in terms of both normalized mean square error and achievable rate under various downsampling scenarios.

Two methods are proposed to reduce the complexity of the belief propagation (BP) decoding for low-density parity-check (LDPC) codes, with the focus on the layered decoding for quasi-cyclic LDPC (QC-LDPC) codes. The two methods can be separately incorporated into the layered decoding procedures and combined seamlessly to provide the two-fold complexity reduction. In the first method, the message updates regarding a subset of variable nodes are omitted. To construct the subset, the most reliable bit-level channels, derived from M-ary input channels, are identified. Then, we select the variable nodes with initial log-likelihood ratios (LLRs) exceeding a threshold among the ones assigned to the most reliable bit-level channels to form the subset. The threshold is designed and adjusted based on characteristics of the most reliable bit-level channels. Our numerical results demonstrate that the first method significantly reduces the complexity with negligibly impacting the error-correcting performance. In the second method, each layer of base graphs of QC-LDPC codes uses either sum-product algorithm (SPA) or min-sum algorithm (MSA) to update LLRs. A degree-based algorithm is presented to determine the layers employing either SPA or MSA. We compare with two intuitive methods: 1) applying SPA to randomly selected layers and MSA to the remaining layers, and 2) applying SPA to the core layers of 5G NR LDPC codes and MSA to the remaining layers. According to the numerical results, the degree-based algorithm maximally reduces the complexity while maintaining the same performance. To evaluate the joint method, the first method is applied on top of the second method. Further complexity reduction is achieved without degrading the performance of the second method.

UE feedback of downlink precoder to the base station (gNB) is important for achieving high spectrum efficiency in massive MIMO wireless networks. Considering the delay of precoding matrix (PM) feedback, rapid channel change due to UE mobility can lead to a mismatch between PM from UE feedback and the current downlink channel. To mitigate performance degradation due to such outdated PM feedback, we propose a profile-based deep learning framework for PM forecast by incorporating radio environment information. Our method models PMs as temporally correlated sequential frames across Transmission Time Intervals (TTIs) and designs a gNB-side DL network to forecast the current PM based on historical feedback. To improve adaptability and generalization, we extract Doppler and multipath delay spread variables to implement environment-aware clustering and to customize a DL model for each cluster. The gNB selects the appropriate DL model in real time according to the current propagation profile. Test results under a Uniform Planar Array (UPA) configuration demonstrate the performance enhancement. This framework provides a scalable and practical solution for robust precoding in dynamic and high-mobility radio environments.

Large language models (LLMs) have shown strong potential across a variety of tasks, but their application in the telecom field remains challenging due to domain complexity, evolving standards, and specialized terminology. Therefore, general-domain LLMs may struggle to provide accurate and reliable outputs in this context, leading to increased hallucinations and reduced utility in telecom operations. To address these limitations, this work introduces KG-RAG—a novel framework that integrates knowledge graphs (KGs) with retrieval-augmented generation (RAG) to enhance LLMs for telecom-specific tasks. In particular, the KG provides a structured representation of domain knowledge derived from telecom standards and technical documents, while RAG enables dynamic retrieval of relevant facts to ground the model’s outputs. Such a combination improves factual accuracy, reduces hallucination, and ensures compliance with telecom specifications. Experimental results across benchmark datasets demonstrate that KG-RAG outperforms both LLM-only and standard RAG baselines, e.g., KG-RAG achieves an average accuracy improvement of 14.3% over RAG and 21.6% over LLM-only models. These results highlight KG-RAG’s effectiveness in producing accurate, reliable, and explainable outputs in complex telecom scenarios.

The uplink of wireless communication networks is becoming increasingly critical to support emerging user-initiated applications, yet faces significant challenges due to limited user transmission power and highly dynamic environment. While artificial intelligence (AI) techniques have shown considerable promise for enhancing physical-layer performance in simulation-based studies, there remains a critical gap with very few prototype implementations validating AI effectiveness under realistic conditions. This paper presents a comprehensive study of AI-based channel estimation for Physical Uplink Shared Channel (PUSCH) reception, implemented and validated in a real-world testbed. Our demonstration at Mobile World Congress 2025 achieved more than 30% system throughput gain compared to conventional channel estimation methods. This paper reveals the end-to-end implementation embracing the practical challenges and design considerations, showcasing potential impact of AI in a realistic network, offering valuable insights for bridging the gap between simulation-based research and real-world deployment.

Low-altitude uncrewed aerial vehicles (UAVs) can pose growing risks to airspace safety, security, and privacy. Cellular infrastructure can passively sense them without dedicated radar hardware by exploiting integrated sensing and communication (ISAC) technology. Most prior work exploits monostatic sensing or bistatic/multistatic configurations based on downlink measurements. To the best of our knowledge, this paper presents the first uplink framework, where multiple user equipments (UEs) transmit sounding reference signal (SRS) pilots and the base station (BS) receives the UAV-scattered echoes. Sensing from uplink SRS, however, introduces new challenges. Each UE has its own oscillator and timing loop, so the channel estimate at the BS carries residual timing, frequency, and amplitude impairments that corrupt the UAV delay and Doppler. Moreover, the UAV echo is weaker than both the line-of-sight (LOS) path and urban clutter, so detection from a single UE transmission is not reliable. We address these challenges by designing a LOS-referenced synchronization scheme and a joint detector. The synchronization reuses the existing timing advance (TA) command and an adjacent-occasion conjugate product to remove the residuals without additional signaling. Then the detector searches a shared 3D state space and accumulates evidence across UEs. It leverages a normalized contrast that exploits the bistatic geometry. We evaluate the framework in a cluttered urban scene at frequency range 1 (FR1) with four pedestrian UEs and a 100 MHz 5G New Radio (NR) waveform. The proposed pipeline achieves sub-nanosecond synchronization and a 4.84 m median 3D position error.

Large language models (LLMs) have received considerable attention recently due to their outstanding comprehension and reasoning capabilities, leading to great progress in many fields. The advancement of LLM techniques also offers promising opportunities to automate many tasks in the telecommunication (telecom) field. After pre-training and fine-tuning, LLMs can perform diverse downstream tasks based on human instructions, paving the way to artificial general intelligence (AGI)-enabled 6G. Given the great potential of LLM technologies, this work aims to provide a comprehensive overview of LLM-enabled telecom networks. In particular, we first present LLM fundamentals, including model architecture, pre-training, fine-tuning, inference and utilization, model evaluation, and telecom deployment. Then, we introduce LLM-enabled key techniques and telecom applications in terms of generation, classification, optimization, and prediction problems. Specifically, the LLM-enabled generation applications include telecom domain knowledge, code, and network configuration generation. After that, the LLM-based classification applications involve network security, text, image, and traffic classification problems. Moreover, multiple LLM-enabled optimization techniques are introduced, such as automated reward function design for reinforcement learning and verbal reinforcement learning. Furthermore, for LLM-aided prediction problems, we discussed time-series prediction models and multi-modality prediction problems for telecom. Finally, we highlight the challenges and identify the future directions of LLM-enabled telecom networks.

Nonlinear self-interference (SI) cancellation is essential for mitigating the impact of transmitter-side nonlinearity on overall SI cancellation performance in flexible duplex systems, including in-band full-duplex (IBFD) and sub-band full-duplex (SBFD). Digital SI cancellation (SIC) must address the nonlinearity in the power amplifier (PA) and the in-phase/quadrature-phase (IQ) imbalance from up/down converters at the base station (BS), in addition to analog SIC. In environments with rich signal reflection paths, however, the required number of delayed taps for time-domain nonlinear SI cancellation increases exponentially with the number of multipaths, leading to excessive complexity. This paper introduces a novel, low-complexity, frequency domain nonlinear SIC, suitable for flexible duplex systems with multiple-input and multiple-output (MIMO) configurations. The key approach involves decomposing nonlinear SI into a nonlinear basis and categorizing them based on their effectiveness across any flexible duplex setting. The proposed algorithm is founded on our analytical results of intermodulation distortion (IMD) in the frequency domain and utilizes a specialized pilot sequence. This algorithm is directly applicable to orthogonal frequency division multiplexing (OFDM) multi-carrier systems and offers lower complexity than conventional digital SIC methods. Additionally, we assess the impact of the proposed SIC on flexible duplex systems through system-level simulation (SLS) using 3D ray-tracing and proof-of-concept (PoC) measurement.

This paper investigates a full-duplex (FD) multiple-input multiple-output (MIMO) setting for integrated sensing and communication (ISAC), which enables simultaneous monostatic sensing and communication with a single base station (BS). We consider frequency bands that exhibit sparse propagation characteristics, such as mmWave bands. A key challenge for these systems is the design of hybrid precoders and combiners that facilitate ISAC functionalities while mitigating self-interference (SI) that is caused by concurrent transmission and reception. While prior work has focused on hybrid precoder and combiner design for FD ISAC with prior communication channel and target parameter knowledge, the problem of joint channel and target parameter estimation remains unexplored. We address this gap by designing SI-aware hybrid training precoders and combiners that form a beam codebook optimized for sparse channel estimation. Our design minimizes the mutual coherence, a key metric in compressed sensing, while effectively suppressing the SI to enable accurate joint estimation. We evaluate our approach in terms of estimation accuracy, as well as SI mitigation performance.

We propose a machine learning (ML) based end-to-end framework for pilotless communications that consists of two key components. The first component is an asymmetric modulation constellation that enables pilotless communications under channel impairments. The second component is a neural network (NN) receiver featuring an architecture that has a core of several serially-connected ResNet-like blocks. The transmitter only sends data symbols (without any pilots), and the NN receiver enables pilotless communications by using the received data symbols from the asymmetric constellation to perform implicit channel estimation/compensation and generate log-likelihood ratios (LLRs) for the bits comprising the data symbols. The combination of the asymmetric modulation constellation and the NN receiver achieves similar or superior performance to a traditional zero-forcing (ZF) receiver that relies on pilot symbols for channel estimation for 64-ary and 256-ary modulations for channels with limited time and frequency selectivity.

Integrated sensing and communication (ISAC) is expected to be a key enabling technology for diversifying the use cases of future telecommunication systems. Recently, there has been an active push in the industry on ISAC, with several strides being made in standardization efforts in both Wi-Fi and cellular communication domains. This paper provides a comprehensive overview of the current efforts towards ISAC in Wi-Fi and cellular communication (5G and 6G), highlighting current research directions and standardization efforts. The paper also discusses the current advances in channel modeling used for evaluating ISAC systems, and the impact of the different transceiver hardware impairments on ISAC. We also discuss the limitations of current progress and summarize key research directions for the future.

CP-OFDM exhibits degraded performance in high mobility situations, primarily due to selective fading in both frequency and time. To address this issue, we propose a novel waveform, discrete Fourier transform phase rotated and permuted frequency division multiple access (DFT-p-FDMA), which efficiently handles a range of mobility scenarios. Unlike CP-OFDM, which assigns distinct modulation symbols to fixed subcarriers, DFT-p-FDMA spreads the modulation symbols in both time and frequency. By leveraging frequency-time spreading, DFT-p-FDMA captures both delay and Doppler diversity, resulting in a near-uniform effective SNR across the demodulated symbols. The DFT-p-FDMA waveform can be generated by applying a DFT and a permutation with phase rotation before performing the IDFT of CP-OFDM, and allows for UE-specific implementation on top of the CP-OFDM framework. Performance evaluation demonstrates that DFT-p-FDMA outperforms CP-OFDM by providing up to 2.5 dB gain at 1e-2 block error rate (BLER).

Large language models (LLMs) have made significant progress in general-purpose natural language processing tasks. However, LLMs are still facing challenges when applied to domain-specific areas like telecommunications, which demands specialized expertise and adaptability to evolving standards. This paper presents a novel framework that combines knowledge graph (KG) and retrieval-augmented generation (RAG) techniques to enhance LLM performance in the telecom domain. The framework leverages a KG to capture structured, domain-specific information about network protocols, standards, and other telecom-related entities, comprehensively representing their relationships. By integrating KG with RAG, LLMs can dynamically access and utilize the most relevant and up-to-date knowledge during response generation. This hybrid approach bridges the gap between structured knowledge representation and the generative capabilities of LLMs, significantly enhancing accuracy, adaptability, and domain-specific comprehension. Our results demonstrate the effectiveness of the KG-RAG framework in addressing complex technical queries with precision. The proposed KG-RAG model attained an accuracy of 88% for question answering tasks on a frequently used telecom-specific dataset, compared to 82% for the RAG-only and 48% for the LLM-only approaches.

In upcoming sixth-generation (6G) communications, sub-terahertz (sub-THz) bands will be employed to meet data rate requirements. For designing systems operating in that band, accurate modeling of the propagation channels, and in particular the multipath propagation, is vital. Measurements show that multipath components (MPCs) occur in clusters; incorporation of cluster structure is beneficial to make models more realistic as well as simpler. In this paper, we present a methodology for extracting cluster statistics from channel measurements that are noisy (and thus necessitate thresholding for noise suppression) and/or have limited dynamic range; this affects the relationships between intra-cluster decay constants and intra-cluster RMS spreads, as well as the observed number of clusters. To address this, we analytically derive the deviations between actual cluster parameters and their observable statistics, followed by validation through simulations. By solving the corresponding inverse problem using the derived expressions and the dynamic range, more realistic cluster parameters are retrieved. Moreover, the retrieved cluster parameters enable the extrapolation of channel characteristics across varying dynamic range conditions. The methodology is applicable at all frequency bands, and we demonstrate the impact of this modeling on the parameterization of ultra-wideband, double-directionally resolved sub-THz measurements of device-to-device (D2D) propagation in outdoor urban environments.

Generative artificial intelligence (GAI) is a promising technique towards 6G networks, and generative foundation models such as large language models (LLMs) have attracted considerable interest from academia and telecom industry. This work considers a novel edge-cloud deployment of foundation models in 6G networks. Specifically, it aims to minimize the service delay of foundation models by radio resource allocation and task offloading, i.e., offloading diverse content generation tasks to proper LLMs at the network edge or cloud. In particular, we first introduce the communication system model, i.e., allocating radio resources and calculating link capacity to support generated content transmission, and then we present the LLM inference model to calculate the delay of content generation. After that, we propose a novel in-context learning method to optimize the task offloading decisions. It utilizes LLM's inference capabilities, and avoids the difficulty of dedicated model training or fine-tuning as in conventional machine learning algorithms. Finally, the simulations demonstrate that the proposed edge-cloud deployment and in-context learning task offloading method can achieve satisfactory generation service quality without dedicated model training or fine-tuning.