In this paper, we propose an integrated sensing and communication (ISAC) system enabled by movable antennas (MAs), where the base station (BS) transmitter is equipped with MAs to enhance both sensing and communication performance. To characterize the benefits of MA-enabled ISAC systems, we focus on the line-of-sight (LoS) channel scenario and derive the Cramér-Rao bound (CRB) for angle estimation error, which is then minimized by jointly optimizing the antenna position vector (APV) and beamforming design, subject to a pre-defined signal-to-noise ratio (SNR) constraint to ensure the communication performance. Despite the non-convexity of the resulting problem, we develop a boundary traversal breadth-first search (BT-BFS) algorithm to obtain the global optimal solution, along with a lower-complexity boundary traversal depth-first search (BT-DFS) algorithm to find a local optimal solution efficiently. Extensive numerical results are presented to verify the effectiveness of the proposed algorithms, and demonstrate the superiority of the considered MA-enabled ISAC system over conventional ISAC systems with fixed-position antennas (FPAs).

Integrated sensing and communication (ISAC) in multi-hop wireless networks is a key technology for supporting a wide range of emerging Internet of Things (IoT) applications, addressing challenges such as spectrum scarcity and limited network coverage. To investigate the performance trade-off between end-to-end communication and sensing in these networks, this paper focuses on maximizing the end-to-end communication rate while ensuring the overall sensing performance in multiple-input multiple-output (MIMO) ISAC multi-hop wireless networks. In order to achieve this, we formulate a mixed-integer nonlinear programming (MINLP) problem that is highly non-convex and difficult to solve directly. To address this difficulty, we first transform the MINLP problem into a more tractable form through a series of equivalent transformations. We then propose an efficient algorithm based on generalized Benders decomposition (GBD) to solve the transformed problem optimally. Finally, numerical results demonstrate that the proposed algorithm achieves the optimal performance obtained by the exhaustive search method but with much lower complexity.

The Barrage Relay Network (BRN) is a mobile ad hoc network architecture designed for tactical edge communication, emphasizing robustness and low-latency data delivery. In BRNs, unicast transmission is confined to a Controlled Barrage Region (CBR) established via cooperative communication. However, dynamic channel conditions can impair the successful formation of CBRs, thereby reducing transmission reliability. Increasing the excess width parameter can improve connectivity but at the cost of significant node redundancy. To address this tradeoff, we propose a deep Q-network (DQN)-based relay selection algorithm utilizing a multi-step temporal difference (TD) target. The proposed scheme aims to minimize relay node usage while preserving CBR reliability. To enable effective decision-making, an extended routing state is introduced, along with two specialized routing packets that facilitate efficient information dissemination. The action space, state space, and reward function are carefully defined to incorporate both local and global network state information. Simulation results show that the proposed method reduces the number of active relay nodes by 66.67%, while maintaining a 100% packet delivery ratio. Furthermore, the algorithm demonstrates strong adaptability across varying network topologies.

Barrage Relay Networks (BRNs), a subclass of mobile ad hoc networks (MANETs), are tailored for tactical edge environments where broadcast-based communication is pre-dominant. By establishing Controlled Barrage Regions (CBRs), BRNs enable spatial reuse and facilitate pipeline-based packet forwarding. This paper investigates unicast transmission within a formed CBR, where a source node generates status updates at fixed intervals and relays them to a destination node via a multihop network. The timeliness of update delivery, quantified by the Age of Information (AoI), is analyzed using a stochastic hybrid system (SHS) framework. Additionally, a version age analysis is conducted for a parallel relay transmission topology, highlighting the tradeoff between information freshness and node utilization.

Due to network delays and scalability limitations, clustered ad hoc networks widely adopt Reinforcement Learning (RL) for on-demand resource allocation. Albeit its demonstrated agility, traditional Model-Free RL (MFRL) solutions struggle to tackle the huge action space, which generally explodes exponentially along with the number of resource allocation units, enduring low sampling efficiency and high computational complexity. To mitigate these limitations, Model-Based RL (MBRL) offers a solution by generating simulated samples through an environment model, which boosts sample efficiency and stabilizes the training by avoiding extensive real-world interactions. However, establishing an accurate dynamic model for complex and noisy environments necessitates a careful balance between model accuracy and computational complexity & stability. To address these issues, we propose a conditional Diffusion Model (DM) as high-dimensional offline resource allocation planner in multifrequency time division multiple access (MF-TDMA) wireless ad hoc networks. By leveraging the astonishing generative capability of DMs, our approach takes advantage of generated high-quality samples to guide exploration and learn optimal policy. Extensive experiments show that our model outperforms MFRL in average reward and Quality of Service (QoS) while demonstrating comparable performance to other MBRL algorithms.

The Wave Relay network represents an innovative large-scale wireless ad hoc network and has typical applications in tactical communication scenarios. This paper proposes an enhanced carrier sense multiple access with collision avoidance (CSMA/CA) strategy, integrated with multi-agent reinforcement learning (MARL), to improve network performance, focusing on communication latency and throughput. The state space, action space, and reward function are carefully designed to enable each agent to perform distributed deep Q-learning. This approach allows for adaptive and dynamic adjustments to the backoff parameters of the CSMA/CA strategy. Simulation results demonstrate that the proposed scheme effectively optimizes communication latency and throughput of the Wave Relay network.

Semantic communication (SemCom) has been deemed as a promising communication paradigm to break through the bottleneck of traditional communications. Nonetheless, most of the existing works focus more on point-to-point communication scenarios and its extension to multi-user scenarios is not that straightforward due to its cost-inefficiencies to directly scale the joint source-channel coding (JSCC) framework to the multi-user communication system. Meanwhile, previous methods optimize the system by differentiable bit-level supervision, easily leading to a “semantic gap”. Therefore, we delve into multi-user broadcast communication (BC) based on the universal transformer (UT) and propose a reinforcement learning (RL) based self-critical alternate learning (SCAL) algorithm, named SemanticBC-SCAL, to capably adapt to the different BC channels from one transmitter (TX) to multiple receivers (RXs) for sentence generation task. In particular, to enable stable optimization via a non-differentiable semantic metric, we regard sentence similarity as a reward and formulate this learning process as an RL problem. Considering the huge decision space, we adopt a lightweight but efficient self-critical supervision to guide the learning process. Meanwhile, an alternate learning mechanism is developed to provide cost-effective learning, in which the encoder and decoders are updated asynchronously as independent agents. Notably, the incorporation of RL makes SemanticBC-SCAL compliant with any user-defined semantic similarity metric and simultaneously addresses the channel non-differentiability issue by alternate learning. Besides, the convergence of SemanticBC-SCAL is also theoretically established. Extensive simulation results have been conducted to verify the effectiveness and superiorness of our approach, especially in low signal-to-noise ratio regions.

Due to the ability to provide superior error-correction performance, the successive cancellation list (SCL) algorithm is widely regarded as one of the most promising decoding algorithms for polar codes with short-to-moderate code lengths. However, the application of SCL decoding in low-latency communication scenarios is limited due to its sequential nature. To reduce the decoding latency, developing tailored fast and efficient list decoding algorithms of specific polar constituent codes (special nodes) is a promising solution. Recently, fast list decoding algorithms are proposed by considering special nodes with low code rates. Aiming to further speedup the SCL decoding, this paper presents fast list decoding algorithms for two types of high-rate special nodes, namely single-parity-check (SPC) nodes and sequence rate one or single-parity-check (SR1/SPC) nodes. In particular, we develop two classes of fast list decoding algorithms for these nodes, where the first class uses a sequential decoding procedure to yield decoding latency that is linear with the list size, and the second further parallelizes the decoding process by pre-determining the redundant candidate paths offline. Simulation results show that the proposed list decoding algorithms are able to achieve up to 70.7% lower decoding latency than state-of-the-art fast SCL decoders, while exhibiting the same error-correction performance.

In modern clustered ad hoc networks, efficient and dynamic resource allocation is crucial for ensuring Quality of Service (QoS) under dynamic and uncertain environments. However, the challenges posed by limited sample efficiency, high interaction cost, and high-dimensional action space limit the effectiveness of the widely adopted Model-Free Reinforcement Learning (MFRL) solutions. In contrast, Model-Based RL (MBRL) offers an alternative approach to boost sample efficiency and stabilize the training by explicitly leveraging a learned environment model. Nevertheless, designing accurate and stable dynamics models in noisy, real-world communication scenarios remains a key bottleneck. To address these issues, we propose a Conditional Diffusion Model Planner (CDMP) for high-dimensional offline resource allocation in clustered ad hoc networks. By leveraging the powerful generative capability of Diffusion Models (DMs), our approach enables the accurate modeling of complex environmental dynamics and utilizes an inverse dynamics model for effective policy planning. Beyond simply adopting DMs in offline RL, we further incorporate the CDMP algorithm with a theoretically guaranteed, uncertainty-aware penalty metric, which theoretically and empirically manifests itself in mitigating the Out-of-Distribution (OOD)-induced distributional shift, a common issue for offline settings with scarce training data. Extensive experiments also show that our model outperforms MFRL in average reward and QoS, while demonstrating superior performance over other MBRL algorithms. These results highlight the practicality and scalability of our model in real-world network resource allocation tasks.

Being able to provide excellent error-correction performance for polar codes with short-to-moderate code length, successive-cancellation list (SCL) is regarded as one of the most promising decoding algorithms. However, the application of SCL decoding in low-latency communication scenarios is limited due to its sequential nature. Recently, fast list decoding algorithms are proposed by considering special nodes with low code rates. Aiming at achieving further speedup for SCL decoding, this paper presents fast list decoding algorithms for two types of high-rate special nodes, namely single-parity-check (SPC) and sequence rate-1(SRI) nodes, based on the minimum-combinations set (MCS) which is able to significantly narrow the search space of candidate paths. Typically, SPC nodes can be directly decoded within one round of path splitting procedure, whereas SR1 nodes, as a group of parallel SPC nodes, can also be decode efficiently. Simulation results show that the proposed fast SCL decoder is able to reduce the decoding latency by 68.4% as compared to the state of the art, without any error-correction performance degradation.

Semantic broadcast communications (SemanticBC) for image and text transmission have achieved significant performance gains for single tasks. Nevertheless, extending these methods to a multi-task scenario presents challenges, as different tasks often require distinct objective functions, and the shared encoder must handle potential conflicts effectively. In this paper, we propose a tri-level multiple gradient descent algorithm (MGDA) based reinforcement learning (RL) approach MagicRL for multi-task SemanticBC to effectively balance the multi-task learning conflicts and serve multiple different tasks simultaneously at the receiver sides, including classification and content-reconstruction tasks. In particular, we provide optimized decoders with given encoder parameters by a self-critical RL approach. Subsequently, MGDA-based weight assignment is applied to balance multiple decoders and a first-order Frank-Wolfe algorithm efficiently solves the underlying quadratic programming problem. On this basis, the encoder gets improved through a proper weighted summation of multi-task objective functions. Extensive simulation results have been conducted to verify the effectiveness, especially under low signal-to-noise ratio.

In this paper, we consider a cooperative sensing framework in the context of future multi-functional network with both communication and sensing ability, where one base station (BS) serves as a sensing transmitter and several nearby BSs serve as sensing receivers. Each receiver receives the sensing signal reflected by the target and communicates with the fusion center (FC) through a backhaul-limited multiple access channel (MAC) for cooperative localization of the target. Different from schemes on only information domain or signal domain cooperation, we present a hybrid information-signal domain cooperative sensing (HISDCS) design, where each sensing receiver transmits both the estimated time delay/effective reflecting coefficient and the received sensing signal sampled around the estimated time delay to the FC. Then, we propose to minimize the number of channel uses by utilizing an efficient Karhunen-Loéve transformation (KLT) encoding scheme for signal quantization and proper node selection, under the Cramér-Rao lower bound (CRLB) constraint and the capacity limits of MAC. A novel matrix-inequality constrained successive convex approximation (MCSCA) algorithm is proposed to optimize the backhaul resource allocation, together with a greedy strategy for node selection. Finally, numerical simulations are presented to show that the proposed HISDCS design is able to outperform the baseline schemes significantly.

Simultaneous transmitting and reflecting reconfigurable intelligent surface (STAR-RIS) is able to perform reflection and refraction of the incident signals simultaneously, making it a promising technology for omnidirectional localization. In this work, we study a STAR-RIS based millimeter-wave (mmWave) localization system in the near field. In particular, we exploit the multipath components (MPCs) as signals emitted from a virtual STAR-RIS and propose a multi-stage localization algorithm. Specifically, in the initialization stage, we present a practical two-step localization method to obtain coarse estimates of UE positions, based on the second-order Fresnel approximation of the near-field channels. In the optimization stage, to effectively leverage the MPCs for localization performance improvement, we propose to add signal weights to the received signals. Then, the signal weights, STAR-RIS energy splitting (ES) coefficients and phase shifts are jointly optimized to minimize the Cramér-Rao lower bound (CRLB). Finally, in the refinement stage, the localization accuracy is further improved based on the information obtained during the first two stages. Simulation results demonstrate the effectiveness of the proposed multipath assisted localization algorithm and show that STAR-RIS surpasses conventional RIS in the omnidirectional localization scenario.

Semantic communication is a promising framework for the next generation communication systems, which generally adopts deep learning based joint source and channel coding and has been verified to offer superior efficacy. A key ingredient in augmenting the reliability of this framework is the incorporation of hybrid automatic repeat request (HARQ) techniques. However, existing semantic HARQ architectures, such as fixed-length HARQ or chase combining HARQ (CC-HARQ), utilize predefined retransmission code lengths, lacking the flexibility to adjust to different channel signal-to-noise ratio (SNR) conditions. To address this issue, this paper develops an adaptive HARQ scheme by leveraging the double deep Q-network (DDQN) to determine the retransmission code lengths. Specifically, we first propose a basic model which consists of an image reconstruction module and a performance estimation module. The performance estimation module replaces the conventional error detection method like cyclic redundancy check (CRC) to estimate the structural similarity index measure (SSIM) of the reconstructed image at the receiver. Building on this basic model, our proposed HARQ scheme works by feeding back an NACK signal and an appropriate code length determined by the proposed DDQN algorithm to the semantic transmitter for the next transmission, if the estimated SSIM performance of the previous transmission does not exceed a predefined threshold. Experimental results demonstrate that our HARQ scheme is able to achieve the same SSIM performance as the existing semantic HARQ schemes, but with significantly reduced communication cost.

For reliable communication in high mobility scenarios, we need to estimate the channel in orthogonal time frequency space (OTFS) systems, which can be considered as a sparse signal recovery problem with an uncertain sensing matrix and solved by compressed sensing (CS) algorithms. However, conventional expectation maximization (EM)-based CS algorithms only output the point estimation of grid parameters to approximate the sensing matrix, which leads to an unavoidable approximation error. To address this problem, we present a turbo inverse-free successive linear approximation variational Bayesian inference (Turbo-IFSLA-VBI) algorithm, which provides the Bayesian estimation of both channel and grid parameters, thus the approximation error can be eliminated by iteratively approximating the sensing matrix with updated grid parameters. Besides, the proposed method employs a majorization-minimization (MM) framework to simplify the matrix inverse operations, achieving a lower computational complexity. Finally, simulation results are presented to verify the superiority of the proposed scheme over the state-of-the-art schemes.

Traditional millimeter-wave beam selection or prediction algorithms typically rely on beam scanning measurements at the transceivers, incurring substantial training overhead and exhibiting limited adaptability in diverse environments. Recent efforts have aimed to mitigate these challenges by incorporating sensing information, thereby reducing or eliminating the need for extensive beam training. However, existing works predominantly concentrate on exploiting a single sensing modality and often overlook the potential benefits of utilizing historical sensing information. In this paper, we introduce an intelligent beam prediction framework that leverages a deep integration of multimodal sensing data, encompassing GPS, camera, radar, and LiDAR data. The design proposed involves the application of customized deep neural networks to extract features from camera, radar, and LiDAR data. These extracted features, combined with user position and selected beam index, are concatenated to form an aggregated feature vector at each time instance. Subsequently, a time series of these concatenated feature vectors is utilized to exploit temporal correlation for beam prediction through a dedicated long short-term memory network module. Numerical simulations confirm the effectiveness of the proposed design and its superiority over several considered state-of-the-art baselines.

Intelligent reflecting surface (IRS) is a promising technology to reconfigure the wireless channel cost-effectively, thereby improving transmission reliability in mobile ad hoc networks (MANETs). Prior works related to IRS primarily rely on channel estimation for configuring IRS, which, however, will introduce additional overhead and impact the efficiency of IRS-assisted MANETs, leading to increased delay and energy consumption during data transmission. To overcome this difficulty, we propose a multi-IRS-assisted double deep Q-Network (MIRS-DDQN) routing algorithm to find paths with higher end-to-end data rate and lower energy consumption. Routing packets are designed to collect experience tuples for DDQN to optimize the joint routing and transmit power selection policy. Moreover, these packets are also used to execute a blind beamforming strategy to configure IRS without incurring additional communication overhead. In particular, the IRSs can effectively enhance the links related to the IRS-assisted nodes and thus provide better solutions for DDQN to find an energy-efficient path with higher end-to-end data rate. Simulation results are presented to demonstrate the advantages of the proposed algorithm as compared to benchmark schemes in terms of end-to-end delay, energy consumption and end-to-end data rate.

Optimizing beamforming is crucial in mitigating pronounced propagation loss and ensuring reliable communication at millimeter-wave (mmWave) frequencies. Traditional beam optimization methods rely on either precise channel estimation or extensive beam training in the mm Wave band, both of which entail substantial pilot overhead. To alleviate this overhead, we leverage the spatial congruence between sub-6 GHz (sub-6G) and mm Wave channels and propose a sub-6G information and few pilots aided beam prediction network (SPBPNet) through deep learning. Specifically, the proposed SPBPNet comprises two cascaded modules: i) the angular information extraction module, which extracts angular features from the available sub-6G channel estimate and maps them to a minimal set of narrow beam directions to be measured in the mm Wave band; and ii) the beam prediction module, which takes limited beam training along the selected directions and then fuses measurements in the mm Wave band with the sub-6G channel information to generate mmWave beam predictions. Numerical results demonstrate that SPBPNet efficiently maps sub-6G channel estimates to mmWave beams and achieves a superior balance between performance and pilot overhead compared to state-of-the-art benchmarks. Moreover, SPBPNet exhibits resilience to varying sub-6G channel estimates at different signal-to-noise ratio levels.

In this paper, we present a novel hybrid precoding scheme, named group-and-codebook based two-timescale hybrid precoding (GC-THP), for reducing complexity in millimeter wave massive MIMO systems. The scheme clusters users into groups based on their statistical similarity and selects analog beams from a codebook with orthogonal beams. The base station time-shares among multiple analog precoders, each serving a subset of user groups, for fairness. At the slow timescale, the analog precoders, together with their corresponding group scheduling vectors and time-sharing factors, are jointly optimized, while at the fast timescale, short-term user scheduling and digital precoding are performed for a given analog precoder and its associated group scheduling vector. The proposed scheme employs beam-domain channel statistics with reduced dimension to optimize the proportional fairness utility. By applying user grouping and codebook based analog precoding, the proposed scheme significantly reduces complexity while retaining high performance levels. Moreover, the implementation complexity is tunable by adjusting the codebook size and the number of user groups, allowing for flexible performance-complexity tradeoffs. Simulation results confirm a superior performance-complexity tradeoff for the proposed GC-THP in comparison to the existing hybrid precoding schemes.

This paper considers an integrated sensing and communication system, where some radar targets also serve as communication scatterers. A location domain channel modeling method is proposed based on the position of targets and scatterers in the scattering environment, and the resulting radar and communication channels exhibit a two-dimensional (2-D) joint burst sparsity. We propose a joint scattering environment sensing and channel estimation scheme to enhance the target/scatterer localization and channel estimation performance simultaneously, where a spatially non-stationary Markov random field (MRF) model is proposed to capture the 2-D joint burst sparsity. An expectation maximization (EM) based method is designed to solve the joint estimation problem, where the E-step obtains the Bayesian estimation of the radar and communication channels and the M-step automatically learns the dynamic position grid and prior parameters in the MRF. However, the existing sparse Bayesian inference methods used in the E-step involve a high-complexity matrix inverse per iteration. Moreover, due to the complicated non-stationary MRF prior, the complexity of M-step is exponentially large. To address these difficulties, we propose an inverse-free variational Bayesian inference algorithm for the E-step and a low-complexity method based on pseudo-likelihood approximation for the M-step. In the simulations, the proposed scheme can achieve a better performance than the state-of-the-art method while reducing the computational overhead significantly.

Due to the sequential nature of the successive-cancellation (SC) algorithm, the decoding of polar codes suffers from significant decoding latencies. Fast SC decoding is able to speed up the SC decoding process, by implementing parallel decoders at the intermediate levels of the SC decoding tree for some special nodes with specific information and frozen bit patterns. To further improve the parallelism of SC decoding, this paper present a new class of special nodes composed of a sequence of rate one or single-parity-check (SR1/SPC) nodes, which can be typically found in high-rate polar codes and is able to envelop a wide variety of existing special node types. Then, we analyse the parity constraints caused by the frozen bits in each descendant node, such that the decoding performance of the SR1/SPC node can be preserved once the parity constraints are satisfied. Finally, a generalized fast decoding algorithm is proposed to decode SR1/SPC nodes efficiently, where the corresponding parity constraints are taken into consideration. Simulation results show that the proposed decoding algorithm of the SR1/SPC node can nearly achieve maximum-likelihood (ML) performance, and the overall SC decoding latency can be reduced by 43.8% as compared to the state-of-the-art fast SC decoder.

Mobile ad-hoc networks (MANETs) have garnered significant interest in various specific scenarios owing to their capability to provide flexible and decentralized communication. However, in MANETs, link failures caused by obstacles, traffic surges and inefficient routing algorithms, are commonly en-countered. To address these issues, we propose an intelligent reflecting surface assisted anti-blocking routing (IRS-ABR) algorithm that incorporates the deep Q-network (DQN) for dynamic obstacles avoidance and traffic control. Moreover, by employing IRSs as intermediate nodes in the network, the proposed algorithm can achieve enhanced path routing. The simulation results validate the effectiveness of the proposed algorithm, as it achieves a 50% higher packet delivery rate compared to the traditional algorithm, while also reducing the transmission delay and energy consumption by 34% and 12%, respectively, through the utilization of IRS.

Federated learning (FL) is an edge learning framework that has received significant attention recently. However, the cost of communication has become a major challenge for FL as the number of edge devices grows and the complexity of training models increases. Besides, data samples across all edge devices are usually not independent and identically distributed (non-IID), posing additional challenges to the convergence and model accuracy of FL. Therefore, we propose a novel personalized FL framework based on deep coded over-the-air computation, named DipFL. In this framework, we design a deep AirComp aggregation (DACA) module for n-to-1 information aggregation. Besides, a joint source-channel coding (JSCC) module is designed based on the variational auto-encoder (VAE) model, which not only encodes the transmitted data, but also reduces the bias of local samples by introducing certain regularisation terms. In addition, we propose a personalized mix module that allows local models to be more personalized by mixing the global model and the local models. Simulation results confirm that the proposed DipFL framework is able to significantly reduce the amount of transmitted data, while improving FL performance especially at low signal-to-noise regimes.

In this paper, we combine two-timescale beamforming and multi-IRS aided joint spatial division and multiplexing (JSDM) in a mmWave multiuser system. Specifically, all the users are first divided into different groups and each group is associated with an IRS. Then, we propose a novel two-stage grouping-based randomized beamforming (TS-GRB) scheme, where the analog beamformer is designed based on the statistical CSI (S-CSI) in the first stage, and the short-term digital beamformer at the BS and long-term passive beam pattern control policy at the IRSs are jointly optimized in the second stage with both S-CSI and dimension-reduced effective I-CSI. In particular, in the first stage, the analog beamformer is designed to reduce the inter-group interference (IGI) and effective channel dimension, while in the second stage, a two-timescale randomized joint beamforming (TRJB) algorithm is proposed to maximize the proportional fairness utility (PFU). We show that through two-timescale beamforming, JSDM and proper problem reformulation, the pilot overhead of our TS-GRB scheme is significantly lower than existing schemes. Finally, simulation results are presented to illustrate the effectiveness of the proposed TS-GRB scheme.

In this work, we propose a neural adjusted min-sum (NAMS) decoder for low-density parity-check (LDPC) codes. In particular, we improve the traditional normalized min-sum (NMS) decoder by introducing a selection mechanism to adjust the check-node update step, where either the min-sum rule or the belief propagation (BP) rule is selected. Besides, we unfold the modified decoder into a model-driven neural network, where layer-dependent trainable parameters are introduced as weights in the Tanner graph and optimized by gradient descent-based methods during network training. Simulation results demonstrate that the proposed NAMS decoder is able to provide superior error-correction performance as compared to the neural NMS decoder, with only slightly increased computational complexity. Moreover, in certain circumstances, the proposed NAMS decoder even outperforms the neural BP decoder, with much lower computational complexity.

Intelligent reflecting surface (IRS)-aided millimeter wave (mmWave) communication systems have gained considerable attention recently. However, the benefits brought by IRS require the instantaneous channel state information (I-CSI) of the cascaded base station (BS)-IRS and IRS-user channel which is difficult to obtain in practice, especially for the multiuser scenario. To address this issue, in this paper, we combine two-timescale beamforming and multi-IRS aided joint spatial division and multiplexing (JSDM) in a mmWave multiuser system. Specifically, all the users are first divided into different groups and each group is associated with an IRS. Then, we propose a novel two-stage grouping-based randomized beamforming (TS-GRB) scheme, where the analog beamformer is designed based on the statistical CSI (S-CSI) in the first stage, and the short-term digital beamformer at the BS and long-term passive beam pattern control policy at the IRSs are jointly optimized in the second stage with both S-CSI and dimension-reduced effective I-CSI. In particular, in the first stage, the analog beamformer is designed to reduce the inter-group interference (IGI) and effective channel dimension, while in the second stage, a two-timescale randomized joint beamforming (TRJB) algorithm is proposed to maximize the proportional fairness utility (PFU). We show that through two-timescale beamforming, JSDM and proper problem reformulation, the pilot overhead of our TS-GRB scheme is significantly lower than existing schemes. Finally, simulation results are presented to illustrate the effectiveness of the proposed TS-GRB scheme.

Orthogonal time frequency space (OTFS) modulation shows significant advantages over orthogonal frequency division multiplexing (OFDM), specially in environments with high frequency dispersion. However, high peak-to-average power ratio (PAPR) has been one of the major drawbacks of OTFS systems, which impairs the efficiency of the power amplifier. To resolve the problem, we propose a PAPR reduction method based on the autoencoder (AE) architecture through deep learning (DL) techniques, where the encoder is trained to reduce the PAPR and the decoder is trained to reconstruct the original signal. By carefully designing the loss function, the bit error rate (BER) and the PAPR are simultaneously minimized, and a hyper-parameter is introduced to achieve a good compromise between BER and PAPR in the proposed scheme. Simulation results validate the advantages of the proposed scheme as compared to the other conventional schemes.

The flying ad-hoc network (FANET) is widely applied to unmanned aerial vehicles (UAV s) but it is vulnerable to the frequency jamming in reality. Therefore, this paper proposes a federated deep Q-network (DQN) based frequency hopping strategy to solve the problem of periodic frequency jamming. We developed a DQN mechanism with an exploration-exploitation epsilon-greedy policy, directed by a federated learning mechanism to obtain a frequency hopping strategy. The simulation results show that our proposed algorithm has better convergence and decision accuracy performance compared with the DQN based frequency hopping strategy. And the performance will improve when the number of UAVs increases.

In statistical priority-based multiple access protocol (SPMA), to ensure the performance of high-priority packets, the absolute prioritization mechanism is adopted and the threshold of each priority queue is usually set to a fixed value. However this limits the transmission of low-priority packets when there are few high-priority packets. Moreover, the existing backoff algorithms do not consider the factor of channel occupancy. In order to solve the above problems, in this paper, we propose a Deep Q-Network (DQN) based transmission and backoff (DQN-TB) approach that consists of two sub networks, where the first one aims to transmit more low-priority packets but not affect the performance of high-priority packets, while the second one determines the better backoff duration by utilizing the action of former sub-network and channel occupancy. Numerical experiments demonstrate that the proposed DQN-TB approach has better throughput and backoff delay performance.

In this work, we consider the use of model-driven deep learning (DL) techniques for signal detection in massive multiple-input multiple-output (MIMO) system. Massive MIMO promises improved spectral efficiency, coverage and reliability, compared to conventional MIMO systems. Unfortunately, these benefits usually come at the cost of significantly increased computational complexity. To address this difficulty, a learned conjugate gradient descent network, referred to as LcgNet, is presented by unfolding the iterative conjugate gradient descent (CG) detector. In the proposed network, instead of calculating the exact values of the scalar step-sizes for every problem instance, we explicitly learn their universal values. We show that the performance of the proposed network can be greatly improved by augmenting the dimensions of these step-sizes. Furthermore, due to the limited learnable parameters to be optimized, the proposed networks are easy and fast to train. Numerical results demonstrate that this approach can achieve superior performance over some state-of-the-art MIMO detectors such as the CG detector, the linear minimum mean squared error (LMMSE) detector etc., with much lower computational complexity.

The wireless channel conditions in the future mobile communication systems will become more and more complex as we are developing higher frequency bands, thus it is necessary to adjust the transmission parameters frequently. To ensure the reliability of data transmission, hybrid automatic repeat request (HARQ) techniques are widely used to improve the data throughput of wireless communication systems. This paper develops a polar coded incremental redundancy HARQ (IR-HARQ) scheme based on deep reinforcement learning (DRL) to combat the unexpected channel fluctuations in practice. Specifically, the IR bits are generated by performing quasi-uniform puncturing and polarizing matrix extension on polar codes, and the number of IR bits are optimized by utilizing the deep deterministic policy gradient (DDPG) algorithm in the considered IR-HARQ scheme. Simulation results show that compared with the conventional chase combing scheme and the fixed-length IR-HARQ scheme, the proposed IR scheme can significantly improve the system throughput.

In mobile ad-hoc networks (MANETs), network performance depends strongly on routing, therefore intensive researches about routing protocol have been developed. However, most traditional routing protocols suffer from significant performance degradation as they attempt to find a shortest path without considering the impact of congestion and channel quality. In this work, we propose an intelligent routing algorithm based on prioritized replay double deep Q-network (PRD-DQN). We design two kinds of packet, fast routing (FR) packet and experience transfer (ET) packet, to explore network, and a reward function is defined in which congestion and channel quality are both considered for adaptive routing decision. The simulation results demonstrate that the proposed algorithm can effectively learn a routing strategy with low congestion and high signal to noise ratio (SNR) and outperform the Q-Learning algorithm and the optimized link state routing (OLSR) protocol in terms of convergence speed, end-to-end latency and network throughput.

Media-based modulation (MBM) is a novel method that embedding part or all of the information in the variations of the transmission media. The traditional detection algorithms for MBM system have to estimate the channel state informations (CSIs) to recover the transmitted symbols and they are affected by the number of receiving antennas (RAs). In this paper we propose a deep neural networks (DNN) detector with a novel loss function (LF) which does not require estimation of CSIs and is independent of the number of RA. Numerical results validate that the proposed DNN detector has better bit-error-ratio (BER) performance than that of the traditional detection algorithms and the performance advantage of the proposed LF is also confirmed.

Reliable data delivery is a major challenge for underwater wireless sensor networks (UWSNs) due to the harsh acoustic communication environment. In order to address this challenge, we propose a new erasure coding based multipath routing (ECMR) scheme to reduce the packet loss ratio in UWSNs. Specifically, a multipath routing algorithm with the modified link weight is designed to adapt the delay tolerant link state routing (DTLSR) algorithm to the underwater acoustic channel environment. As a reliable transmission process, each packet is split into several blocks, and erasure coding is employed to generate redundant blocks. Based on the Bayes' theorem, we propose a new mechanism to distribute different numbers of coded blocks to multiple paths with specific reliability, in order to optimize the transmission success rate. Simulation results show that the ECMR scheme can effectively reduce the packet loss ratio during data delivery in UWSNs.

Unmanned aerial vehicle (UAV) swarms (or UAV networks) are vulnerable to jamming attacks due to the shared wireless transmission medium. In order to address this difficulty, frequency hopping based anti-jamming schemes are commonly used in the literature, however, their performance is usually limited due to the unique mobility feature of UAV swarms. In this work, a practical local reaction anti-jamming (LRAJ) scheme is proposed to reduce the packet transmission delay when the jammed nodes are dynamically changing. In the proposed scheme, the jammed nodes and their one-hop neighbors determine their node types at each frequency (channel) by exchanging information about the states of their corresponding frequencies, and performing adaptive frequency hopping (AFH) accordingly. In the mean time, the unjammed nodes can still maintain their normal operations. Therefore, with the aid of the proposed LRAJ scheme, the considered UAV swarm is able to resist malicious jamming attacks in the local area. Simulation results validate the effectiveness of the proposed scheme.

Intelligent reflecting surface (IRS) has been proposed as a promising technology to smartly control the wireless signal propagation and enhance the spectral efficiency of wireless communication systems cost-effectively. The channel state information (CSI) is a crucial factor for the design of optimal passive beamforming in the IRS assisted communication systems. However, acquiring such CSI is very challenging for IRS due to its lack of radio frequency (RF) chains. In this paper, we consider an IRS aided multiple-in single-out (MISO) orthogonal frequency-division multiplexing (OFDM) system and propose a deep learning (DL) based channel estimation method to address the above challenges. In particular, a convolutional neural network is designed to estimate both the direct and cascaded channels of the system considered. Simulation results validate that the proposed DL approach achieves better performance than traditional channel estimation techniques.

The statistical priority-based multiple access protocol (SPMA) is a MAC (medium access control) protocol adopted for the Tactical Targeting Network Technology (TTNT), by virtue of high reliability and low latency. In SPMA, in order to keep the network in a favorable loading situation and ensure 99 percent first time success rate for packets of the highest priority, priority thresholds are normally set to fixed values. However, this fixed threshold strategy is not necessarily optimal when the transmission rate at each node varies due to the change of channel condition. To improve the performance, in this paper, we propose an adaptive priority-setting strategy for SPMA, accounting for the change of transmission rates at network nodes. The performance enhancement from the proposed strategy is validated through numerical simulations. In particular, it is shown that higher throughput and lower latency are achieved when the number of nodes supporting high transmission rate increases, while 99 percent first time success rate for the packets of the highest priority is guaranteed when it decreases.

Orthogonal time frequency space (OTFS) modulation is a two-dimensional modulation technique designed in the delay-Doppler domain, specially suitable for doubly-dispersive fading channels. In general, the conventional message passing (MP) algorithm is capable of eliminating the negative impacts of inter-symbol interferences for data detection in OTFS at the expense of high computational complexity. To reduce the receiver complexity in OTFS systems, we propose a damped generalized approximate message passing (GAMP) algorithm, where the damping factors are optimized based on deep learning (DL) techniques. Specifically, each iteration of the GAMP algorithm is unfolded into a layer-wise structure analogous to a neural network and the damping factors are learned to improve the detection performance. The optimized damping factors can be directly employed in the original GAMP algorithm without increasing its computational complexity. Simulation results demonstrate the effectiveness of the proposed algorithm and show that it can outperform the classical GAMP algorithm and the MP algorithm.

In this paper, we present a secure transmission scheme to improve the secrecy capacity of wiretap systems by blending the benefits of polar code and artificial noise (AN). In the considered system, a transmitter tries to communicate with a receiver without leaking any confidential information to an eavesdropper. We propose a new system utility function, referred to as message secrecy capacity (MSC), which is obtained by integrating the code rate of polar code into the conventional secrecy capacity. Then, in order to maximize the secrecy capacity of the information bits in polar code, we formulate a max-min optimization problem to optimize the powers allocated among useful signals and the AN. To address the highly non-convexity of the considered problem, we propose a concave-convex procedure (CCCP)-based algorithm by introducing some carefully designed auxiliary variables, and convergence to the set of KarushKuhn-Tucker (KKT) solutions is guaranteed. Numerical results demonstrate the effectiveness of the proposed MSC objective function and power allocation scheme.

The scheduling in a dense wireless network with interfering links is a very challenging problem, especially in the environments with additional jammers. In this work, we propose a joint jamming detection and link scheduling method based on deep neural networks (DNN). The proposed method admits a branched structure and mainly consists of two subnetworks, where the first subnetwork aims to detect and locate the jammer by utilizing the geographical information and received signal power, while the second one determines the link scheduling with the aid of the previously obtained jamming detection results. Furthermore, inspired by the multi-task learning method, we propose a hybrid-goal training approach to accelerate the training process. Numerical experiments have confirmed that the proposed DNN-based solution can achieve both superior jamming localization accuracy and highly competitive link scheduling performance.

Mobile Ad-hoc Network (MANET) is a self-configuring network that is widely used but vulnerable to the malicious jammers in practice. In this paper, we consider a jamming channel problem in MANET where a jammer intermittently interrupts the communication channels and the transmitter needs to determine which time slot to send data in order to avoid the interruption. Learning from the historical experience, a Deep Q-Network (DQN) based approach is proposed to generate transmission decisions at the transmitter. In addition, a variant of DQN, termed adaptive DQN, is introduced to cope with the change of jamming conditions. The simulation results demonstrate that the proposed scheme can learn an optimal policy to guide the transmitter to avoid jamming more quickly and efficiently than a Q-learning baseline. Moreover, the effectiveness and robustness of the adaptive DQN is also numerically verified.

Filter Bank MultiCarrier with Offset Quadrature Amplitude Modulation (FBMC-OQAM) has been intensively studied, and becomes a very potential candidate in future wireless communication system because of its numerous advantages. This paper presents a framework of Artifical Neural Network (ANN)-aided receiver design for the FBMC system. Specifically, two new joint channel estimation and equalization architectures are developed, which are based on two classical ANN algorithms, Multi-layer Perceptron (MLP) and Functinal Link Artificial Neural Network (FLANN). In addition, a powerful Loss Function (LF) is proposed by combining intrinsic characteristics of FBMC and is applied in the ANN-aided FBMC receiver. Numerical results validate the effectiveness of the proposed ANN-aided design and demonstrate its remarkable bit-error-ratio (BER) performance under multi-path channel environment. Furthermore, the performance advantage of the proposed LF is also confirmed by simulations.

The filter bank multicarrier with offset quadrature amplitude modulation (FBMC-OQAM) has attracted great interest in recent years for its low Adjacent Channel Leakage Ratio (ACLR). However, the problem of high peak-to-average power ratio (PAPR) has negative impact on the energy efficiency of the FBMC system. In this paper, we investigate PAPR reduction of FBMC-OQAM signals based on partial transmit sequence techniques (PTS). It has been shown that a trellis-based PTS scheme with multi-block joint optimization (MBJO) is superior to most schemes based on symbol-by-symbol PTS approach. However, the complexity of this scheme is undesirable in most cases. Several suboptimal solutions have been introduced in order to reduce the complexity at the expense of significant performance loss. To achieve a better complexity-performance tradeoff, we propose an improved scheme to approach the performance of trellis-based PTS with a lower complexity. The simulation results confirm that the proposed scheme effectively reduces the PAPR of FBMC-OQAM signals.

Infrastructure-less and energy-limited multi-hop ad- hoc networks are vulnerable to the Denial-of-service (DoS) attacks launched by malicious jammers. In this paper, we propose a power allocation scheme with cooperative anti-jamming strategy to enhance communication reliability against jamming in ad-hoc networks under limited resource. In the proposed anti-jamming strategy, multiple relays are employed to assist the transmission in multiple hops. Under this strategy, we propose the optimal power allocation scheme with the objective of minimizing the outage probability at the destination. We decompose the outage minimization problem into two sub-problems and obtain the near-optimal solution of each sub-problem based on the upper bound of the outage probability. Simulation results demonstrate the effectiveness of the cooperative jamming strategy and the power allocation scheme proposed.

In this paper, we investigate the carrier frequency offsets (CFOs) effect on uplink universal filtered multi-carrier-interleave division multiple access (UFMC-IDMA), which has improved robustness against inter-carrier interference (ICI) compared with orthogonal frequency division multiplexing-IDMA (OFDM-IDMA). In particular, a multiuser detection algorithm for uplink UFMC-IDMA which is robust against CFOs is proposed. The proposed algorithm takes the CFOs impact into account in the detection process and iteratively mitigates the combined interference from other users and CFOs. In addition, a corresponding approximation method with reduced complexity is also developed, which omits the interference elements whose values are small. Simulation results validate the superior interference cancellation performance of the proposed algorithm and reveal that the performance loss of the low complexity approximation method is small compare with the proposed algorithm with full calculation.

This paper investigates the resource optimization problem for the multiple-input multiple-output (MIMO) cognitive radio (CR) networks. Different from the conventional works, a novel optimization metric, namely the bandwidth-power product (BPP), is used to achieve the optimal spectrum sharing. We apply the direct-channel singular value decomposition (DC-SVD) method to design the optimal source precoding matrix. Then the unified power and channel allocation problem is derived and found to be a mixed-integer programming problem. Hence, a sub-optimal and tractable algorithm with low complexity called the joint iterative power and channel optimization (JIPCO) algorithm is proposed. Furthermore, the discrete particle swarm optimization (DPSO) algorithm is introduced to approximately provide an upper bound. Numerical results show that the JIPCO algorithm achieves the acceptable performance with great reduction of computation complexity and validate the superiority of the proposed scheme compared to conventional power optimization scheme using the water-filling method.

Cooperative communications and full-duplex (FD) relaying have been proposed to meet the ever increasing data traffic demands and fully exploit the scarce spectrum resources. In this paper, we consider FD relaying networks with multiple source-destination (S-D) pairs and multiple FD relays. We propose the joint optimal relay assignment and power allocation (ORAPA) scheme to maximize the sum rate of the network. The sum rate maximization problem is formulated as a mixed-integer nonlinear programming (MENLP) problem. We provide an equivalent maximum weighted bipartite matching (MWBM) problem to solve the MINLP issue and reduce computational complexity by adopting the Hungarian algorithm. Simulations results are provided to verify the effectiveness of our proposed scheme.

Multiple-input-multiple-output (MIMO) Device-to-Device (D2D) communication underlaying cellular networks can improve user throughout and extend battery life of user equipment. This paper studies MIMO precoding and power allocation schemes for the D2D and cellular uplink communications to improve the energy efficiency of both the D2D and the cellular user (CU). Due to the co-channel interference caused by the shared resources, %between the D2D link and the cellular uplink, the precoder design and power allocation of both the D2D transmitter (DTX) and the CU become highly inter- dependent. A distributed cooperative iterative optimization algorithm (DCIOA) is introduced, where either the DTX or the CU cooperatively exchanges the interference information and iteratively updates its optimum precoder and power level. Meanwhile, a centralized stochastic search optimization algorithm using the particle swarm optimization (PSO) method is applied to provide an upper bound. Simulation results show that the performance of the DCIOA is close to that of the centralized algorithm with significant reduction of computation and slight increment of exchange overhead.

This paper aims to investigate the resource allocation problem in a relay-assisted OFDMA cognitive radio (CR) system. Different from conventional CR resource allocation problems, a joint bandwidth and power optimization framework using the bandwidth-power product metric is proposed. Besides, rate requirement of the secondary system is satisfied and interference power at the primary receiver is limited. Meanwhile channel pairing at the relay terminal is operated so that signal received over a particular channel at the first hop can be forwarded over a different channel at the second hop. The problem is found to be nonconvex and intractable to seek for an optimal solution. Hence, using the Lagrangian-Dual and Gauss-Newton methods, a sub-optimal algorithm with low complexity is proposed. Numerical results are provided to validate the superiority of the proposed method compared to conventional power optimization methods using the waterfilling scheme.

This study investigates resource allocation problems for a point‐to‐point multi‐carrier multiple‐input multiple‐output cognitive radio network. Different from conventional resource optimisation problems, a joint power and bandwidth resource optimisation framework using the novel optimisation metric, namely bandwidth‐power product, is developed. Besides, rate requirement of the secondary system is satisfied, and the interferences introduced to the primary users (PUs) are below threshold of tolerance. The optimal source precoding matrix is designed and two methods, namely the project‐channel singular value decomposition (SVD) and direct‐channel SVD methods, are applied to satisfy interference power constraints for PUs. Then the unified power and channel allocation problem is derived and found to be a mixed‐integer programming problem. Hence, a sub‐optimal and tractable algorithm with low complexity is proposed. The innovative idea is to determine the channel resource budget by selecting the best channels, where the criterion for evaluating the quality of channel is detailed discussed. Then the power optimisation subproblem and channel allocation subproblem can be performed independently using the Lagrange‐duality theory and Gauss–Newton method, respectively. The simulation results show significant improvement in spectral efficiency by using this framework compared to classical power optimisation framework using the waterfilling scheme.

A novel frequency offset tracking algorithm for Space-Time Block Coded (STBC) Orthogonal Frequency Division Multiplexing (OFDM) systems is proposed in this work. Tracking of a frequency offset between the transmitter and the receiver is often aided by transmitting pilots embedded in the data payload. The proposed algorithm mainly exploits the specific construction of the OFDM symbol in STBC-OFDM systems, which does not need any additional pilots or sequences in the data field, providing high efficiency in spectrum. The estimator is derived on the basis of the maximum likelihood (ML) theory. Simulation results show that in a 2 × 2 multiple input multiple output (MIMO) system, under the assumption that the antennas are uncorrelated to each other, this method can provide a significant performance improvement in terms of the estimation accuracy of the frequency offset.