The multiple-input multiple-output dual functional radar communication (MIMO-DFRC) system is a promising platform for future integrated sensing and communication applications. Ensuring reliable performance of both radar and communication functions, the beam selection is a critical technology in MIMO-DFRC systems. However, the beam selection problem is known to be NP-hard, and efficiently addressing it remains an open issue, especially in distributed systems. In this paper, we address the beam selection problem for a MIMO-DFRC system by formulating it as a semi-Markov decision process and propose a novel hierarchical reinforcement learning (HRL) algorithm. In our approach, codebook-based beam selection for transmitting and receiving BS is controlled by an agent deployed in the cloud. Inspired by the mechanism of hierarchical codebook beam training, we employ an option-based policy that enables the agent to explore different layers of the codebook and extract context information across multiple discrete time steps. We utilize an invalid action masking technique to overcome the dynamic action space problem caused by the option-based policy. Simulation results demonstrate that the HRL-based algorithm outperforms existing beam selection methods and achieves remarkable performance even under conditions of a high probability of false alarm and low signal-to-noise ratio. Furthermore, we find that the proposed algorithm exhibits promising capabilities to learn a more efficient policy beyond the full hierarchical codebook training trajectory.

The deployment of extremely large-scale MIMO (XL-MIMO) and short-wavelength signaling enhances communication capabilities and improves spectrum efficiency for future sixth-generation (6G) wireless communication. However, users may potentially be located in the near-field region due to the sharp increase in antenna array aperture. In the near-field region, the signal wave is spherical wave. Thus, the consideration of spatial angle and distance requires the development of novel channel estimation algorithms to reduce codebook overhead. This paper develops a novel scheme based on a low-size adaptive codebook to reconstruct the near-field channel. Initially, it is investigated that the angle spread for one channel path component is confined to a certain angular spatial region, which demonstrates the sparsity inherent in angular domain. Exploiting the angular sparsity inherent, we propose a novel adaptive joint sparse Bayesian learning (JSBL) estimation algorithm on all subcarriers to cater to reduce the codebook size. The proposed algorithm captures all spatial angular sparse information and then refines distance information so that the measurement codebook size only depends on the spatial angular resolution. Further, the proposed adaptive JSBL approach is extended to estimate the time-varying near-field channel. Moreover, Bayesian Cramér-Rao Bounds (BCRBs) are derived for quasi-static and temporal scenarios. Numerical simulations are presented to demonstrate that our approaches with low codebook overhead outperform other algorithms based on the angular-domain and polar-domain codebooks.

The deployment of extremely large-scale arrays and high-frequency signaling holds the potential to enhance communication capabilities and improve spectrum efficiency. However, channel estimation faces challenges due to the simultaneous consideration of spatial angles and distances, leading to storage constraints and energy spread. To cope with this issue, we analyze the sparsity inherent in beamspace domain representation and introduce an adaptive codebook scheme for extremely large-scale massive MIMO (XL-MIMO) channels. In this work, we transform multi-band channel estimation to sparse matrix recovery problem. Then, a novel adaptive joint sparse Bayesian learning algorithm is proposed to capture the angular-domain information and refine distance information iteratively without increasing codebook overhead for XL-MIMO channel estimation. Simulation results demonstrate our approach outperforms other algorithms based on the sampling angular-distance domain codebook with low codebook overhead.

Dual-functional radar-communication (DFRC) is a promising direction in the future integrated sensing and communication system. The joint radar and communication (JRC) beamforming scheme is recently developed in DFRC systems. To address the JRC beamforming challenge, conventional approaches predominantly rely on convex optimization methods, which severely depend on precise channel estimation and entail a high computational complexity. Motivated by this, a deep learning-based optimization approach is investigated for tackling the JRC beamforming problem. To enhance the overall performance, we design a deep alternating neural network architecture. Simulation results verify that our proposed method guarantees the required sensing performance and outperforms numerical algorithms in terms of the average data rate of communication users.

This paper concentrates on addressing the challenging problem of sensing channel estimation in cell-free integrated sensing and communication (ISAC) multiple-input multiple-output (MIMO) system. This challenge arises from the complex mixture of signals from both the direct sensing channel and target reflected sensing channel. To tackle this challenge, we introduce the digital twin (DT), as a powerful tool to exploit and characterize the inherent features of the target sensing channel by sampling data from the real world and interacting with it. To be specific, the DT model, designed as a generative adversarial network (GAN), is trained to be capable of generating the desired results from the coarse observations, where the distribution of the sensing channel in a particular cell-free ISAC system is implicitly learned via the adversarial process. With this basis, we propose a novel digital-twin-enabled channel estimation (DTE-CE) approach to enhance the performance of channel estimation, where the DTE-CE network is meticulously designed by utilizing the virtual channel matrix (VCM) model to facilitate the estimation process. Simulation results show the excellent performance of the proposed approach, as well as the effectiveness of our designed DTE-CE network, in terms of sensing channel estimation with different transmitting power and numbers of targets.

Cell-free massive multiple-input multiple-output (MIMO) is known for its potential to enhance overall system performance. Thanks to the principle of channel reciprocity, it becomes possible to implement downlink beamforming by exploiting the estimated uplink channel in time-division duplex (TDD) mode. However, the assumption of perfect hardware conditions, as made in prior studies, is not reflective of practical realities. The involvement of hardware impairments disrupts this reciprocity, resulting in performance degradation. This paper investigates the impact of hardware impairments in downlink data transmission, where a novel model is established by jointly considering the radio frequency (RF) mismatches and nonlinear distortions. We first derive closed-form achievable user rate expressions and prove that the impact of RF mismatches vanishes as the number of access points (APs)$M \to \infty $in certain distributions of RF gains. Then, we study the scenarios when the number of user equipments (UEs)$K \to \infty $, as well as various degrees of hardware impairments’ severity scaling M. Finally, we introduce a channel calibration process and theoretically derive its performance, observing that in certain scenarios, the need for calibration becomes redundant as$M \to \infty $. These findings are further validated through numerical results, confirming the scaling laws derived in our study.

Dual-functional radar-communication (DFRC) is a promising technology in future integrated sensing and communication systems. Since communication and sensing performance need to be taken into consideration for joint radar and communication (JRC) beamforming in the DFRC system, existing approaches mainly transform JRC beamforming problems into convex optimization problems and then solve them with classical convex solvers. These traditional solutions heavily rely on precise channel estimation and entail high computational complexity. In this paper, we investigate a deep learning-based optimization approach for JRC beamforming to enhance the spectral efficiency for communication users and guarantee the probability of detecting targets. To achieve better performance, we leverage the theoretical optimal structures of JRC beamforming and design an effective deep neural network architecture. To further reduce the computational burden in the training phase of neural network, we develope an improved orthogonal beamforming technique. Simulation results verify that our proposed algorithm guarantees the required sensing performance and outperforms numerical algorithms in terms of communication performance. The orthogonal beamforming technique achieves satisfactory performance with low computational complexity.

Channel estimation stands as a pivotal and challenging task for millimeter wave (mmWave) massive multiple-input multiple-output (MIMO) communication system, especially in a time-varying scenario, where exists a massive number of channel coefficients and severe propagation loss due to the Doppler shifts. Conventional estimation schemes may fail to track the fast varying channels and not be able to fully exploit the unique characteristics of mmWave channels in their model designs. In this work, we leverage the Generative Adversarial Networks (GANs) and meticulously design a novel framework named Homogeneous Denoising Generative Adversarial Network (HDnGAN) to tackle the challenge of time-varying channel estimation for mmWave MIMO system. Our framework incorporates the distinctive traits of mmWave channels, such as temporal and spatial correlations, as well as angular sparsity, into the network architecture design. Theoretically, a special case of our proposed HDnGAN with a linear structure is demonstrated to be not inferior to the linear minimum mean squared error (LMMSE) estimator. Numerical simulations underscore the superiority of HDnGAN over existing channel estimation methods, particularly in low signal-to-noise ratio (SNR) regions. Furthermore, it exhibits robustness across varying scenarios. Notably, it remains applicable in out-of-distribution situations and in the absence of ground truth.

Large antenna arrays and beamforming are necessary for the mmWave communication system, resulting in heavy time and energy consumption in the beam training stage. Therefore, dual-band operations are expected to be deployed in future communication systems, where low-frequency channels are used to meet basic communication needs, and millimeter wave (mmWave) channels are exploited when the high-rate transmission is required. Existing works utilize deep learning methods to extract low-frequency channel state information (CSI) to reduce the mmWave beam training overheads. However, an important limitation of deep learning approaches is that the model is usually trained in a given environment. When employed in an unseen environment, it usually requires a large amount of data to retrain. In this paper, a model-agnostic optimization algorithm based on meta-learning is proposed to provide a general mmWave beam prediction model. This model can be deployed to edge base stations and effectively adapted to the environment without the need for a heavy collection of data. Simulation results demonstrate that the proposed approach could reduce the model adaptation overheads. The meta-learning-based beam prediction model is robust and achieves high prediction accuracy and spectral efficiency in different signal-to-noise ratio (SNR) regimes.

Federated learning (FL) is an emerging paradigm for distributed machine learning that uses the data and the computational power of user devices while maintaining user privacy (e.g., position and motion track). It has been proved a promising way to help the learning-based millimeter wave (mmWave) system achieve efficient link configuration. However, FL systems have an inherent vulnerability to backdoor attacks during training, and this has not received attention in current FL-based beam selection research. The goal of a backdoor attacker is to implant a backdoor in the model such that at test time, the model will mispredict a certain family of inputs, and corrupt the performance of the trained model on specific sub-tasks. We study backdoor attacks in an FL-based beam selection system based on a deep neural network that utilizes user location information. Specifically, we propose a backdoor attack scheme that can be configured in the real world. The attacker’s trigger is an obstacle placed in certain locations. When the model encounters an input with these obstacles, the backdoor will be triggered, and the model will output the beam specified by the attacker. Through experiments, we show that the proposed attack can achieve a high attack success rate in a system without a defense mechanism. Moreover, we show that the traditional norm-clipping defense method cannot effectively defend against our attack. Furthermore, we propose a new backdoor attack defense method and verify the effectiveness of this scheme through experiments. In addition, we propose a backdoor detection method: the federated noise titration method, which can diagnose whether the model has a backdoor. Overall, our work explored backdoor attacks, defenses, and detection of the FL-based mmWave beam selection system.