Fluid antenna systems (FAS) signify a pivotal advancement in 6G communication by enhancing spectral efficiency and robustness. However, obtaining accurate channel state information (CSI) in FAS poses challenges due to its complex physical structure. Traditional methods, such as pilot-based interpolation and compressive sensing, are not only computationally intensive but also lack adaptability. Current extrapolation techniques relying on rigid parametric models do not accommodate the dynamic environment of FAS, while data-driven deep learning approaches demand extensive training and are vulnerable to noise and hardware imperfections. To address these challenges, this paper introduces a novel self-supervised learning network (SSNet) designed for efficient and adaptive channel extrapolation in FAS. We formulate the problem of channel extrapolation in FAS as an image reconstruction task. Here, a limited number of unmasked pixels (representing the known CSI of the selected ports) are used to extrapolate the masked pixels (the CSI of unselected ports). SSNet capitalizes on the intrinsic structure of FAS channels, learning generalized representations from raw CSI data, thus reducing dependency on large labeled datasets. For enhanced feature extraction and noise resilience, we propose a mix-of-expert (MoE) module. In this setup, multiple feedforward neural networks (FFNs) operate in parallel. The outputs of the MoE module are combined using a weighted sum, determined by a gating function that computes the weights of each FFN using a softmax function. Extensive simulations validate the superiority of the proposed model. Results indicate that SSNet significantly outperforms benchmark models, such as AGMAE and long short-term memory (LSTM) networks by using a much smaller labeled dataset. A key observation is that the proposed model is more effectively trained using a small unmasked ratio of known CSI. Specifically, the proposed SSNet trained using CSI of 10 % total ports outperforms that trained using CSI of 25 % and 50 % total ports. This is because using a smaller number of known CSIs during training, the proposed model is forced to learn more effective channel correlation for channel extrapolation at the expense of higher training complexities. Ablation experiments reveal substantial performance gains from the MoE module’s integration. Furthermore, zero-shot learning experiments show a moderate performance degradation of about 3-5 dB, underscoring the model’s robust generalization ability. Finally, the inference speed experiments illustrate that the proposed model outperforms the benchmark models dramatically at the expense of a slightly longer execution time of 1.13 ms, 2.9 ms, and 3.12 ms on NVIDIA RTX 4090, 4060, and 3060 graphics processing units (GPU)s, respectively.

In 6G mobile communications, acquiring accurate and timely channel state information (CSI) becomes increasingly challenging due to the growing antenna array size and bandwidth. To alleviate the CSI feedback burden, the channel knowledge map (CKM) has emerged as a promising approach by leveraging environment-aware techniques to predict CSI based solely on user locations. However, how to effectively construct a CKM remains an open issue. In this paper, we propose F4-CKM, a novel CKM construction framework characterized by four distinctive features: radiance Field rendering, spatial-Frequency-awareness, location-Free usage, and Fast learning. Central to our design is the adaptation of radiance field rendering techniques from computer vision to the radio frequency (RF) domain, enabled by a novel Wireless Radiator Representation (WiRARE) network that captures the spatial-frequency characteristics of wireless channels. Additionally, a novel shaping filter module and an angular sampling strategy are introduced to facilitate CKM construction. Extensive experiments demonstrate that F4-CKM significantly outperforms existing baselines in terms of wireless channel prediction accuracy and efficiency.

In this paper, by integrating the advantages of synchronous federated learning (SFL) and asynchronous federated learning (AFL), an efficient federated learning (FL) framework with hybrid aggregation is proposed for the connected and autonomous vehicles (CAVs) over multiple base stations (BSs). Specifically, to cope with the stragglers caused by traffic accidents, extreme weather or other uncontrollable factors, the AFL with periodic aggregation is proposed to perform edge model aggregation within a single BS. Furtherly, taking the freshness of local model updates and the training data distribution into account, a novel weighting strategy is designed correspondingly. Moreover, to reconcile the contradiction between the scarce wireless communication resources and the enormous communication overhead caused by frequent exchanges of model parameters, an effective model compression mechanism is constructed based on the inherent statistical property of FL. In addition, considering a relatively small number of autonomous vehicles (AVs) within the limited coverage of a single BS and the instant guidance required for the CAVs, the SFL based on FedAvg is introduced to aggregate edge models trained from multiple BSs at network edge instead of remote cloud. The superior performance of the proposed method is verified by various simulations on real dataset.

On unlicensed spectrum, the beam scanning in directional listen before talk (LBT) channel access is similar to the beam training procedure adopted for channel estimation before data transmission. Therefore, repeating these operations not only increases signalling overhead but also wastes limited resources. To address this issue, this paper proposes an efficient data transmission strategy that combines the two similar aforementioned steps. Considering the limited capacity of each beam, the optimal beam pairing problem between the transmitter and the receiver is formulated as a matching game problem and solved by means of the Gale-Shapley algorithm. Finally, the results validate the effectiveness of the proposed mechanism in reducing the complexity of beam search, minimizing the signalling overhead, and optimizing the system capacity.

In this work, we propose a multi-aircraft cooperative handover scheme for satellite-to-aircraft communication systems. Specifically, considering the dual characteristics of aircraft resource demands and three satellite states (normal, congested, and failed), multiple aircraft collaborate to make handover decisions while maintaining network stability and avoiding congestion. We formulate the cooperative handover problem as a multi-objective optimization problem to minimize communication latency and network congestion while maximizing connection stability. To solve this problem, we first model the handover scheme into the Markov decision process to facilitate seamless satellite-aircraft handover. Then we develop a multi-agent deep deterministic policy gradient (MADDPG) algorithm with centralized training and decentralized execution architecture. Due to the time-varying nature of the action space and the constraint that action selection is limited to currently visible and undamaged satellites, we implement an action mask approach to effectively filter out illegal actions instead of using conventional negative reward methods. The simulation results demonstrate that the proposed framework effectively reduces handover frequency, minimizes communication latency, and achieves better network load balancing, validating its feasibility and effectiveness in satellite-toaircraft communication systems.

In decentralized learning systems over wireless device-to-device (D2D) networks, training latency is a key metric that needs to be minimized by link selection and resource allocation, thereby accelerating model training. However, it may cause large computational complexity in general. To tackle the challenge, this paper proposes a graph neural network (GNN)-based algorithm to minimize the training latency. Under modeling the D2D network as a graph, the link selection and resource allocation can be efficiently obtained based on the local computing power and link quality. By the constraint on the network connectivity, the training latency can be significantly reduced while guaranteeing accuracy with a low complexity. The simulation results demonstrate that the GNN-based approach outperforms traditional approaches, offering superior scalability and robustness in heterogeneous large-scale D2D networks.

Recently, decentralized learning has been proposed for model training among mobile devices without center nodes. However, large resource overhead for model aggregation and synchronization would be incurred, which may reduce the learning performance under a given resource budget. To cope with these issues, we propose a novel cluster-based collaborative learning framework over device-to-device (D2D) network, where one device is selected as the cluster head for model aggregation. Within the proposed framework, the learning performance (evaluated by model divergence) and learning latency are analyzed with the consideration of imbalanced data and unreliable D2D communication. Then, an optimization problem is formulated to maximize the learning performance under a given latency constraint by joint cluster head selection and resource allocation. To solve this problem, a lower bound on latency constraint is first obtained for error-free model aggregation. The optimal learning performance is also derived with different degrees of data distribution. After that, an adaptive cluster head selection and resource allocation algorithm is developed for erroneous case by introducing the outage probability. Finally, comprehensive experiments are conducted on well-known models and datasets to illustrate the effectiveness of the proposed algorithm. The results show that our proposal can improve the learning performance while reducing communication and signaling overheads.

We consider the problem of reducing the learning latency of layerwise federated learning through joint device selection and bandwidth allocation. Specifically, we examine practical scenarios with heterogeneous devices with varying system parameters (e.g., CPU frequency, transmit power, etc.) and energy budgets. We formulate a long-term optimization problem, which is difficult to solve even with perfect channel state information. To address the issue, we employ Lyapunov theory to transform the problem into a series of online optimization problems, each of which can be efficiently solved using an alternating optimization-based method. Simulation results show that our scheduling scheme surpasses baseline schemes not only in terms of reducing the learning latency but also in reducing the energy deficit.

Deep mutual learning (DML) is one of the emerging technologies for mobile intelligent applications that has attracted much attention in recent years. To effectively deploy DML at a large scale, in this paper, we propose a novel hierarchical asynchronous deep mutual learning (HADML) system that enables devices to collaborate in model training without the exchange of local datasets. To further improve the learning efficiency, the average energy cost for model exchanging is minimized by jointly optimizing the link scheduling and communication resource allocation. To efficiently solve this problem, the graph neural network and deep unfolding network are employed to obtain the link scheduling and resource allocation, respectively. Finally, the simulation results demonstrate that our proposed algorithm can achieve a balance between knowledge sharing and communication energy consumption.

Unmanned Aerial Vehicle (UAV) technologies have garnered significant attention, particularly in the context of UAV-assisted wireless networks, which are seen as a pivotal component in the development of Sixth-Generation (6G) mobile communication systems. In this research, we delve into the realm of UAV-assisted wireless communication networks, where a single UAV efficiently caters to numerous random mobile users on the ground. Our focus lies in optimizing user movement tracking, beamforming, and UAV trajectory to maximize the data transmission rates for users within a specified time frame, all while adhering to stringent power constraints and the UAV's limited flight range. We harness the power of deep reinforcement learning (DRL) to monitor mobile users and predict the ever-changing channel state information. As beamforming and UAV trajectory adjustments operate on different timescales, we introduce a dual-layer deep unfolding network to fine-tune the transmit beamformer and UAV trajectory simultaneously. The outcomes of our simulations demonstrate the effectiveness and commendable performance of the proposed scheme.

To satisfy the requirements of many industrial applications, realizing ultrareliable low-latency communication (URLLC) has become one of the major challenges for future wireless networks. This article considers a downlink multiuser multiple-input-single-output (MISO) system in the Internet of Things (IoT) networks, in which a multiantenna base station (BS) serves multiple delay-sensitive IoT users, each equipped with a single antenna. To minimize the overall end-to-end delay, we jointly optimize the beamforming vectors and the packet blocklength to balance the queuing delay and the transmission delay. The problem is formulated as a Markov decision process (MDP), whose optimal solution can be theoretically found. However, the complexity on finding the optimal resource allocation and blocklength selection strategy is prohibitively high for real-system deployments due to the large state and action space. To overcome this issue, we simplify the original problem and develop an iterative algorithm to solve the simplified problem based on the uplink-downlink duality theory. Since solving the simplified problem would result in suboptimal solutions and may degrade the latency performance, we further develop a deep-reinforcement-learning (DRL)-based beamforming and blocklength selection framework to efficiently learn the optimal strategy of the original MDP. Simulation results demonstrate that the proposed algorithms can effectively improve the latency performance compared with the benchmark algorithm.

Recently, deep learning (DL) based semantic communication systems have shown great potential to improve transmission efficiency in various tasks. However, the coexisting mechanism between semantic communications and other services remains unexplored, which limits the application of semantic communications in practical communication systems. In this paper, we propose a dynamic multiplexing and co-scheduling scheme for the semantic and ultra-reliable low-latency communication (URLLC) traffic coexisting systems. In particular, a joint resource allocation and model training problem is formulated, which aims at maximizing the utility of semantic service while satisfying the latency requirement of URLLC traffic. To reduce the computational complexity, the original problem is simplified and decoupled into a joint resource allocation and model selection problem and a robust model training problem. In the resource allocation and model selection phase, the original problem is decomposed into three subproblems and an alternating optimization algorithm is then proposed to obtain the optimal resource allocation result. In the model training phase, a two-stage semantic communication network is designed, which can efficiently mitigate the impact of feature erasure brought by the random arrival of URLLC traffic. Simulation results show that the proposed method can effectively improve the quality of semantic service while satisfying the latency requirement of URLLC traffic.

We study efficient federated learning (FL) using random pruning in resource-constrained edge intelligence networks. We propose an edge device selection strategy to identify appropriate edge devices for participating in FL at the beginning of each training iteration. We then formulate an optimization problem that jointly optimizes the pruning ratio, CPU frequency, uplink power, and bandwidth allocation for the selected edge devices. Since the optimization problem is non-convex and challenging to solve directly, we decompose it into three subproblems and propose efficient algorithms or closed-form solutions for each subproblem. Based on the solutions to the subproblems, an alternating optimization algorithm is constructed to solve the original problem. Simulation results demonstrate that our scheme outperforms baseline schemes in terms of both learning accuracy and energy consumption.

Device-to-device (D2D)-assisted decentralized learning has been proposed for mobile devices to collaboratively train artificial intelligence networks without the centralized parameter server. However, a densely connected network will cause large learning latency and energy consumption due to the limited computation and communication resources. In addition, link selection and aggregation weight have a significant impact on the learning performance. To cope with these challenges, we propose a joint computing power adjustment, wireless resource allocation, link selection, and aggregation weight adaptation mechanism to improve both communication and energy efficiencies. Specifically, the learning performances including the convergence rate, per-iteration learning latency, and per-iteration energy consumption are first analyzed. Then, an optimization problem is formulated to minimize the total learning cost, which is defined as the weighted sum of total learning latency and energy consumption. Given a network topology, the computing power and wireless resource allocation are optimized by the alternating optimization algorithm. Moreover, the optimal aggregation weight is obtained by semidefinite programming. With respect to link selection, we propose a tabu search based meta-heuristic algorithm to approximately achieve feasible solutions with a low computational complexity. Finally, extensive experiments demonstrate that the proposed link selection algorithm can significantly reduce the learning cost under the given learning accuracy requirement.

Wireless hierarchical federated learning (HFL) has been proposed for large-scale model training over multi-cell network while preserving the data privacy. However, the imbalanced data distribution and load have a significant impact on the convergence rate, the learning accuracy, and the learning latency in wireless HFL with non-independent identically distributed training data. To cope with these challenges, we first derive the learning latency and the upper bound of the model error. Then, an optimization problem is formulated to minimize the weighted sum of total data distribution distance and learning latency. Joint user association and wireless resource allocation algorithms are investigated to achieve the optimal learning performance. Finally, the effectiveness of the proposed algorithms are demonstrated by the simulations.

In recent years, wirelessfederated learning(FL) has been proposed to support the mobile intelligent applications over the wireless network, which protects the data privacy and security by exchanging the parameter between mobile devices and thebase station(BS). However, the learning latency increases with the neural network scale due to the limited local computing power and communication bandwidth. To tackle this issue, we introduce model pruning for wireless FL to reduce the neural network scale. Device selection is also considered to further improve the learning performance. By removing the stragglers with low computing power or bad channel condition, the model aggregation loss caused by model pruning can be alleviated and the communication overhead can be effectively reduced. We analyze the convergence rate and learning latency of the proposed model pruning method and formulate an optimization problem to maximize the convergence rate under the given learning latency budget via jointly optimizing the pruning ratio, device selection, and wireless resource allocation. By solving the problem, the closed-form solutions of pruning ratio and wireless resource allocation are derived and the threshold-based device selection strategy is developed. Finally, extensive experiments are carried out to demonstrate that the proposed model pruning algorithm outperforms other existing schemes.

In this work, hierarchical federated learning (HFL) over wireless multi-cell networks is proposed for large-scale model training while preserving data privacy. However, the imbalanced data distribution has a significant impact on the convergence rate and learning accuracy. In addition, a large learning latency is incurred due to the traffic load imbalance among base stations (BSs) and limited wireless resources. To cope with these challenges, we first provide an analysis of the model error and learning latency in wireless HFL. Then, joint user association and wireless resource allocation algorithms are investigated under independent identically distributed (IID) and non-IID training data, respectively. For the IID case, a learning latency aware strategy is designed to minimize the learning latency by optimizing user association and wireless resource allocation, where a mobile device selects the BS with the maximal uplink channel signal-to-noise ratio (SNR). For the non-IID case, the total data distribution distance and learning latency are jointly minimized to achieve the optimal user association and resource allocation. The results show that both data distribution and uplink channel SNR should be taken into consideration for user association in the non-IID case. Finally, the effectiveness of the proposed algorithms are demonstrated by the simulations.

In wireless federated learning, the unreliable communication has a significant impact on the convergence rate and learning latency, which cannot be ignored. To cope with this problem, we propose a novel modulation selection mechanism to achieve the balance between learning latency and convergence rate loss caused by stochastic channel error. Different from the traditional one, the modulation mode in wireless FL system should be adjusted according to devices’ computing power, channel conditions, and training data importance. Then, an optimization problem to maximize the learning efficiency is formulated to obtain the optimal modulation scheme. Finally, extensive experiments are implemented to demonstrate the effectiveness of the proposed mechanism.

Unlike the conventional device-to-device (D2D) networks, the unlicensed D2D (D2D-U) pairs can not only reuse the licensed channels with the base station (BS) but also share the unlicensed channels with the WiFi stations. One challenge arises from the fact that the co-channel interference on licensed channels and the collision probability on unlicensed channels may cause extra power consumption at the terminals. Accordingly, we first propose a channel access method for the D2D-U pairs on unlicensed channels. Then, a decentralized joint spectrum and power allocation scheme is designed to minimize the power consumption at D2D-U pairs. Different from the existing distributed schemes, the proposed scheme can guarantee the global minimization of power consumption across the D2D-U pairs. Simulation results validate the theoretical analysis and verify the performance from the proposed scheme.

In wireless federated learning system, wireless communication and local computation have a significant impact on the learning latency due to the limited bandwidth and computing power of mobile devices. To reduce the learning latency, local stochastic gradient methods and gradient compression can be applied, which however would decrease the convergence rate. To tackle such issues, in this paper, the trade-off between the convergence rate and the learning latency is taken into account. We first formulate an optimization problem to maximize the convergence rate under the given training latency constraint via jointly optimizing the batchsize, compression ratio, and spectrum allocation. Then, by decomposing the problem into two subproblems, an adaptive algorithm is proposed to obtain the optimal solution. The results show that batchsize and compression ratio should be selected according to the computing power and channel state information of the devices to improve the convergence rate. Finally, experimental results are presented to verify the effectiveness of the proposed algorithm.

In licensed assisted access (LAA) based small cell networks (SCNs), the small base station (SBS) can reuse the uplink licensed bands with the macro cell while sharing the unlicensed bands with the Wi-Fi networks to improve its throughput. To mitigate the severe co-channel interference to the macro cell and guarantee the harmonious coexistence with the Wi-Fi networks, the spectrum and power should be jointly allocated at the SBSs. Moreover, to overcome the overwhelming signaling overheads introduced by the traditional centralized scheme and adapt to the variable radio environments, an adaptive decentralized scheme is necessary. Therefore, in this paper, an adaptive semi-distributed scheme is proposed to jointly allocate the power and spectrum on both licensed and unlicensed bands, which can achieve the global optimal spectrum efficiency (SE) of the SCNs. The proposed scheme can enable the SBSs to work independently and adaptively without sharing the whole information of the SBSs, but requires some Lagrangian parameters exchange via the coordination of the macro base station (MBS). Theoretical analysis and numerical results are presented to show that the proposed scheme is capable of achieving the optimal SE on both licensed and unlicensed bands adaptively while confining the co-channel interference to the MBS and guaranteeing the fair coexistence with the Wi-Fi network.

Facing the challenges brought by the surge in the demand for mobile data traffic and increasingly scarce spectrum resources, two well-known channel access mechanisms named as duty-cycle muting (DCM) and listen- before-talk (LBT) have been proposed. In this article, we propose a novel adaptive hybrid channel access scheme which takes advantages of both mechanisms. Based on the WiFi traffic and the available licensed spectrum resource, our proposal can adaptively adjust the important parameters, such as the back-off window size and the duty-cycle time fraction, while ensuring fair and harmonious network coexistence between the WiFi and LTE-U systems. It can realize the flexible handoff between the DCM and LBT mechanisms to meet the requirements of different markets as well. Moreover, joint transmission power and spectrum resource allocation is also studied to improve the spectral efficiency on both licensed and unlicensed bands. The effectiveness of the proposed scheme is finally validated by numerical simulations.

With the development of the fifth generation mobile communication, long-term evolution in unlicensed spectrum (LTE-U) has been proposed as a promising means to solve the spectrum scarcity problem. In this paper, we investigate the mobile data offloading in a LTE-U system where one LTE small cell base station (SBS) coexists with several WiFi APs. Different from the traditional unidirectional mobile offloading, the proposed bidirectional mobile offloading can improve the system throughput and achieve the load balance among different WiFi APs as well. We first formulate a multi-objective optimization problem (MOOP) to analyze the bidirectional mobile offloading. Then, we propose two different algorithms to achieve the effective solutions to the MOOP. The first one utilizes the Nash bargaining solution (NBS) to obtain the closed-form solution to the optimal unlicensed resource allocation. The second one leverages the Non-dominated Sorting Genetic Algorithm II (NSGA-II) to develop a low-complexity heuristic algorithm. Our proposals are finally validated by numerical simulations.