Ultra-Reliable and Low-Latency Communication (URLLC) requires stable transmission under strict latency and high-reliability constraints. Consequently, passive retransmission, which merely waits for bit errors to occur, cannot meet these stringent control and security requirements. This paper proposes an active feedback mechanism, namely the Transmission Prediction Feedback Network (TPFNet). It leverages trajectory-aware modeling and post-decoding statistics to perform risk assessment and guide strategy switching, thereby proactively managing subsequent retransmissions. The TPFNet concept is based on utilizing the historical progression of symbols on the constellation diagram as an initial reference, integrating memory-augmented trajectory-aware modeling. This approach ensures that decision-making is not confined to the instantaneous state received at a single point in time. Initially, it uses the in-phase and quadrature path trajectories of the received sequence to jointly assess instantaneous deviations and temporally cumulative morphological changes. Secondly, it employs residual regression for the calibration of symbol positions, complemented by a concentric constraint loss to mitigate shrinkage bias toward the constellation origin. Finally, it provides a confidence metric for the transceiver link, which facilitates adaptive modulation and coding scheme switching. Simulation results demonstrate that the proposed method significantly enhances transmission fidelity under diverse channel conditions. It markedly improves the block error rate and undetected error rate while enhancing the robustness of symbol decisions at the boundaries.

With advancements in wireless communication and localization technologies, cellular networks are evolving towards integrated sensing and communication (ISAC) capabilities. To address the challenges of sensing-assisted communication, we introduce the digital twin of channel (DToC). Specifically, locations of user terminals (UTs) and their statistical channel state information (sCSI) are treated as physical objects and virtual counterparts in the concept of digital twin (DT), respectively. In this work, we establish a probabilistic model that characterizes sCSI as a location-conditioned distribution. To enable precise sCSI generation, we enhance the latent diffusion model (LDM) and propose an improved latent diffusion model (ILDM) with deterministic sampling. We further propose an accelerated LDM method to speed up the generation process by skipping certain sampling steps. Simulation results demonstrate that the proposed ILDM achieves high accuracy in generating sCSI, while the accelerated LDM delivers significant speedups with minor performance degradation. Our results also validate that the DToC framework can effectively generate sCSI without pilot overhead.

With the growth of positioning demand, fingerprintbased positioning, which utilizes multi-path information to improve positioning accuracy, has attracted academic and industrial attentions. Current position fingerprints often incur significant storage overhead and time complexity for machine learning due to their large dimension. This paper investigates positioning in massive multiple-input multiple-output (MIMO) communication systems using a low dimensional fingerprints called the angledelay power coordinates matrix (ADECM). We first use a dimensionality reduction algorithm for the widely used position fingerprint angle-delay channel power matrix (ADCPM) in existing works, and define the reduced matrix as the ADECM. Subsequently, recognizing the excellent performance of generative adversarial network in regression tasks, we propose the Feature Separation Generative Adversarial Network (FSGAN). This innovative network is specifically designed to accurately estimate the position of user terminals. Simulation results show that the proposed positioning method outperforms existing methods.

Low Earth Orbit (LEO) satellite communication is crucial for developing sixth-generation (6G) networks. The integration of massive multiple-input multiple-output (MIMO) technology is being actively researched to enhance the performance of LEO satellite communication systems. However, the limited power resources of LEO satellites pose significant challenges to improving energy efficiency (EE) under power-constrained conditions. Typical optimization-based methods often lack real-time adaptability and computational efficiency. This paper proposes innovative solutions to address the challenges of precoding in massive MIMO LEO satellite communications. Specifically, we introduce a deep unfolding of the Dinkelbach algorithm and the weighted minimum mean square error (WMMSE) approach to achieve enhanced EE. This transformation of iterative optimization procedures into a graph neural network (GNN) leads to faster convergence and improved computational efficiency. Furthermore, we apply the Taylor expansion method to approximate matrix inversion within the GNN framework. Numerical experiments demonstrate the superiority of our proposed method in terms of complexity and robustness, achieving significant improvements over other state-of-the-art methods.

Multi-frequency massive multi-input multi-output (MIMO) communication is a promising strategy for both 5G and future 6G systems, ensuring reliable transmission while enhancing frequency resource utilization. Statistical channel state information (CSI) has been widely adopted in multi-frequency massive MIMO transmissions to reduce overhead and improve transmission performance. In this paper, we propose efficient and accurate methods for obtaining statistical CSI in multi-frequency massive MIMO systems. First, we introduce a multi-frequency massive MIMO channel model and analyze the mapping relationship between two types of statistical CSI, namely the angular power spectrum (APS) and the spatial covariance matrix, along with their correlation across different frequency bands. Next, we propose an autoregressive (AR) method to predict the spatial covariance matrix of any frequency band based on that of another frequency band. Furthermore, we emphasize that channels across different frequency bands share similar APS characteristics. Leveraging the maximum entropy (ME) criterion, we develop a low-complexity algorithm for high-resolution APS estimation. Simulation results validate the effectiveness of the AR-based covariance prediction method and demonstrate the highresolution estimation capability of the ME-based approach. Furthermore, we demonstrate the effectiveness of multi-frequency cooperative transmission by applying the proposed methods to obtain statistical CSI from low-frequency bands and utilizing it for high-frequency channel transmission. This approach significantly enhances high-frequency transmission performance while effectively reducing system overhead.

Low Earth Orbit (LEO) satellite communication is a critical component in the development of sixth generation (6G) networks. The integration of massive multiple-input multipleoutput (MIMO) technology is being actively explored to enhance the performance of LEO satellite communications. However, the limited power of LEO satellites poses a significant challenge in improving communication energy efficiency (EE) under constrained power conditions. Artificial intelligence (AI) methods are increasingly recognized as promising solutions for optimizing energy consumption while enhancing system performance, thus enabling more efficient and sustainable communications. This paper proposes approaches to address the challenges associated with precoding in massive MIMO LEO satellite communications. First, we introduce an end-to-end graph neural network (GNN) framework that effectively reduces the computational complexity of traditional precoding methods. Next, we introduce a deep unfolding of the Dinkelbach algorithm and the weighted minimum mean square error (WMMSE) approach to achieve enhanced EE, transforming iterative optimization processes into a structured neural network, thereby improving convergence speed and computational efficiency. Furthermore, we incorporate the Taylor expansion method to approximate matrix inversion within the GNN, enhancing both the interpretability and performance of the proposed method. Numerical experiments demonstrate the validity of our proposed method in terms of complexity and robustness, achieving significant improvements over state-of-the-art methods.

With the advancement of communication technology and the improvement of localization accuracy, cellular networks are gradually evolving from communication to perception-integrated networks. Addressing the research challenges of sensing-assisted communication, we propose, for the first time, the concept of Digital Twin of Channel (DToC). Specifically, we regard user terminal (UT) positions as physical objects, and statistical channel state information (CSI) as virtual digital objects. Observing the change trend of UTs’ statistical CSI caused by the changes of UT’s physical position enables predictive analytics for subsequent communication tasks. Then, we establish the relationship between physical and virtual digital objects using a Diffusion Model (DM) to achieve the DToC. Indeed, the DM can generate the desired objects by gradually denoising from noisy data using neural networks. Furthermore, we propose a conditional DM utilizing UTs’ positions, which completes the task of generating the corresponding statistical CSI under known user-specific position conditions, thus mapping UT positions to statistical CSI. Simulation results demonstrate that our DToC framework outperforms previous statistical CSI estimation methods. Without the need of pilots, our method can simultaneously generate statistical CSIs from a large number of UTs’ positions, achieving satisfactory results.

Wireless positioning is crucial for Internet of Things (IoT) landscape, enhancing precision and reliability in location-based services. This article addresses the challenges of existing massive multiple-input–multiple-output fingerprint positioning methods, which typically require accurate channel estimation and one-by-one labeled data sets. We propose a semisupervised representation contrastive learning technique that leverages a partially labeled received pilot signal data set readily available from the base station. Our approach employs data augmentation to generate a large number of positive and negative sample pairs, which are then used to pretrain an encoder with a contrastive loss function in the self-supervision way. During pretraining, the encoder learns to encode positive samples close to an anchor, while keeping negative samples far away in the representation space. A fully connected layer is added on top of the encoder for position regression, and the encoder and regression networks are fine-tuned with a small labeled subdataset for the downstream positioning task. Simulation results demonstrate that our pretraining and fine-tuning approach outperforms the previous methods, significantly improving positioning accuracy, avoiding exact channel estimation and achieving labeling efficiency.