Federated Learning (FL) enables distributed model training across edge devices in a privacy-friendly manner. However, its efficiency heavily depends on effective device selection and high-dimensional resource allocation in dynamic and heterogeneous wireless environments. Conventional methods demand a confluence of domain-specific expertise, extensive hyperparameter tuning, and/or heavy interaction cost. This paper proposes a Tool-aided Evolutionary Large Language Model (T-ELLM) framework to generate a qualified policy for device selection in a wireless FL environment. Unlike conventional optimization methods, T-ELLM leverages natural language-based scenario prompts to enhance generalization across varying network conditions. The framework decouples the joint optimization problem mathematically, enabling tractable learning of device selection policies while delegating resource allocation to convex optimization tools. To facilitate the evolutionary process, T-ELLM interacts with a sample-efficient, model-based virtual learning environment that captures the relationship between device selection and learning performance. This developed virtual environment reduces reliance on real-world interactions, thus minimizing communication overhead while refining the LLM-based decision-making policy through group relative policy optimization. Theoretical analysis proves that the discrepancy between virtual and real environments is bounded, ensuring the advantage function learned in the virtual environment maintains a provably small deviation from real-world conditions. Experimental results demonstrate that T-ELLM outperforms benchmark methods in energy efficiency and exhibits robust adaptability to environmental changes.

Federated learning (FL) in multi-service provider (SP) ecosystems is fundamentally hampered by non-cooperative dynamics, where privacy constraints and competing interests preclude the centralized optimization of multi-SP communication and computation resources. In this paper, we introducePAC-MCoFL, a game-theoretic multi-agent reinforcement learning (MARL) framework where SPs act as agents to jointly optimize client assignment, adaptive quantization, and resource allocation. Within the framework, we integrate Pareto Actor-Critic (PAC) principles with expectile regression, enabling agents to conjecture optimal joint policies to achieve Pareto-optimal equilibria while modeling heterogeneous risk profiles. To manage the high-dimensional action space, we devise a ternary Cartesian decomposition (TCAD) mechanism that facilitates fine-grained control. Further, we developPAC-MCoFL-p, a scalable variant featuring a parameterized conjecture generator that substantially reduces computational complexity with a provably bounded error. Alongside theoretical convergence guarantees, our framework's superiority is validated through extensive simulations –PAC-MCoFLachieves approximately$5.8\%$and$4.2\%$improvements in total reward and hypervolume indicator (HVI), respectively, over the latest MARL solutions. The results also demonstrate that our method can more effectively balance individual SP and system performance in scaled deployments and under diverse data heterogeneity.

With the prevalence of emerging artificial intelligence services in next-generation wireless edge networks, Split Federated Learning (SFL), which divides a learning model into server-side and client-side models, has emerged as an appealing technology to deal with the heavy computational burden for network edge clients. However, existing SFL frameworks would frequently upload smashed data and download gradients between the server and each client, leading to severe communication overheads. To address this issue, this work proposes a novel communication-and-computation efficient SFL framework, which allows dynamic model splitting (server- and client-side model cutting point selection) and broadcasting of aggregated smashed data gradients. We theoretically analyze the impact of the cutting point selection on the convergence rate, revealing that model splitting with a smaller client-side model size leads to a better convergence performance and vise versa. Based on the above insights, we formulate an optimization problem to minimize the model convergence rate and latency under the consideration of data privacy via a joint Cutting point selection, Communication and Computation resource allocation (CCC) strategy. To deal with the proposed mixed integer nonlinear programming optimization problem, we develop an algorithm by integrating the Double Deep Q-learning Network (DDQN) with convex optimization methods. Extensive experiments validate our theoretical analyses across various datasets, and the numerical results demonstrate the effectiveness and superiority of the proposed communication-efficient SFL compared with existing schemes, including parallel split learning and traditional SFL mechanisms.

This paper presents STS-T, a novel spatial-temporal-spectral transformer for end-to-end prediction of jamming spectrum situations. To address challenges such as node mobility, legitimate interference, and noisy or missing data, STS-T employs spatiotemporal positional encoding and axial attention mechanisms that effectively capture global features while reducing computational complexity. Specifically, the model integrates a Temporal Predictive Decoder with dilated convolution for improved temporal forecasting, and a Spatial Interpolation Decoder with cross-attention to seamlessly map between input and output spaces. A synthetic dataset generated via OMNeT++ simulations is also introduced to compensate for the lack of real-world UAV jamming data. Experimental results confirm that STS-T outperforms existing baselines, providing a robust solution for predicting jamming spectrum situations in complex UAV swarm environments.

Integrated communication and control serves as a critical ingredient in Multi-Agent Reinforcement Learning. However, partial observability limitations will impair collaboration effectiveness, and a potential solution is to establish consensus through well-calibrated latent variables obtained from neighboring agents. Nevertheless, the rigid transmission of less informative content can still result in redundant information exchanges. Therefore, we propose a Consensus-Driven Event-Based Graph Information Bottleneck (CDE-GIB) method, which integrates the communication graph and information flow through a GIB regularizer to extract more concise message representations while avoiding the high computational complexity of innerloop operations. To further minimize the communication volume required for establishing consensus during interactions, we also develop a variable-threshold event-triggering mechanism. By simultaneously considering historical data and current observations, this mechanism capably evaluates the importance of information to determine whether an event should be triggered. Experimental results demonstrate that our proposed method outperforms existing state-of-the-art methods in terms of both efficiency and adaptability.

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.

Multiagent reinforcement learning (MARL) has emerged as a foundational approach for addressing diverse, intelligent control tasks in various scenarios like the Internet of Vehicles, Internet of Things, and unmanned aerial vehicles. However, the widely assumed existence of a central node for centralized, federated learning-assisted MARL might be impractical in highly dynamic environments. This can lead to excessive communication overhead, potentially overwhelming the system. To address these challenges, we design a novel communication-efficient, fully distributed algorithm for collaborative MARL under the frameworks of soft actor-critic (SAC) and decentralized federated learning (DFL), named regulated segment mixture-based multiagent SAC (RSM-MASAC). In particular, RSM-MASAC enhances multiagent collaboration and prioritizes higher communication efficiency in dynamic systems by incorporating the concept of segmented aggregation in DFL and augmenting multiple model replicas from received neighboring policy segments, which are subsequently employed as reconstructed referential policies for mixing. Distinctively diverging from traditional reinforcement learning (RL) approaches, RSM-MASAC introduces new bounds under the framework of maximum entropy reinforcement learning (MERL). Correspondingly, it adopts a theory-guided mixture metric to regulate the selection of contributive referential policies, thus guaranteeing soft policy improvement during the communication-assisted mixing phase. Finally, the extensive simulations in mixed-autonomy traffic control scenarios verify the effectiveness and superiority of our algorithm.

Optimizing the deployment of large language models (LLMs) in edge computing environments is critical for enhancing privacy and computational efficiency. In the path toward efficient wireless LLM inference in edge computing, this study comprehensively analyzes the impact of different splitting points in mainstream open-source LLMs. Accordingly, this study introduces a framework taking inspiration from model-based reinforcement learning to determine the optimal splitting point across the edge and user equipment. By incorporating a reward surrogate model, our approach significantly reduces the computational cost of frequent performance evaluations. Extensive simulations demonstrate that this method effectively balances inference performance and computational load under varying network conditions, providing a robust solution for LLM deployment in decentralized settings.

Abstract The sixth-generation (6G) mobile network implements the social vision of digital twins and ubiquitous intelligence. Contrary to the fifth-generation (5G) mobile network that focuses only on communications, 6G mobile networks must natively support new capabilities such as sensing, computing, artificial intelligence (AI), big data, and security while facilitating Everything as a Service. Although 5G mobile network deployment has demonstrated that network automation and intelligence can simplify network operation and maintenance (O&M), the addition of external functionalities has resulted in low service efficiency and high operational costs. In this study, a technology framework for a 6G autonomous radio access network (RAN) is proposed to achieve a high-level network autonomy that embraces the design of native cloud, native AI, and network digital twin (NDT). First, a service-based architecture is proposed to re-architect the protocol stack of RAN, which flexibly orchestrates the services and functions on demand as well as customizes them into cloud-native services. Second, a native AI framework is structured to provide AI support for the diverse use cases of network O&M by orchestrating communications, AI models, data, and computing power demanded by AI use cases. Third, a digital twin network is developed as a virtual environment for the training, pre-validation, and tuning of AI algorithms and neural networks, avoiding possible unexpected losses of the network O&M caused by AI applications. The combination of native AI and NDT can facilitate network autonomy by building closed-loop management and optimization for RAN.

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.

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.

Vehicle-to-everything communications-assisted autonomous driving has witnessed remarkable advancements in recent years, with pragmatic communications (PragComm) emerging as a promising paradigm for real-time collaboration among vehicles and other agents. Simultaneously, extensive research has explored the interplay between collaborative perception and decision-making in end-to-end driving frameworks. In this work, we revisit the collaborative driving problem and propose the Select2Drive framework to optimize the utilization of limited computational and communication resources. Particularly, to mitigate cumulative latency in perception and decision-making, Select2Drive introduces distributed predictive perception by formulating an active prediction paradigm and simplifying high-dimensional semantic feature prediction into a computationally efficient, motion-aware reconstruction. Given the “less is more” principle that an over-broadened perceptual horizon possibly confuses the decision module rather than contributing to it, Select2Drive utilizes area-of-importance-based PragComm to prioritize the communication of critical regions, thus boosting both communication efficiency and decision-making efficacy. Empirical evaluations on the V2Xverse and real-world DAIR-V2X datasets demonstrate that Select2Drive achieves a 2.60% and 1.99% improvement in offline perception tasks under limited bandwidth (resp., pose error conditions). Moreover, it delivers at most 8.35% and 2.65% enhancement in closed-loop driving scores and route completion rates, particularly in scenarios characterized by dense traffic and high-speed dynamics.

Generally, Reinforcement Learning (RL) agent updates its policy by repetitively interacting with the environment, contingent on the received rewards to observed states and undertaken actions. However, the environmental disturbance, commonly leading to noisy observations (e.g., rewards and states), could significantly shape the performance of agent. Furthermore, the learning performance of Multi-Agent Reinforcement Learning (MARL) is more susceptible to noise due to the interference among intelligent agents. Therefore, it becomes imperative to revolutionize the design of MARL, so as to capably ameliorate the annoying impact of noisy rewards. In this paper, we propose a novel decomposition-based multi-agent distributional RL method by approximating the globally shared noisy reward by a Gaussian Mixture Model (GMM) and decomposing it into the combination of individual distributional local rewards, with which each agent can be updated locally through distributional RL. Moreover, a Diffusion Model (DM) is leveraged for reward generation in order to mitigate the issue of costly interaction expenditure for learning distributions. Furthermore, the monotonicity of the reward distribution decomposition is theoretically validated under nonnegative weights and increasing distortion risk function, while the design of the loss function is carefully calibrated to avoid decomposition ambiguity. We also verify the effectiveness of the proposed method through extensive simulation experiments with noisy rewards. Besides, different risk-sensitive policies are evaluated in order to demonstrate the superiority of distributional RL in different MARL tasks.

Multiple quadrotor uncrewed aerial vehicles (UAVs) systems have garnered widespread research interest and fostered tremendous interesting applications, especially in multiconstrained pursuit-evasion games (MC-PEGs). The cooperative evasion and formation coverage (CEFC) task, where the UAV swarm aims to maximize formation coverage across multiple target zones while collaboratively evading predators, belongs to one of the most challenging issues in MC-PEGs, especially under communication-limited constraints. This multifaceted problem, which intertwines responses to obstacles, adversaries, target zones, and formation dynamics, brings up significant high-dimensional complications in locating a solution. In this article, we propose a novel two-level framework [i.e., consensus inference-based hierarchical reinforcement learning (CI-HRL)], which delegates target localization to a high-level policy, while adopting a low-level policy to manage obstacle avoidance, navigation, and formation. Specifically, in the high-level policy, we develop a novel multiagent reinforcement learning (RL) module, consensus-oriented multiagent communication (ConsMAC), to enable agents to perceive global information and establish consensus from local states by effectively aggregating neighbor messages. Meanwhile, we leverage an alternative training-based MAPPO (AT-M) and policy distillation to accomplish the low-level control. The experimental results, including the high-fidelity software-in-the-loop (SITL) simulations, validate that CI-HRL provides a superior solution with enhanced swarm's collaborative evasion and task completion capabilities.

Semantic Communication (SemCom) demonstrates strong superiority over conventional bit-level accurate transmission, by only attempting to recover the essential semantic information of data. Nevertheless, most SemCom works train the whole system in an End-to-End (E2E) way, with the assumption of a differentiable channel which is rare in reality applications. In this paper, to tackle the non-differentiability of channels, we propose an alternate learning-based sparse SemCom system with an SNR-adaptive capability for visual transmission, named SparseSBC-SADM. Specially, SparseSBC-SADM leverages two separate Deep Neural Network (DNN)-based models at the transmitter (TX) and receiver (RX), respectively. It alternates between learning the encoding and decoding processes, rather than the joint optimization commonly found in existing literature, to solve the non-differentiability in the channel. In particular, a “self-critic” training scheme is leveraged for stable training. Moreover, the DNN-based TX generates a sparse set of bits in deduced “semantic bases”, by further incorporating a binary quantization module by combining Compressive Sensing (CS) and DNN on the basis of minimal detrimental effect to the semantic accuracy. Furthermore, enlightened from the denoising steps in the Denoising Diffusion Model (DDM), a lightweight, SNR-Adaptive Denoising Module (SADM) is provisionally deployed at RX to improve data reconstruction with a gate mechanism to determine the activation under poor channel conditions. Extensive simulation results validate that SparseSBC-SADM shows efficient and effective transmission performance under various channel conditions, and outperforms typical SemCom solutions.

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.

Faced with the abundance of mobile ad-hoc network (MANET) applications, there emerges a strong incentive to provision MANET in millimeter wave. However, the deafness of directional antennas and the decentralized structure of MANET make it difficult to achieve consistent medium access control (MAC) among nodes through random competition. Therefore, graph coloring-based MAC (GC-MAC) scheme is proposed to implement time division multiplexed scheduling, but it allocates equal slots to links, disregarding the unbalanced and piecewise stationary traffic distribution in task-driven MANETs. Here, we propose a continual multi-agent reinforcement learning (RL)-assisted traffic-adaptive scheme to enhance the agility of slot allocation. Specifically, we add a contention period to frames of GC-MAC, during which nodes analyze stochastic characteristics of traffic to derive link traffic distribution for each task, and adjust slot assignment to reach cooperation through decentralized multi-agent deep Q-network (MA-DQN). Besides, considering the traffic variation due to task switch, continual RL is incorporated to accommodate changes of the environment more sensitively. Finally, simulation results prove the proposed scheme achieves faster convergence speed, lower delay and higher throughput.

The wide applications of artificial intelligence (AI) on massive Internet-of-things or smartphones raises significant concerns about privacy, heterogeneity, and resource efficiency. Correspondingly, federated learning emerges as an effective way to enable AI over massively distributed nodes without uploading the raw data. Conventional works mostly focus on learning a single unified model for one solitary task. Multi-task learning (MTL) outperforms single-task learning by training multiple models concurrently, leading to reduced model sizes and increased flexibility. However, existing federated learning efforts often face challenges in efficiently managing MTL scenarios, particularly with the presence of stragglers, without incurring prohibitive computation and communication costs. In this paper, inspired by the natural cloud-BS-terminal hierarchy of cellular networks, we provide a viable resource-aware hierarchical federated MTL (RHFedMTL) solution to meet the task heterogeneity corresponding to different non-IID (independent and identically distributed) training datasets. Specifically, a primal-dual method has been leveraged to effectively transform the coupled MTL into some local optimization sub-problems within BSs. Therefore, it enables solving different tasks within a BS and aggregating the multi-task result in the cloud without uploading the raw data. Furthermore, compared with existing methods that reduce resource costs by simply changing the aggregation frequency, we dive into the intricate relationship between resource consumption and learning accuracy, and develop a resource-aware learning strategy for adjusting the iteration number on local terminals and BSs to meet the resource budget. Extensive simulation results demonstrate the effectiveness and superiority of RHFedMTL in terms of improving the learning accuracy and boosting the convergence rate.

The surging demand for high-definition video streaming services and large neural network models implies a tremendous explosion of Internet traffic. To mitigate the traffic pressure, architectures with in-network storage have been proposed to cache popular contents at devices in closer proximity to users. Correspondingly, in order to maximize caching utilization, it becomes essential to devise an effective popularity prediction method. In that regard, predicting popularity with dynamic graph neural network (DGNN) models achieves remarkable performance. However, DGNN models still suffer from tackling sparse datasets where most users are inactive. Therefore, we propose a reformative temporal graph network, named semantics-enhanced temporal graph network (STGN), which attaches extra semantic information into the user-content bipartite graph and could better leverage implicit relationships behind the superficial topology structure. On top of that, we customize its temporal and structural learning modules to further boost the prediction performance. Specifically, in order to efficiently aggregate the diversified semantics that a content might possess, we design a user-specific attention (UsAttn) mechanism for the temporal learning. Unlike the attention mechanism that only analyzes the influence of genres on content, UsAttn also considers the attraction of semantic information to a specific user. Meanwhile, as for the structural learning, we introduce the concept of positional encoding into our attention-based graph learning and novelly adopt a semantic positional encoding (SPE) function, which effectively boost the performance of lightweight algorithms. Finally, extensive simulations verify the superiority of our models and demonstrate their effectiveness in content caching.

The paper considers independent reinforcement learning (IRL) for multi-agent collaborative decision-making in the paradigm of federated learning (FL). However, FL generates excessive communication overheads between agents and a remote central server, especially when it involves a large number of agents or iterations. Besides, due to the heterogeneity of independent learning environments, multiple agents may undergo asynchronous Markov decision processes (MDPs), which will affect the training samples and the model’s convergence performance. On top of the variation-aware periodic averaging (VPA) method and the policy-based deep reinforcement learning (DRL) algorithm (i.e., proximal policy optimization (PPO)), this paper proposes two advanced optimization schemes orienting to stochastic gradient descent (SGD): 1) A decay-based scheme gradually decays the weights of a model’s local gradients with the progress of successive local updates, and 2) By representing the agents as a graph, a consensus-based scheme studies the impact of exchanging a model’s local gradients among nearby agents from an algebraic connectivity perspective. This paper also provides novel convergence guarantees for both developed schemes, and demonstrates their superior effectiveness and efficiency in improving the system’s utility value through theoretical analyses and simulation results.

Multi-Agent Reinforcement Learning (MARL) has become a classic paradigm to solve diverse, intelligent control tasks like autonomous driving in Internet of Vehicles (IoV). However, the widely assumed existence of a central node to implement centralized federated learning-assisted MARL might be impractical in highly dynamic scenarios, and the excessive communication overheads possibly overwhelm the IoV system. Therefore, in this paper, we design a communication efficient cooperative MARL algorithm, named RSM-MAPPO, to reduce the communication overheads in a fully distributed architecture. In particular, RSM-MAPPO enhances the multi-agent Proximal Policy Optimization (PPO) by incorporating the idea of segment mixture and augmenting multiple model replicas from received neighboring policy segments. Afterwards, RSM-MAPPO adopts a theory-guided metric to regulate the selection of contributive replicas to guarantee the policy improvement. Finally, extensive simulations in a mixed-autonomy traffic control scenario verify the effectiveness of the RSM-MAPPO algorithm.

The enormous amount of network equipment and users implies a tremendous growth of Internet traffic for multi-media services. To mitigate the traffic pressure, architectures with in-network storage have been proposed to cache popular content at devices in close proximity to users in order to decrease the number of backhaul hops. Meanwhile, the reduced transmission distance also contributes to energy saving. However, due to limited storage, only a fraction of the content can be cached, while caching the most popular content is cost-effective. Correspondingly, it becomes essential to devise an effective popularity prediction method. In this regard, some existing efforts manifest the effectiveness of dynamic graph neural network (DGNN) models, but it remains challenging to tackle sparse datasets. Herein, we first propose a reformative temporal graph network, named STGN, to address the challenge and improve prediction performance. Specifically, the STGN model leverages extra semantic messages to help establish implicit paths within the sparse interaction graph and enhance the temporal and structural learning of a DGNN model. Furthermore, we devise a user-specific attention mechanism to aggregate various semantics in a fine-grained manner. Finally, extensive simulations verify the superiority of our STGN models and demonstrate the potential in terms of energy-saving.

Semantic communication (SemCom) demonstrates strong superiority over conventional bit-level accurate transmission, by only attempting to recover the essential semantic information of data. In this paper, in order to tackle the non-differentiability of channels, we propose an alternate learning based SemCom system for visual transmission, named Spars-eSBC. Specially, SparseSBC leverages two separate Deep Neural Network (DNN)-based models at the transmitter and receiver, respectively, and learns the encoding and decoding in an alternate manner, rather than the joint optimization in existing literature, so as to solving the non-differentiability in the channel. In particular, a "self-critic" training scheme is leveraged for stable training. Moreover, the DNN-based transmitter generates a sparse set of bits in deduced "semantic bases", by further incorporating a binary quantization module on the basis of minimal detrimental effect to the semantic accuracy. Extensive simulation results validate that SparseSBC shows efficient and effective transmission performance under various channel conditions, and outperforms typical SemCom solutions.

In recent years, unmanned aerial vehicle (UAV) ad-hoc networks have achieved rapid development due to their autonomy and high reliability. Typically, clustering is widely adopted to reduce the degradation of network performance in large-scale UAV networks. However, due to the limited channel resources and complex communication environment, it becomes challenging to obtain complete information needed for clustering from remote nodes. In addition, the high dimension and non-linear relationships of data also make large-scale clustering diffcult. Therefore, in this paper, we propose a noval distributed clustering scheme. First, a matrix factorization (MF) based time series algorithm is proposed to predict and complement the incomplete information. Second, we adopt deep autoencoder to wisely incorporate the non-linear relationship of information needed for clustering. Finally we extract features as input to k-means to obtain clustering results. Simulation results demonstrate the effectiveness of the proposed clustering scheme.

Multi-Agent Reinforcement Learning (MARL) has reached astonishing achievements in various fields such as the traffic control of vehicles in a wireless connected environment. In MARL, how to effectively decompose a global feedback into the relative contributions of individual agents belongs to one of the most fundamental problems. However, the volatility of the environment (e.g., the vehicle movement and wireless disturbance) could significantly shape the time-varying topological relationships among agents, thus making the Value Decomposition (VD) challenging. Therefore, in order to cope with this annoying volatility, it becomes imperative to design a dynamic VD framework. Hence, in this paper, we propose a novel Stochastic VMIX (SVMIX) methodology by embedding the dynamic topological features into the VD and incorporating the corresponding components into a multi-agent actor-critic architecture. In particular, the Stochastic Graph Neural Network (SGNN) is leveraged to effectively extract underlying dynamics embedded in topological features and improve the flexibility of VD against the environment volatility. Finally, the superiority of SVMIX is verified through extensive simulations.

The imaging genetics approach generates large amount of high dimensional and multi-modal data, providing complementary information for comprehensive study of Schizophrenia, a complex mental disease. However, at the same time, the variety of these data in structures, resolutions, and formats makes their integrative study a forbidding task. In this paper, we propose a novel model called Joint Sparse Collaborative Regression (JSCoReg), which can extract class-specific features from different health conditions/disease classes. We first evaluate the performance of feature selection in terms of Receiver operating characteristic curve and the area under the ROC curve in the simulation experiment. We demonstrate that the JSCoReg model can achieve higher accuracy compared with similar models including Joint Sparse Canonical Correlation Analysis and Sparse Collaborative Regression. We then applied the JSCoReg model to the analysis of schizophrenia dataset collected from the Mind Clinical Imaging Consortium. The JSCoReg enables us to better identify biomarkers associated with schizophrenia, which are verified to be both biologically and statistically significant.

Synthetic aperture radar interferometry (InSAR) technology has been widely used for landslide monitoring in mountainous areas. The troposphere in steep mountainous areas is affected by the variable topography, temperature, and humidity, which differs from that in plain areas and thus exhibits large spatial heterogeneity. Traditional InSAR troposphere correction methods are limited in this area, and the accuracy of InSAR measurements will be significantly affected. In this paper, we proposed a tropospheric delay correction method based on deep learning (AtmNet) without external data considering the spatial-heterogenetiy in each individual interferogram. The tropospheric correction and landslides monitoring based on Sentinel-1 SAR data was carried out in Mao County, a high landslide-prone area in southwest Sichuan Province (China). A simulation experiment was conducted to analyze the adaptability of the model and evaluate the effectiveness of the AtmNet method. Furthermore, we demonstrated the good performance of the AtmNet method through a comparison with the linear model (LM) and GACOS method, revealing that the proposed method could effectively model the spatial heterogeneity of tropospheric delay in steep mountains. The slope displacements that cannot be seen in the interferogram were very clear after the tropospheric delay correction. This method provides important technical support for the accurate DInSAR and time-series InSAR for landslide monitoring in steep mountainous areas in the future.

Mobile ad-hoc network (MANET) has attracted ex-tensive attention in many applications with nodes operating over bandwidth-constrained wireless links in a self-organized manner. Typically, to avoid potential collisions, CSMA/CA method is widely adopted in MANETs and affects the provisioning quality of service (QoS). However, due to the potentially severe collisions in a large-scale MANET, current CSMA/CA methods fail to support QoS effectively, especially for scenarios with different priority services. Here, we propose a priority-aware CSMA/CA strategy to use the higher-priority packets to collect the queueing information from adjacent nodes in a piggyback manner and model the channel access process by a partial observation Markov decision process (POMDP). Furthermore mean-field multi-agent reinforcement learning (MARL) is adopted to allow each individual node in the MANET to dynamically select the appropriate contention window (CW) and gradually compute the global optimal policy. Finally, extensive simulation results verify significantly lower delay and packet loss rate for the higher- priority services and comparable provisioning quality for lower- priority services, which reflects the superiority over baselines.

The paper considers independent reinforcement learning (IRL) for multi-agent decision-making process in the paradigm of federated learning (FL). We show that FL can clearly improve the policy performance of IRL in terms of training efficiency and stability. However, since the policy parameters are trained locally and aggregated iteratively through a central server in FL, frequent information exchange incurs a large amount of communication overheads. To reach a good balance between improving the model's convergence performance and reducing the required communication and computation over-heads, this paper proposes a system utility function and develops a consensus-based optimization scheme on top of the periodic averaging method, which introduces the consensus algorithm into FL for the exchange of a model's local gradients. This paper also provides novel convergence guarantees for the developed method, and demonstrates its superior effectiveness and efficiency in improving the system utility value through theoretical analyses and numerical simulation results.

Federated learning is an effective way to enable artificial intelligence over massive distributed nodes with security and communication efficiency. Some previous works primarily focus on learning a single global model for a unique task across the network, which is less competent to handle multi-task scenarios with stragglers and fault, after adopting the general gradient update methods in a federated environment. Others aim to learn a distinct model for each node, which is expensive in terms of the computation and communication cost. Using hierarchical network to reduce communication cost is becoming a new candidate. Thus, we propose a primal-and-dual method-based hierarchical federated multi-task learning system, supported with HFedMTL algorithm that allows massive nodes from distributed areas to join in the federated multi-task learning process. Empirical experiments verify the analysis and demonstrate the benefits of improving the learning performance and convergence rate.

With the development of network technology and the rapid growth of network equipment, the data throughput in the network is sharply increasing. To meet people’s requirements for low latency, the network architecture like Information-Centric Network (ICN) proposes to keep part of the content at the edge of network. In this paper, to maximize the cache hit rate, we propose a prediction model based on dynamic graph neural network (DGNN) to jointly learn the structural and temporal patterns embedded in the bipartite graph between users and visited content for predicting the content popularity. Furthermore, in order to strengthen the dynamic learning of graphs, we propose an age of information (AoI) based attention mechanism to extract useful historical information while avoiding the problem of message staleness. Extensive simulation results demonstrate that our model can obtain higher prediction accuracy, and generate a caching policy with boosted caching hits.

In this paper, we aim to propose a novel transmission control protocol (TCP) congestion control method from a cross-layer-based perspective and present a deep reinforcement learning (DRL)-driven method called DRL-3R (DRL for congestion control with Radio access network information and Reward Redistribution) so as to learn the TCP congestion control policy in a superior manner. In particular, we incorporate the RAN information to timely grasp the dynamics of RAN, and empower DRL to learn from the delayed RAN information feedback potentially induced by several consecutive actions. Meanwhile, we relax the implicit assumption (that the feedback to one specific action returns at a round-trip-time (RTT) after the action is applied) in previous researches, by redistributing the rewards and evaluating the merits of actions more accurately. Experiment results show that besides maintaining a reasonable fairness, DRL-3R significantly outperforms classical congestion control methods (e.g., TCP Reno, Westwood, Cubic, BBR and DRL-CC) on network utility by achieving a higher throughput while reducing delay in various network environments.

With the rapid evolution of wireless mobile devices, there emerges an increased need to design effective collaboration mechanisms between intelligent agents to gradually approach the final collective objective by continuously learning from the environment based on their individual observations. In this regard, independent reinforcement learning (IRL) is often deployed in multiagent collaboration to alleviate the problem of a nonstationary learning environment. However, behavioral strategies of intelligent agents in IRL can be formulated only upon their local individual observations of the global environment, and appropriate communication mechanisms must be introduced to reduce their behavioral localities. In this article, we address the problem of communication between intelligent agents in IRL by jointly adopting mechanisms with two different scales. For the large scale, we introduce the stigmergy mechanism as an indirect communication bridge between independent learning agents, and carefully design a mathematical method to indicate the impact of digital pheromone. For the small scale, we propose a conflict-avoidance mechanism between adjacent agents by implementing an additionally embedded neural network to provide more opportunities for participants with higher action priorities. In addition, we present a federal training method to effectively optimize the neural network of each agent in a decentralized manner. Finally, we establish a simulation scenario in which a number of mobile agents in a certain area move automatically to form a specified target shape. Extensive simulations demonstrate the effectiveness of our proposed method.

Network slicing (NS) devotes to provisioning various services with distinct requirements over the same physical communication infrastructure. Considering a dense cellular network scenario that contains several NS over multiple base stations (BSs), it remains challenging to design a proper resource management strategy in real time, so as to cope with frequent BS handover and meet distinct service requirements. In this paper, we propose to formulate this challenge as a multiagent reinforcement learning (MARL) problem and leverage graph attention network (GAT) to strengthen the cooperation between agents. Furthermore, we incorporate GAT into deep Q network (DQN) and correspondingly design an intelligent resource management strategy for NS. Finally, we verify the superiority of the GAT-based DQN algorithm through extensive simulations.

The maximum capacity of fractal D2D (device-to-device) social networks with both direct and hierarchical communications is studied in this paper. Specifically, the fractal networks are characterized by the direct social connection and the self-similarity. First, for a fractal D2D social network with direct social communications, it is proved that the maximum capacity is$ \Theta (\frac{1}{\sqrt{n\,\log n}})$if a user communicates with one of his/her direct contacts randomly, where$ n$denotes the total number of users in the network, and it can reach up to$ \Theta (\frac{1}{\log n})$if any pair of social contacts with distance$ d$communicate according to the probability in proportion to$ d^{-\beta }$. Second, since users might get in touch with others without direct social connections through the inter-connected multiple users, the fractal D2D social network with these hierarchical communications is studied as well, and the related capacity is further derived. Our results show that this capacity is mainly affected by the correlation exponent$\epsilon$of the fractal structure. The capacity is reduced in proportional to$ \frac{1}{{\log n}}$if$ 2<\epsilon <3$, while the reduction coefficient is$ \frac{1}{n}$if$ \epsilon >3$.

Nowadays, deep neural networks (DNNs) have been rapidly deployed to realize a number of functionalities like sensing, imaging, classification, recognition, etc. However, the computational-intensive requirement of DNNs makes it difficult to be applicable for resource-limited Internet of Things IoT devices. In this paper, we propose a novel pruning-based paradigm that aims to reduce the computational cost of DNNs, by uncovering a more compact structure and learning the effective weights therein, on the basis of not compromising the expressive capability of DNNs. In particular, our algorithm can achieve efficient end-to-end training that transfers a redundant neural network to a compact one with a specifically targeted compression rate directly. We comprehensively evaluate our approach on various representative benchmark datasets and compared with typical advanced convolutional neural network (CNN) architectures. The experimental results verify the superior performance and robust effectiveness of our scheme. For example, when pruning VGG on CIFAR-10, our proposed scheme is able to significantly reduce its FLOPs (floating-point operations) and number of parameters with a proportion of 76.2% and 94.1%, respectively, while still maintaining a satisfactory accuracy. To sum up, our scheme could facilitate the integration of DNNs into the common machine-learning-based IoT framework, and establish distributed training of neural networks in both cloud and edge.

In this paper, we propose a novel network training mechanism called “dynamic channel propagation” to prune the neural networks during the training period. In particular, we pick up a specific group of channels in each convolutional layer to participate in the forward propagation in training time according to the significance level of channel, which is defined as channel utility. The utility values with respect to all selected channels are updated simultaneously with the error back-propagation process and will adaptively change. Furthermore, when the training ends, channels with high utility values are retained whereas those with low utility values are discarded. Hence, our proposed scheme trains and prunes neural networks simultaneously. We empirically evaluate our novel training scheme on various representative benchmark datasets and advanced convolutional neural network (CNN) architectures, including VGGNet and ResNet. The experiment results verify the superior performance and robust effectiveness of our approach.

Multi-agent system (MAS) needs to mobilize multiple simple agents to complete complex tasks. However, it is difficult to coherently coordinate distributed agents by means of limited local information. In this demo, we propose a decentralized collaboration method named as "stigmergy" in network-assisted MAS, by exploiting digital pheromones (DP) as an indirect medium of communication and utilizing deep reinforcement learning (DRL) on top. Correspondingly, we implement an experimental platform, where KHEPERA IV robots form targeted specific shapes in a decentralized manner. Experimental results demonstrate the effectiveness and efficiency of the proposed method. Our platform could be conveniently extended to investigate the impact of network factors (e.g., latency, data rate, etc).

Network slicing is a key technology in 5G communications system. Its purpose is to dynamically and efficiently allocate resources for diversified services with distinct requirements over a common underlying physical infrastructure. Therein, demand-aware resource allocation is of significant importance to network slicing. In this paper, we consider a scenario that contains several slices in a radio access network with base stations that share the same physical resources (e.g., bandwidth or slots). We leverage deep reinforcement learning (DRL) to solve this problem by considering the varying service demands as the environment state and the allocated resources as the environment action. In order to reduce the effects of the annoying randomness and noise embedded in the received service level agreement (SLA) satisfaction ratio (SSR) and spectrum efficiency (SE), we primarily propose generative adversarial network-powered deep distributional Q network (GAN-DDQN) to learn the action-value distribution driven by minimizing the discrepancy between the estimated action-value distribution and the target action-value distribution. We put forward a reward-clipping mechanism to stabilize GAN-DDQN training against the effects of widely-spanning utility values. Moreover, we further develop Dueling GAN-DDQN, which uses a specially designed dueling generator, to learn the action-value distribution by estimating the state-value distribution and the action advantage function. Finally, we verify the performance of the proposed GAN-DDQN and Dueling GAN-DDQN algorithms through extensive simulations.

Coordinating multiple unmanned aerial vehicles (multi-UAVs) is a challenging technique in highly dynamic and sophisticated environments. Based on digital pheromones as well as current mainstream unmanned system controlling algorithms, we propose a strategy for multi-UAVs to acquire targets with limited prior knowledge. In particular, we put forward a more reasonable and effective pheromone update mechanism, by improving digital pheromone fusion algorithms for different semantic pheromones and planning individuals’ probabilistic behavioral decision-making schemes. Also, inspired by the flocking model in nature, considering the limitations of some individuals in perception and communication, we design a navigation algorithm model on top of Olfati-Saber’s algorithm for flocking control, by further replacing the pheromone scalar to a vector. Simulation results show that the proposed algorithm can yield superior performance in terms of coverage, detection and revisit efficiency, and the capability of obstacle avoidance.

The traffic in wireless networks has become diverse and fluctuating both spatially and temporally due to the emergence of new wireless applications and the complexity of scenarios. The purpose of this paper is to quantitatively analyze the impact of the wireless traffic, which fluctuates both spatially and temporally, on the performance of the wireless networks. Specially, we propose to combine the tools from stochastic geometry and queueing theory to model the spatial and temporal fluctuation of traffic, which to our best knowledge has seldom been evaluated analytically. We derive the spatial and temporal statistics, the total arrival rate, the stability of queues and the delay of users by considering two different spatial properties of traffic, i.e., the uniformly and non-uniformly distributed cases. The numerical results indicate that although the fluctuation of traffic (reflected by the variance of total arrival rate) when the users are clustered is much fiercer than that when the users are uniformly distributed, the unstable probability is smaller. Our work provides a useful reference for the design of wireless networks when the complex spatio-temporal fluctuation of the traffic is considered.

Network slicing is a key technology in 5G communications system, which aims to dynamically and efficiently allocate resources for diversified services with distinct requirements over a common underlying physical infrastructure. Therein, demand-aware allocation is of significant importance to network slicing. In this paper, we consider a scenario that contains several slices in one base station on sharing the same bandwidth. Deep reinforcement learning (DRL) is leveraged to solve this problem by regarding the varying demands and the allocated bandwidth as the environment state and action, respectively. In order to obtain better quality of experience (QoE) satisfaction ratio and spectrum efficiency (SE), we propose generative adversarial network (GAN) based deep distributional Q network (GAN-DDQN) to learn the distribution of state-action values. Furthermore, we estimate the distributions by approximating a full quantile function, which can make the training error more controllable. In order to protect the stability of GAN-DDQN's training process from the widely-spanning utility values, we also put forward a reward-clipping mechanism. Finally, we verify the performance of the proposed GAN-DDQN algorithm through extensive simulations.

Software-defined security (SDS) overcomes the limitations of traditional security mechanisms, which brings significant merits for design, deployment and management. However, existing researches are usually limited to some independent algorithms, while not able to apply multiple algorithms to accommodate various types of attack in actual deployment. In this paper, we propose and implement a novel SDS framework, which aims to flexibly deploy a variety of security functions and artificial intelligence (AI) algorithms to automatically learn ongoing threats and proactively protect the network from attacks.

In this study, the authors consider the utilisation of multicast technology in cellular networks given different arrival patterns for the content requests of mobile users. Traditionally, the performance evaluation of multicast in the literature usually depends on the adoption of temporal Poisson processes for content requests, which is not accurate any more according to many real data measurements. Therefore, to make use of the bursty nature of content requests, they propose a hybrid unicast/multicast strategy where the base station (BS) can perform the unicast or multicast procedure according to its serving status. By modelling the complete process into a circular Markov chain, they derive the average latency of content requests and the average power consumption of BSs under different arrival patterns and serving configurations in theoretical and/or simulative ways. Moreover, the multicast threshold introduced in their strategy can be dynamically adjusted to achieve a joint optimisation between average latency and power consumption when confronted with varied demands. Numerous results show that the proposed strategy can not only reduce the average latency of content requests but also decrease the average power consumption of BSs, especially under the bursty request arrival patterns.

Software defined Internet of Things (SD-IoT) networks profit from centralized management and interactive resource sharing, which enhances the efficiency and scalability of Internet of Things applications. But with the rapid growth in services and applications, they are vulnerable to possible attacks and face severe security challenges. Intrusion detection has been widely used to ensure network security, but classical detection methods are usually signature-based or explicit-behavior-based and fail to detect unknown attacks intelligently, which makes it hard to satisfy the requirements of SD-IoT networks. In this paper, we propose an artificial intelligence-based two-stage intrusion detection empowered by software defined technology. It flexibly captures network flows with a global view and detects attacks intelligently. We first leverage Bat algorithm with swarm division and binary differential mutation to select typical features. Then, we exploit Random Forest through adaptively altering the weights of samples using the weighted voting mechanism to classify flows. Evaluation results prove that the modified intelligent algorithms select more important features and achieve superior performance in flow classification. It is also verified that our solution shows better accuracy with lower overhead compared with existing solutions.

To understand the spatial deployment of base stations (BSs) is the first step to analyze the performance of cellular networks and further design efficient networking protocols. Poisson point process (PPP), which has been widely adopted to characterize the deployment of BSs and established the reputation to give tractable results in the stochastic geometry analyses, usually assumes a static BS deployment density in homogeneous PPP (HPPP) models or delicately designed location-dependent density functions in in-homogeneous PPP models. However, the simultaneous existence of attractiveness and repulsiveness among BSs practically deployed in a large-scale area defies such an assumption, and the α-stable distribution, one kind of heavy-tailed distributions, has recently demonstrated superior accuracy to statistically model the varying BS density in different areas. In this paper, we start with these new findings and investigate the intrinsic feature (i.e., the spatial self-similarity) embedded in the BSs. Afterwards, we refer to a generalized PPP setup with α-stable distributed density and theoretically derive the related coverage probability. In particular, we give an upper bound of the derived coverage probability for high signal-to-interference-plus-noise ratio thresholds and show the monotonically decreasing property of this bound with respect to the variance of BS density. Besides, we prove that our model could reduce to the single-tier HPPP for some special cases and demonstrate the superior accuracy of the α-stable model to approach the real environment.

Traffic prediction plays an important role in evaluating the performance of telecommunication networks and attracts intense research interests. A significant number of algorithms and models have been put forward to analyse traffic data and make prediction. In the recent big data era, deep learning has been exploited to mine the profound information hidden in the data. In particular, Long Short-Term Memory (LSTM), one kind of Recurrent Neural Network (RNN) schemes, has attracted a lot of attentions due to its capability of processing the long-range dependency embedded in the sequential traffic data. However, LSTM has considerable computational cost, which can not be tolerated in tasks with stringent latency requirement. In this paper, we propose a deep learning model based on LSTM, called Random Connectivity LSTM (RCLSTM). Compared to the conventional LSTM, RCLSTM makes a notable breakthrough in the formation of neural network, which is that the neurons are connected in a stochastic manner rather than full connected. We apply the RCLSTM to predict traffic and validate that the RCLSTM with even 35% neural connectivity still shows a satisfactory performance. When we gradually add training samples, the performance of RCLSTM becomes increasingly closer to the baseline LSTM. Moreover, for the input traffic sequences of enough length, the RCLSTM exhibits even superior prediction accuracy than the baseline LSTM.

The rapid urbanisation has transformed cities to the preferential human settlement and allowed cities to quietly witness all range of human activities. As the key enabler in the information and communications technology industry, cellular networks play a decisive role in delivering communication messages and entertainment content. In particular, cellular network operators respond to human initiated service requests by gradually deploying necessary infrastructure and calibrating transmission protocols. Hence, cellular network records encompass the interesting interaction between human‐initiated messages and network‐triggered responses. In this study, the authors collect the ‘big data’ in urban cellular networks and try to dig out the human and urban planning properties. Specifically, they focus on the statistical modelling of three representative scenarios like spatial deployment density of base stations, packet length or traffic volume of mobile services, as well as inter‐arrival time and dwell time of human mobility. Through extensive data mining, they validate the heavy‐tailed feature universally existing in these scenarios. Afterwards, they discuss the implications of this heavy‐tailed feature and talk about its fundamental contribution to intelligent resource adjustment, proactive content caching, and enhanced connection management in cellular networks. Finally, they highlight the applications of this feature towards smarter cellular networks and cities.

When we talk about the queueing theory in cellular networks, M/M/1 always comes first, which supposes the inter-arrival time and service time are both exponentially distributed. However, with the prosperity of Mobile Instant Messaging (MIM), many researchers found that the inter-arrival times of users' MIM follow lognormal distribution and there are also many keep-alive(KA) messages produced by the keep-alive mechanism accompanied with the users' messages. Based on this discovery, we rethink the queueing model that people might apply to analyze the MIM traffic. Firstly, we focus on a queueing system with lognormally distributed inter-arrival time and exponentially distributed service time, and find that the time a customer spends for waiting in queue and the length of the queue are shorter and smaller than the situation that the inter-arrival time is exponential one as proposed by 3GPP. Secondly, we study the influence of the keep-alive mechanism on queueing process. After that, we carry out simulations for this model, and the numerical results show that there exists an optimal KA duration, which determines the minimal average waiting time.

Full-duplex (FD) networks promise nearly doubled spectrum efficiency for single links with beyond 100 dB self-interference cancellation capability. However, the performance of FD networking is not that straightforward, as there will be several new types of interference introduced. Hence, it is critical to exploit some radio resource management techniques to combat the annoying interference and inevitably needs to reserve part of radio resources in HD mode to obtain the essential auxiliary information. It is naturally to ask which physical channels and signals could be left in the FD mode and how much system gain FD cellular networks could bring? In this paper, we try to answer these questions. We first propose the feasibility to leave one kind of physical channels or signals in the FD mode and then highlight some potential resource management schemes. We will also demonstrate the system gain of FD networks by leveraging our proposed design and the management techniques. Our simulation results show FD communications leads to promising performance improvement.

In conventional cellular networks, for base stations (BSs) that are deployed far away from each other, it is general to assume them to be mutually independent. Nevertheless, after long-term evolution of cellular networks in various generations, the assumption no longer holds. Instead, the BSs, which seem to be gradually deployed by operators in a casual manner, have embedded many fundamental features in their locations, coverage and traffic loading. Their features can be leveraged to analyze the intrinstic pattern in BSs and even human community. According to large-scale measurement datasets, we build spatial correlation model of BSs by utilizing one of the most important features, that is, traffic. Coupling with the theory of complex networks, we make further analysis on the structure and characteristics of this spatial correlation model. Simulation results show that its degree distribution follows scale-free property. Moreover, the datasets also unveil the characteristics of fractality and small-world. Furthermore, we study how to choose the appropriate metric to measure the importance of each BS with a combination of network efficiency and demonstrate that degree is relatively more important.

Recent years have witnessed the unprecedented surge of mobile traffic and base stations (BSs) deployment, which poses severe requirement for future communications systems. Understanding the distribution dynamics of traffic and BSs in time‐space domain is of vital importance for better network design and resource management in cellular networks. In this study, a study on the statistical characteristics of cellular traffic series is carried out and ‐stable distribution is verified to be valid for modelling the traffic series of each BS. On the other hand, inspired by the fact that BSs traffic series are spatially correlated, the authors study the statistical relationship between the correlation coefficient and the distance between BSs. Moreover, ‐stable model is also suitable to describe the BSs deployment, thus conducing to prove the existence of self‐similarity. In addition, both the traffic time series and the BSs spatial distribution are deeply associated with heterogeneity, so they come up with the density‐based and distance‐based methods to quantify their heterogeneous degree.

In this paper, a delay diversity orthogonal frequency-division multiplexing (DD OFDM) transmission scheme in amplify-and-forward (AF) full-duplex relay systems is investigated. One direct source-to-destination link, one relay forwarding link, and residual self-interference are considered in the system. The necessary cyclic prefix length is investigated and a suitable AF relay protocol in the full-duplex relay OFDM system is proposed. This paper demonstrates that the AF relay link and the direct source-to-destination link can be combined to provide spatial diversity. The key is that the DD OFDM scheme is used to transform the spatial diversity into increased channel frequency diversity that is further exploited by using the bit-interleaved coding. The bit error rate performance of the proposed system is verified by simulation results.

Traffic learning and prediction is at the heart of the evaluation of the performance of telecommunications networks and attracts a lot of attention in wired broadband networks. Now, benefiting from the big data in cellular networks, it becomes possible to make the analyses one step further into the application level. In this paper, we first collect a significant amount of application-level traffic data from cellular network operators. Afterward, with the aid of the traffic “big data,” we make a comprehensive study over the modeling and prediction framework of cellular network traffic. Our results solidly demonstrate that there universally exist some traffic statistical modeling characteristics at a service or application granularity, including α-stable modeled property in the temporal domain and the sparsity in the spatial domain. But, different service types of applications possess distinct parameter settings. Furthermore, we propose a new traffic prediction framework to encompass and explore these aforementioned characteristics and then develop a dictionary learning-based alternating direction method to solve it. Finally, we examine the effectiveness and robustness of the proposed framework for different types of application-level traffic. Our simulation results prove that the proposed framework could offer a unified solution for application-level traffic learning and prediction and significantly contribute to solve the modeling and forecasting issues.

In this paper, we address the tradeoff between energy efficiency (EE) and spectral efficiency (SE) for cellular networks with full-duplex (FD) communications enabled base stations. To be backward compatible with legacy LTE systems, it is assumed that user devices still work in the conventional half-duplex (HD) mode. There usually exists residual self-interference (RSI) in FD communications after advanced interference suppression techniques are applied. In this paper, we consider two different RSI models: constant RSI model and linear RSI model. First, the necessary conditions for an FD transceiver to achieve better EE-SE tradeoff than an HD one are derived for both the RSI models. Then, for the constant RSI model, a closed-form EE-SE expression is obtained in the scenario of single pair of users. We further extend our result and prove that EE is a quasi-concave function of SE in the scenario of multiple user pairs. Accordingly, an optimal algorithm to achieve the maximum EE based on the Lagrange dual decomposition technique is developed. For the linear RSI model, the EE-SE relation is difficult to deal with and we develop a heuristic algorithm by decoupling the problem into two sub-problems: power control and resource allocation. Our analysis and algorithms are finally verified by comprehensive numerical results.

Full-duplex (FD) communications benefit from recent breakthroughs in analog and digital signal processing and demonstrate nearly doubled spectrum efficiency for point-to-point links with beyond 100 dB self-interference cancellation capability. However, FD cellular networks with base stations (BSs) in FD mode and user terminals (UEs) in half-duplex mode also introduce intra-cell inter-UE interference and other kinds of interference, which might even eat up the potential gain. In this paper, we propose to take advantage of power control and inter-UE interference cancellation (IC) techniques to cope with the bothersome intra-cell inter-UE interference. We verify that the power control solution for single-cell FD network with a given user pair exhibits an interesting binary feature, that is, either both the BS and the uplink UE transmit at the maximum power or one of them completely mutes to achieve the optimal sum rate. Moreover, the binary feature holds even when inter-UE IC is applied, thus showing appealing computational efficiency. We also demonstrate significant performance improvement by applying binary power control with inter-UE IC to both singlecell and multi-cell FD networks.

Understanding traffic characteristics in cellular networks is of great significance for better network design and performance optimization. The rapid development of various social networking applications for smart devices makes it an imperative to carry out cellular data traffic analysis further into the application level. In this paper, based on a plenty of practical mobile data traffic records, we focus on three typical application types and draw conclusions in terms of statistical characteristics and appropriate distribution model for social mobile data traffic. Firstly, the universal existence of burstiness and self-similarity is demonstrated by testing traffic series at different time scales. Afterwards, α-stable distributions are used to model traffic series benefiting from their internal burstiness and self- similarity. The minor fitting errors verify the validity of α-stable model and a preliminary traffic prediction shows the usefulness of α-stable model for further traffic analysis.

As current cellular networks are becoming increasingly heterogeneous, traditional cell association rule based on the downlink reference signal receiving power (RSRP) no longer serves well. Under this circumstance, the concept of downlink (DL) and uplink (UL) decoupling (DUDe), in which user equipments choose the serving Base Station (BS) in DL and UL separately, has drawn great attention during the design of next generation cellular networks. In this paper, using the framework of stochastic geometry, we at first derive the theoretical energy efficiency (EE) distribution of a two-tier heterogeneous network both in DL and UL with DUDe. Then through numerical simulation, we verify that the DUDe rule can improve EE of a heterogeneous network comparing with the traditional RSRP rule. The improvement is shown to be significant through a further comparison with bias-based Cell Range Extension (CRE).

This paper proposes a distributed cooperative framework for improving the energy efficiency of green cellular networks. Based on the traffic load, neighboring base stations (BSs) cooperate to optimize the BS switching (sleeping) strategies so as to maximize the energy saving while guaranteeing users' minimal service requirements. The inter-BS cooperation is formulated following the principle of ecological self-organization. An interaction graph is defined to capture the network impact of the BS switching operation. Then, we formulate the problem of energy saving as a constrained graphical game, where each BS acts as a game player with the constraint of traffic load. The constrained graphical game is proved to be an exact constrained potential game. Furthermore, we prove the existence of a generalized Nash equilibrium (GNE), and the best GNE coincides with the optimal solution of total energy consumption minimization. Accordingly, we design a decentralized iterative algorithm to find the best GNE (i.e., the global optimum), where only local information exchange among the neighboring BSs is needed. Theoretical analysis and simulation results finally illustrate the convergence and optimality of the proposed algorithm.

Rapid growing demand for mobile traffic and severe service bursts generated by various mobile applications are challenging the capacities and management of future cellular networks. The measurement data from real cellular networks indicate that the mobile Internet traffic expresses long-range dependence (LRD) characteristics, which differs from the traditional exponential assumption as well as 3GPP reports, would greatly deteriorate the experience of the subscribers. On the other hand, the mobile cellular networks treat most services without differentiation, although different services have their own requirements on transmission rates and delay. After introducing a new metric called the Quality of Experience (QoE) index for different services and analyzing the influence of LRD traffic, two approaches are proposed to alleviate the experience deterioration caused by traffic bursts: cloud based baseband resource pool to improve the flexibility of networks and wireless context-aware service controller to manage the distinctive QoE requirements of different services. Finally, simulations validate that these approaches not only contribute to service-oriented cross-layer optimization methods to satisfy the network resource constraints, but also provide a software-controlled service management framework for cellular networks.

Poisson Point Process (PPP) has been widely adopted as an efficient model for the spatial distribution of base stations (BSs) in cellular networks. However, real BSs deployment are rarely completely random, due to environmental impact on actual site planning. Particularly, for multi-tier heterogeneous cellular networks, operators have to place different BSs according to local coverage and capacity requirement, and the diversity of BSs' functions may result in different spatial patterns on each networking tier. In this paper, we consider a two-tier scenario that consists of macrocell and microcell BSs in cellular networks. By analyzing these two tiers separately and applying both classical statistics and network performance as evaluation metrics, we obtain accurate spatial model of BSs deployment for each tier. Basically, we verify the inaccuracy of using PPP in BS locations modeling for either macrocells or microcells. Specifically, we find that the first tier with macrocell BSs is dispersed and can be precisely modelled by Strauss point process, while Matern cluster process captures the second tier's aggregation nature very well. These statistical models coincide with the inherent properties of macrocell and microcell BSs respectively, thus providing a new perspective in understanding the relationship between spatial structure and operational functions of BSs.

The spatially distributed sparsity of the mobile devices (MDs) in indoor wireless local area networks (WLANs) makes compressive sensing (CS) based localisation algorithms feasible and desirable. In this Letter, the authors exploit the most recent developments in CS to efficiently perform localisation in WLANs and design an accurate indoor localisation scheme by taking advantage of the theory of multi‐task Bayesian CS (MBCS). The proposed scheme assembles the strength measurements of signals from the MDs to distinct access points (APs) and jointly utilises them at a central unit or a specific AP to achieve localisation, thus being able to alleviate the burden of MDs while simultaneously giving a precise estimation of the locations. Afterwards, they give a deeper insight into the localisation problem in more practical scenarios with varying number of MDs and investigate two different adaptive algorithms to meet the satisfactory localisation error requirement. Compared with the conventional MBCS algorithms, simulation results validate that both adaptive algorithms could provide superior localisation accuracy and exhibit stronger resilience to the changes in the number of MDs.

Recent works have validated the possibility of improving energy efficiency in radio access networks (RANs), achieved by dynamically turning on/off some base stations (BSs). In this paper, we extend the research over BS switching operations, which should match up with traffic load variations. Instead of depending on the dynamic traffic loads which are still quite challenging to precisely forecast, we firstly formulate the traffic variations as a Markov decision process. Afterwards, in order to foresightedly minimize the energy consumption of RANs, we design a reinforcement learning framework based BS switching operation scheme. Furthermore, to speed up the ongoing learning process, a transfer actor-critic algorithm (TACT), which utilizes the transferred learning expertise in historical periods or neighboring regions, is proposed and provably converges. In the end, we evaluate our proposed scheme by extensive simulations under various practical configurations and show that the proposed TACT algorithm contributes to a performance jump start and demonstrates the feasibility of significant energy efficiency improvement at the expense of tolerable delay performance.

Recent works have validated the possibility of energy efficiency improvement in radio access networks (RAN), depending on dynamically turn on/off some base stations (BSs). In this paper, we extend the research over BS switching operation, matching up with traffic load variations. However, instead of depending on the predicted traffic loads, which is still quite challenging to precisely forecast, we formulate the traffic variation as a Markov decision process (MDP). Afterwards, in order to foresightedly minimize the energy consumption of RAN, we adopt the actor-critic method and design a reinforcement learning framework based BS switching operation scheme. In the end, we evaluate our proposed scheme by extensive simulations under various practical configurations and prove the feasibility of significant energy efficiency improvement.

In cellular networks, the base station power consumption is not simply proportional to the traffic loads of its coverage. As the traffic load fluctuates spatially and temporally, the base stations consequently suffer from heavy energy wastage when the traffic loads of their coverage are low. In this paper, we propose a grid-based energy saving scheme over predicted traffic loads. We firstly take advantage of the spatial-temporal pattern of traffic loads and employ the compressed sensing method to predict the future traffic loads. Then, we propose a grid-based energy saving scheme to improve the energy efficiency through turning some base stations into sleeping mode while ensuring the quality of service. Results of the simulation with real traffic loads1finally show the accuracy of the traffic load prediction and large energy efficiency improvement.