Federated fine-tuning of large language models (LLMs) provides a privacy-preserving approach to deploying pervasive generative AI services, yet the substantial memory overhead of first-order (FO) gradient computation presents significant practical challenges.While zeroth-order (ZO) optimization methods offer memory-efficient alternatives, they remain susceptible to performance degradation brought by data heterogeneity.Specifically, direct ZO-for-FO substitution is incompatible with existing strategies tailored for cross-client discrepancies.In response, we propose a new federated LLM fine-tuning framework, with a holistic revamped design of the entire ZO gradient processing pipeline.Crucially, with our proposed global adaptive optimization and local personalized perturbation, we present a unified solution for incorporating ZO gradients in federated learning, from local personalized perturbation sampling and ZO gradient transmission, to global ZO gradient reconstruction and aggregation with adaptive momentum, thereby directly addressing the challenges of inefficiencies and cross-client discrepancies.Our convergence analysis and experimental results demonstrate the superiority of our proposed framework over diverse heterogeneous data settings, both in terms of generalization and efficiency.

We introduce a new analytical framework, developed based on the spatial network calculus, for performance assessment of Low Earth Orbit (LEO) satellite networks. Specifically, we model the satellites' spatial positions as a strong ball-regulated point process on the sphere. Under this model, proximal points in space exhibit a locally repulsive property, reflecting the fact that intersatellite links are protected by a safety distance and would not be arbitrarily close. Subsequently, we derive analytical lower bounds on the conditional coverage probabilities under Nakagami-$m$ and Rayleigh fading, respectively. These expressions have a low computational complexity, enabling efficient numerical evaluations. We validate the effectiveness of our theoretical model by contrasting the coverage probability obtained from our analysis with that estimated from a Starlink constellation. The results show that our analysis provides a tight lower bound on the actual value and, surprisingly, matches the empirical simulations almost perfectly with a 1 dB shift. This demonstrates our framework as an appropriate theoretical model for LEO satellite networks.

The proliferation of mobile devices and real-time status updating applications has motivated the optimization of data freshness in the context of age of information (AoI). Meanwhile, increasing computational demands have inspired research on CPU scheduling. Since prior CPU scheduling strategies have ignored data freshness and prior age-minimization strategies have considered only constant CPU speed, we formulate the first CPU scheduling problem as a constrained semi-Markov decision process (SMDP) problem with uncountable space, which aims to minimize the long-term average age of information, subject to an average CPU power constraint. We optimize strategies that specify when the CPU sleeps and adapt the CPU speed (clock frequency) during the execution of update-processing tasks. We consider the age-minimal CPU scheduling problem for both predictable task size (PTS) and unpredictable task size (UTS) cases, where the task size is realized at the start (PTS) or at the completion (UTS) of the task, respectively. To address the non-convex objective, we employ Dinkelbach's fractional programming method to transform our problem into an average cost SMDP. We develop a value-iteration-based algorithm and prove its convergence to obtain optimal policies and structural results for both the PTS and UTS systems. Compared to constant CPU speed, numerical results show that our proposed scheme can reduce the AoI by 50\% or more, with increasing benefits under tighter power constraints. Further, for a given AoI target, the age-minimal CPU scheduling policy can reduce the energy consumption by 50\% or more, with greater AoI reductions when the task size distribution exhibits higher variance.

Federated learning (FL) and split learning (SL) are two effective distributed learning paradigms in wireless networks, enabling collaborative model training across mobile devices without sharing raw data. While FL supports low-latency parallel training, it may converge to less accurate model. In contrast, SL achieves higher accuracy through sequential training but suffers from increased delay. To leverage the advantages of both, hybrid split and federated learning (HSFL) allows some devices to operate in FL mode and others in SL mode. This paper aims to accelerate HSFL by addressing three key questions: 1) How does learning mode selection affect overall learning performance? 2) How does it interact with batch size? 3) How can these hyperparameters be jointly optimized alongside communication and computational resources to reduce overall learning delay? We first analyze convergence, revealing the interplay between learning mode and batch size. Next, we formulate a delay minimization problem and propose a two-stage solution: a block coordinate descent method for a relaxed problem to obtain a locally optimal solution, followed by a rounding algorithm to recover integer batch sizes with near-optimal performance. Experimental results demonstrate that our approach significantly accelerates convergence to the target accuracy compared to existing methods.

We investigate the role of channel access schemes in enhancing the timeliness of status updates in sensor networks. Specifically, we model the large-scale sensor network as a Poisson cellular network and derive the network average age of information (AoI) under five different channel access schemes: slotted ALOHA, frame slotted ALOHA, random scheduling, round robin, and channel-aware. These schemes are categorized based on random vs. scheduled access and non-channel-aware vs. channel-aware. Our goal is to investigate when the additional overhead and complexity introduced by scheduling and channel state information (CSI) are beneficial, enabling better decisions in network design. Our findings reveal that the effectiveness of these schemes is influenced by the signal-to-interference ratio (SIR) decoding threshold, which often reflects the length of communication data. For short-packet communications, the performance differences among various channel access strategies are minimal, and the gains from scheduling are limited. Additionally, the inclusion of extra CSI does not yield performance improvements; in fact, some simple scheduling strategies, along with channelaware strategy that leverage CSI, may not outperform basic random access methods. Among the protocols we examined, the round robin scheme achieves the best performance. In contrast, scheduled access schemes exhibit a clear performance advantage in long-packet communications. Furthermore, the channel-aware scheme significantly enhances the network AoI performance, particularly in networks with higher transmitter competition.

Large language models (LLMs) have demonstrated remarkable capabilities across a wide range of natural language processing tasks. Exploiting the heterogeneous capabilities of edge LLMs is crucial for diverse emerging applications, as it enables greater cost-effectiveness and reduced latency. In this work, we introduceMixture-of-Edge-Experts (MoE2), a novel collaborative inference framework for edge LLMs. We formulate a joint gating and expert selection problem to optimize inference performance under energy and latency constraints. Unlike conventional MoE problems, LLM expert selection becomes significantly more challenging due to the combinatorial nature and the heterogeneity of edge LLMs across various attributes. To this end, we propose a two-level expert selection mechanism through which we uncover an optimality-preserving property of gating parameters across expert selections. This property enables the decomposition of the training and selection processes, significantly reducing complexity. Furthermore, we leverage the objective’s monotonicity and design a discrete monotonic optimization algorithm for optimal expert selection. We implement edge servers with NVIDIA Jetson AGX Orins and NVIDIA RTX 4090 GPUs, and perform extensive experiments. Our results validate the performance improvements for various LLM models and show that our MoE2 method can achieve optimal trade-offs among different delay and energy budgets, and outperforms baselines under various system resource constraints. We further demonstrate its strong robustness in dynamic, non-stationary environments and its effectiveness in achieving load balancing.

This work considers the problem of remote estimation of Wiener processes and proposes a sample preprocessing method to ensure the convergence of the estimation distortion. Specifically, the autocorrelation of the Wiener process is exploited to counteract the effect of strong quantization noise arising from the linearly increasing variance over time. We first derive the exact expression for the convergent mean squared error (MSE) without considering transmission outages to explain the proposed preprocessing method. Then, the analysis is extended to more complex and general scenarios with outages. Based on the derived MSE, the quantization precision, the sampling interval, and the transmission time of a single piece of update information are optimized individually. We further give two algorithms to obtain two global suboptimal MSEs for practical cases considering low thresholds of quantization precision and sampling interval, following a demonstration of the unsolvability of the joint optimization. The numerical results reveal that dynamic distortion plays a greater role than static distortion due to the fast-varying nature of the Wiener process, which also verifies the effectiveness of the proposed preprocessing method in controlling the quantization error.

Federated learning facilitates collaborative model training across multiple clients while preserving data privacy. However, its performance is often constrained by limited communication resources, particularly in systems supporting a large number of clients. To address this challenge, integrating over-the-air computations into the training process has emerged as a promising solution to alleviate communication bottlenecks. The system significantly increases the number of clients it can support in each communication round by transmitting intermediate parameters via analog signals rather than digital ones. This improvement, however, comes at the cost of channel-induced distortions, such as fading and noise, which affect the aggregated global parameters. To elucidate these effects, this paper develops a theoretical framework to analyze the performance of over-the-air federated learning in large-scale client scenarios. Our analysis reveals three key advantages of scaling up the number of participating clients: (1) Enhanced Privacy: The mutual information between a client’s local gradient and the server’s aggregated gradient diminishes, effectively reducing privacy leakage. (2) Mitigation of Channel Fading: The channel hardening effect eliminates the impact of small-scale fading in the noisy global gradient. (3) Improved Convergence: Reduced thermal noise and gradient estimation errors benefit the convergence rate. These findings solidify over-the-air model training as a viable approach for federated learning in networks with a large number of clients. The theoretical insights are further substantiated through extensive experimental evaluations.

In this paper, we consider the problem of remote estimation of Wiener processes and propose a sample preprocessing method to ensure the convergence of estimation distortion. The autocorrelation of the Wiener process is exploited to counteract the effect of strong quantization noise arising from the linearly increasing variance over time. We first derive the convergent expression for the mean squared error without considering transmission outages to explain the proposed preprocessing method. Then, the analysis is extended to more complex and general scenarios with outages. Based on the analyses, the quantization precision and the sampling interval are optimized individually.

Large language models (LLMs) have demonstrated remarkable capabilities across various natural language processing tasks. Exploiting the heterogeneous capabilities of edge LLMs is crucial for emerging applications, enabling greater cost-effectiveness and reduced latency. In this work, we introduce Mixture-of-Edge-Experts (MoE2), a collaborative inference framework for edge LLMs. We formulate the joint gating and expert selection problem to optimize inference under energy and latency constraints. Unlike conventional MoE problems, expert selection here is more challenging due to the combinatorial nature and heterogeneity of edge LLMs. To address this, we propose a two-level expert selection mechanism and uncover an optimality-preserving property of gating parameters that decouples training and selection, reducing complexity. We further leverage the objective's monotonicity and design a discrete monotonic optimization algorithm. Implemented on Jetson Orin and RTX 4090 platforms, MoE2 achieves optimal trade-offs across delay and energy budgets, outperforming baselines under various resource constraints.

This paper presents a novel analysis of the directional RTS/CTS handshake mechanism using stochastic geometry, treating its impact as a thinning process in a spatial point process. We introduce unique thinning procedures based on the cosine antenna pattern, offering a more accurate representation of directional transmission in millimeter-wave networks. Through these models, we derive expressions for the intensity of concurrent transmitter processes and the mean interference. Our analysis provides key insights into how mean interference varies with system parameters such as antenna configuration and protection distance. Specifically, numerical results demonstrate that increasing the number of transmit antennas reduces interference, though diminishing returns are observed as the number of antennas grows.

Leveraging over-the-air computations for model aggregation is an effective approach to cope with the communication bottleneck in federated edge learning. By exploiting the superposition properties of multi-access channels, this approach facilitates an integrated design of communication and computation, thereby enhancing system privacy while reducing implementation costs. However, the inherent electromagnetic interference in radio channels often exhibits heavy-tailed distributions, giving rise to exceptionally strong noise in globally aggregated gradients that can significantly deteriorate the training performance. To address this issue, we propose a novel gradient clipping method, termed Median Anchored Clipping (MAC), to combat the detrimental effects of heavy-tailed noise. We also derive analytical expressions for the convergence rate of model training with analog over-the-air federated learning under MAC, which quantitatively demonstrates the effect of MAC on training performance. Extensive experimental results show that the proposed MAC algorithm effectively mitigates the impact of heavy-tailed noise, hence substantially enhancing system robustness.

Over-the-air federated learning (OTA FL) provides a joint computation and communication approach to design FL systems with improved efficiency. By leveraging the superposition property of wireless channels, OTA FL enables the automatic aggregation of intermediate parameters-such as gradients-across a large number of clients, significantly reducing communication overhead while concurrently enhancing transmission privacy. However, gradient aggregation requires all clients to use identical model architectures, a condition often impractical in real-world scenarios due to variations in client hardware and computational capabilities. This mismatch can lead to scalability issues and system incompatibilities. To address this challenge, we propose a model-agnostic method based on model prototypes that enables collaborative training across clients with heterogeneous models. The proposed method bypasses the requirements of the conventional model weight/gradient updates with prototype vector aggregation, without requiring the model structures of all clients to be identical. To the best of our knowledge, the proposed method is the first to explore the prototypical model-heterogeneous OTA FL with desirable training performance and extremely low communication cost. We conducted extensive experiments to verify the efficacy of the proposed method. The results show that our approach not only significantly reduces communication overhead but also exploits the superior capabilities of large models to enhance the performance of smaller models.

We investigate the role of channel access schemes in enhancing the timeliness of status updates in sensor networks. Specifically, we model the large-scale sensor network as a Poisson cellular network and derive the network average Age of Information (AoI) under two channel access schemes: random and scheduled access. Our findings reveal that the effectiveness of these schemes is influenced by the signal-to-interference ratio (SIR) decoding threshold, which often reflects the length of communication data. For short-packet communications, performance differences among various channel access strategies are minimal, and the gains from scheduling are limited; in fact, some simple scheduling strategies may not outperform basic random strategies. Conversely, scheduled access schemes demonstrate a distinct performance advantage for long-packet communications. The round robin scheme consistently yields the best performance among the four protocols we examined-slotted ALOHA, frame slotted ALOHA, random scheduling, and round robin scheduling. This is due to its ability to mitigate intra-cell interference and regularize both status updates and channel access periods for each sensor, which is particularly beneficial in reducing AoI.

Federated learning enables multiple clients to collaboratively train a common model while concurrently preserving data privacy. However, its performance is often constrained by limited communication resources, especially when the system encounters a large number of clients. Under those circumstances, integrating over-the-air computations into the model training procedure is considered an effective approach to coping with the communication bottleneck. Specifically, by uploading each client's intermediate parameters via analog transmissions instead of digital ones, the system can dramatically extend the number of clients it simultaneously supports in each communication round. However, that is achieved at the expense of introducing channel distortions, particularly fading and noise, in the aggregated global parameter. To demystify these effects, the present paper develops a theoretical framework to analyze the performance of the over-the-air federated model training process. Our analysis unveils a three-fold benefit from system scaling up, i.e., as the number of participating clients increases: (i) the privacy leakage, quantified by the mutual information between each client's locally possessed gradient and the edge's globally aggregated one, substantially decreases, (ii) the impairment of small-scale fading disappears due to the channel hardening effect, and (iii) the convergence rate is enhanced as thermal noise and gradient estimation error can be reduced. To that end, it establishes over-the-air model training as a viable approach for implementing federated learning in scenarios with a large number of clients. We corroborate the theoretical findings with extensive experiments.

Federated learning (FL) with synchronous model aggregation suffers from the straggler issue because of heterogeneous transmission and computation delays among different agents. In mobile wireless networks, this issue is exacerbated by time-varying network topology due to agent mobility. Although asynchronous FL can alleviate straggler issues, it still faces critical challenges in terms of algorithm design and convergence analysis because of dynamic information update delay (IU-Delay) and dynamic network topology. To tackle these challenges, we propose a decentralized FL framework based on gradient descent with momentum, named decentralized momentum federated learning (DMFL). We prove that DMFL is globally convergent on convex loss functions under the bounded time-varying IU-Delay, as long as the network topology is uniformly jointly strongly connected. Moreover, DMFL does not impose any restrictions on the data distribution over agents. Extensive experiments are conducted to verify DMFL’s performance superiority over the benchmarks and to reveal the effects of diverse parameters on the performance of the proposed algorithm.

The deep autoencoder (DAE) framework has turned out to be efficient in reducing the channel state information (CSI) feedback overhead in massive Multiple-Input Multiple-Output (mMIMO) systems. However, these DAE approaches presented in prior works rely heavily on large-scale data collected through the base station (BS) for model training, thus rendering excessive bandwidth usage and data privacy issues, particularly for mMIMO systems. When considering users’ mobility and encountering new channel environments, the existing CSI feedback models may often need to be retrained. Returning back to previous environments, however, will make these models perform poorly and face the risk of catastrophic forgetting. To solve the above challenging problems, we propose a novel gossiping generative adversarial network (Gossip-GAN)-aided CSI feedback training framework. Notably, Gossip-GAN enables the CSI feedback training with low-overhead while preserving users’ privacy. Specially, each user collects a small amount of data to train a GAN model. Meanwhile, a fully distributed gossip learning strategy is exploited to avoid model overfitting, and to accelerate the model training as well. Simulation results demonstrate that Gossip-GAN can: 1) achieve a similar CSI feedback accuracy as centralized training with real-world datasets; 2) address catastrophic forgetting challenges in mobile scenarios, and 3) greatly reduce the uplink bandwidth usage. Besides, our results show that the proposed approach possesses an inherent robustness.

Federated learning (FL) emerges as a potential solution for enabling multiple terminal devices to collaboratively accomplish computational tasks within an autonomous aerial vehicle (AAV) swarm. However, traditional FL approaches, predicated on synchronous data aggregation, are not feasible for a AAV swarm owing to the inherently variable and dynamic nature of their communication networks compared with terrestrial systems. Furthermore, the data procured by AAVs is often highly heterogeneous, attributable to disparities in deployment environments and device attributes. Considering the distinct flight paths and unique operational conditions encountered by different AAVs, a considerable amount of data remains unlabeled. To tackle the challenges associated with asynchronous operations and the prevalence of unlabeled data, we introduce a novel framework termed semi-synchronous FL with cross-sharpness and layer-freezing (SFedXL), tailored for a AAV swarm. In particular, we devise a cross-sharpness model training strategy aimed at optimizing the utilization of both labeled and unlabeled datasets. Additionally, we propose an innovative semi-synchronous model aggregation protocol, complemented by client-specific layer-freezing and client cluster scheduling, designed to expedite the training process. Our simulation results indicate that the proposed algorithm surpasses current FL methods in terms of object recognition accuracy and communication efficiency, albeit with a tradeoff of increased local computation latency.

We study the age of information (AoI) in a random access network consisting of multiple source-destination pairs, where each source node is empowered by energy harvesting capability. Every source node transmits a sequence of data packets to its destination using only the harvested energy. Each data packet is encoded with finite-length codewords, characterizing the nature of short codeword transmissions in random access networks. By combining tools from bulk-service Markov chains with stochastic geometry, we derive an analytical expression for the network average AoI and obtain closed-form results in two special cases, i.e., the small and large energy buffer size scenarios. Our analysis reveals the trade-off between energy accumulation time and transmission success probability. We then optimize the network average AoI by jointly adjusting the update rate and the blocklength of the data packet. Our findings indicate that the optimal update rate should be set to one in the energy-constrained regime where the energy consumption rate exceeds the energy arrival rate. This also means if the optimal blocklength of the data packet is pre-configured, an energy buffer size supporting only one transmission is sufficient.

This work studies the average Age of Information (AoI) of a remote monitoring setup in which a multi-sensor system observes independent sources and updates the status to a common monitor using orthogonal channels. Considering the limited buffer size at the sensors, we first model each sensor as a first-come-first-served M/M/1/1 queue. Leveraging tools from stochastic hybrid systems, we derive the average AoI of a homogeneous single-source multi-sensor system in which all sensors’ arrival and service rates are the same. We then extend the results to the multi-source, multi-sensor system. For a multi-source dual-sensor system, we present an approximate optimal arrival rate for a given sum arrival rate at a light load. For heterogeneous cases with different arrival and service rates at sensors, the average AoI is derived for the single-source dual-sensor and more general multi-source systems. Our analysis shows that the average AoI decreases by 16.44% and 21.44% for the dual-sensor and three-sensor systems, respectively, compared to the single-sensor system when the service rate and the total arrival rate of the sensors are normalized. Numerical results confirm that the average AoI performance of the single-source dual-sensor system outperforms the M/M/2 system at high system load.

Joint sensing and communication (JSC) systems aim to leverage the same spectral resources for both communication and sensing tasks within a single system. These systems have the potential to enhance sensing capabilities through advanced communication techniques, while also utilizing precise localization and tracking information from sensing technologies to improve communication. However, the integration of information obtained from sensing and transmitted in communication is not yet fully understood. This paper investigates the challenge of guaranteeing timely delivery of sensing information within JSC systems. We introduce a novel metric, termed as the age of estimation information (AoEI), which integrates radar mutual information (MI) and age of information (AoI). This unified metric effectively captures both the passage of time and the accuracy of estimation information, making it well-suited for the JSC system. Further, we delve into the joint optimization of time and power allocation for a single JSC node with both sensing and communication capabilities. Our objective is to minimize the long-term average AoEI while adhering to a long-term average power constraint. To tackle this problem, we formulate it as an average-reward constrained Markov decision process (CMDP) and propose a model-free constrained deep reinforcement learning (CDRL) algorithm, namely the average policy optimization (APO)-Lagrangian based algorithm. Simulation results demonstrate that our proposed algorithm effectively meets the constraint in dynamic and uncertain environments while achieving a favorable balance between AoEI and power consumption. Additionally, our algorithm outperforms four baseline schemes, showcasing its superior performance.

Federated Learning (FL) stands as a privacy-preserving machine learning paradigm that enables collaborative training of a global model across multiple clients. However, the practical implementation of FL models often confronts challenges arising from data heterogeneity and limited communication resources. To address the aforementioned issues simultaneously, we develop a Sparsified Random Partial Update framework for personalized Federated Learning (SRP-pFed), which builds upon the foundation of dynamic partial model updates. Specifically, we decouple the local model into personal and shared parts to achieve personalization. For each client, the ratio of its personal part associated with the local model, referred to as the update rate, is regularly renewed over the training procedure via a random walk process endowed with reinforced memory. In each global iteration, clients are clustered into different groups where the ones in the same group share a common update rate. Benefiting from such design,SRP-pFedrealizes model personalization while substantially reducing communication costs in the uplink transmissions. We conduct extensive experiments on various training tasks with diverse heterogeneous data settings. The results demonstrate that theSRP-pFedconsistently outperforms the state-of-the-art methods in test accuracy and communication efficiency.

We propose an effective approach toward implementing personalized federated learning at the edge of wireless networks. The scheme employs a bi-level optimization framework to personalize the federated learning models, and leverages over-the-air computations for model aggregation.We identify a mutual benefit in such a design. Specifically, personalized federated learning models address the challenge of data heterogeneity in federated learning, while over-the-air computations, which capitalize on the superposition property of multiple access channels, enable all clients to upload their intermediate parameters in each communication round for global aggregation, significantly enhancing system scalability. However, the channel fading and heavy-tailed noise introduced by over-the-air computations pose challenges to the robustness of personalized federated learning models. By adopting a bi-level optimization framework, we improve the stability of personalized federated learning models based on over-the-air computations, establishing a scalable and robust federated edge learning system. We also derive convergence rates of the proposed algorithm, encompassing key factors such as model compression, channel fading, and heavy-tailed noise. The analysis offers a comprehensive understanding of how system configurations affect training performance. We corroborate the efficacy of our framework via extensive experiments.

Federated learning (FL) and split learning (SL) are two prominent distributed learning modes. FL allows for parallel training but demands significant computational resources on devices to train deep neural network models. Conversely, SL reduces the computational burden on devices and can enhance learning performance, though it often leads to longer training time due to its sequential nature. In this paper, we introduce a novel distributed learning framework, hybrid split and federated learning (HSFL), which combines the advantages of both FL and SL over wireless networks. To achieve a lower training loss within a shorter latency, we start with the convergence analysis of HSFL, followed by a joint optimization problem of the learning mode selection, model splitting, and bandwidth allocation. To solve the problem, we propose a two-stage algorithm. First, we find the optimal bandwidth allocation and model splitting with a fixed learning mode. Then, we select the optimal learning mode based on the above optimal values. Experimental results validate the superior learning efficacy of our proposed algorithm.

Along with the advancement of WiFi technologies, commodity devices of WiFi 6E have now prevailed. In this paper, we conduct a wireless sensing task that utilizes WiFi 6E devices to monitor the respiration pattern of a human being. We discover new features, as well as their causes, in the channel state information (CSI) measured from WiFi 6E devices, which significantly differ from those in legacy IEEE 802.11n/ac/ax devices. As a result, directly applying conventional sensing techniques based on the CSI hardly yields satisfactory performance. In response, we propose a new method that effectively rectifies the measurement defects and recovers the targeted signal. We carry out extensive experiments, the results of which show that our approach substantially outperforms state-of-the-art benchmarks in sensing accuracy, verifying the effectiveness of our scheme.

Integrated sensing and communication (ISAC) has been proved as a promising technique to further improve the performance for both the communication and the sensing tasks in wireless networks. However, due to the complicated and dynamic interference circumstances, the performance of ISAC cannot be guaranteed. To provide some insights for keeping a sophisticated balance between communication and sensing with considering co-channel interference, the theoretical performance of ISAC is studied in this paper. First, an analytical system model is provided based on stochastic geometry. Second, the success serving probability (SSP) is defined for both the communication and the sensing tasks based on mutual information, which provided a unified analysis framework for ISAC. The tractable expressions of SSP are also derived. Finally, the simulation results are shown to verify the analytical results of SSP, which can provide some insights for the tradeoff between sensing and communication of ISAC.

Federated Long-Tailed Learning (Fed-LT), a paradigm wherein data collected from decentralized local clients manifests a globally prevalent long-tailed distribution, has garnered considerable attention in recent times. In the context of Fed-LT, existing works have predominantly centered on addressing the data imbalance issue to enhance the efficacy of the generic global model while neglecting the performance at the local level. In contrast, conventional Personalized Federated Learning (pFL) techniques are primarily devised to optimize personalized local models under the presumption of a balanced global data distribution. This paper introduces an approach termed Federated Local and Generic Model Training in Fed-LT (FedLoGe), which enhances both local and generic model performance through the integration of representation learning and classifier alignment within a neural collapse framework. Our investigation reveals the feasibility of employing a shared backbone as a foundational framework for capturing overarching global trends, while concurrently employing individualized classifiers to encapsulate distinct refinements stemming from each client’s local features. Building upon this discovery, we establish the Static Sparse Equiangular Tight Frame Classifier (SSE-C), inspired by neural collapse principles that naturally prune extraneous noisy features and foster the acquisition of potent data representations. Furthermore, leveraging insights from imbalance neural collapse's classifier norm patterns, we develop Global and Local Adaptive Feature Realignment (GLA-FR) via an auxiliary global classifier and personalized Euclidean norm transfer to align global features with client preferences. Extensive experimental results on CIFAR-10/100-LT, ImageNet, and iNaturalist demonstrate the advantage of our method over state-of-the-art pFL and Fed-LT approaches.

The proliferation of real-time status updating applications and ubiquitous mobile devices have motivated the analysis and optimization of data freshness in the context of age of information. At the same time, increasing requirements on computer performance have inspired research on CPU scheduling, with a focus on reducing energy consumption. However, since prior CPU scheduling strategies have ignored data freshness, we formulate the first CPU scheduling problem that aims to minimize the long-term average age of information, subject to an average power constraint. In particular, we optimize CPU scheduling strategies that specify when the CPU sleeps and adapt the CPU speed (clock frequency) during the execution of update-processing tasks. We formulate the age-minimal CPU scheduling problem as a constrained semi-Markov decision process (SMDP) problem with uncountable space. We develop a value-iteration-based algorithm and further prove its convergence in infinite space to obtain the optimal policy. Compared with existing benchmarks in terms of long-term average AoI, numerical results show that our proposed scheme can reduce the AoI by up to 53%, and obtains greater benefits when faced with a tighter power constraint. In addition, for a given AoI target, the age-minimal CPU scheduling policy can save more than 50% on energy consumption.

Federated Learning (FL) is envisioned as the cornerstone of the next-generation mobile system, whereby integrating FL into the network edge elements (i.e., user terminals and edge/cloud servers), it is expected to unleash the potential of network intelligence by learning from the massive amount of users' data while concurrently preserving privacy. In this paper, we develop an analytical framework that quantifies the interplay of user mobility, a fundamental property of mobile networks, and data heterogeneity, the salient feature of FL, on the model training efficiency. Specifically, we derive the convergence rate of a hierarchical FL system operated in a mobile network, showing how user mobility exacerbates the divergence caused by data heterogeneity. The theoretical findings are corroborated by experimental simulations.

We optimize the Age of Information (AoI) in random access networks using the age-threshold slotted ALOHA (TSA) protocol. The network comprises multiple source-destination pairs, where each source sends a sequence of status update packets to its destination over a shared spectrum. The TSA protocol stipulates that a source node must remain silent until its AoI reaches a predefined threshold, after which the node accesses the radio channel with a certain probability. We derive analytical expressions for the transmission success probability and time-average AoI using stochastic geometry tools. Subsequently, we obtain closed-form expressions for the optimal update rate and age threshold that minimize the time-average AoI. In addition, we establish a scaling law for the time-average AoI in random access networks, revealing that the optimal time-average AoI increases linearly with the deployment density. Notably, the growth rate under TSA is half of that under conventional slotted ALOHA.

Age of information (AoI), defined for an information source as the time elapsed since the latest received update was generated, is a recently proposed metric that quantifies the timeliness of information delivery in a communication system. This paper studies a fundamental problem of how the achievable AoI scales in mobile networks. Specifically, we consider a network consisting of n/2 source-destination (S-D) pairs and employ the protocol model to characterize interference incurred by concurrent transmissions. We consider a general class of scheduling policies potentially with the multi-hop transmission and the duplication of packets to multiple nodes. The analysis of AoI faces significant challenges due to potential out-of-order packet delivery, the inherent tradeoffs between packet-centric metrics (throughput and delay), and their unexplored relation to AoI. We first show that the average per-node AoI in static settings scales as [EQUATION]. In the case of networks with i.i.d. mobility, where the node locations vary independently over time, we introduce an episodic technique that allows us to establish lower bounds and design and analyze scheduling policies as constructive upper bounds. Our analytical results reveal that the average per-node AoI scales as [EQUATION] under i.i.d. mobility, which highlights that mobility can enhance timeliness. Finally, we show that, in a more general class of wireless network settings, one can design the age-minimal scheduling policy by balancing throughput and delay.

Federated learning (FL) is a privacy-preserving distributed learning framework where model training is performed locally on distributed devices. Unlike traditional FL, which assumes a fixed local dataset, this paper focuses on the more realistic scenario of FL with streaming data. In Streaming Federated Learning (SFL), new data continuously arrives over time, and due to the limited local storage capacity on devices, some data is inevitably discarded. The discarded data may be forgotten by the model, leading to a decline in model accuracy. To this end, we introduce Federated Data Slimming (FedDS), a data selection scheme designed to determine which data should be stored locally. Particularly, FedDS considers both gradient norms and directions when making data selections. We evaluate the performance of FedDS against several previously proposed schemes using various datasets. Our experimental results demonstrate that FedDS surpasses all baseline schemes, achieving the fastest convergence rate and the highest test accuracy.

We leverage the Multiplicative Weight Update (MWU) method to develop a decentralized algorithm that significantly improves the performance of dynamic time division duplexing (D-TDD) in small cell networks. The proposed algorithm adaptively adjusts the time portion allocated to uplink (UL) and downlink (DL) transmissions at every node during each scheduled time slot, aligning the packet transmissions toward the most appropriate link directions according to the feedback of signal-to-interference ratio information. Our simulation results reveal that compared to the (conventional) fixed configuration of UL/DL transmission probabilities in D-TDD, incorporating MWU into D-TDD brings about a two-fold improvement of mean packet throughput in the DL and a three-fold improvement of the same performance metric in the UL, resulting in the D-TDD even outperforming Static-TDD in the UL. It also shows that the proposed scheme maintains a consistent performance gain in the presence of an ascending traffic load, validating its effectiveness in boosting the network performance. This work also demonstrates an approach that accounts for algorithmic considerations at the forefront when solving stochastic problems.

Timely status updating is the premise of emerging interaction-based applications in the Internet of Things (IoT). Using redundant devices to update the status of interest is a promising method to improve the timeliness of information. However, parallel status updating leads to out-of-order arrivals at the monitor, significantly challenging timeliness analysis. This work studies the Age of Information (AoI) of a multi-queue status update system where multiple devices monitor the same physical process. Specifically, two systems are considered: theBasic System, which only has type-1 devices that are ad hoc devices located close to the source, and theHybrid System, which contains additional type-2 devices that are infrastructure-based devices located in fixed points compared to theBasic System. Using the Stochastic Hybrid Systems (SHS) framework, a mathematical model that combines discrete and continuous dynamics, we derive the expressions of the average AoI of the considered two systems in closed form. Numerical results verify the accuracy of the analysis. It is shown that when the number and parameters of the type-1 devices/type-2 devices are fixed, the logarithm of average AoI will linearly decrease with the logarithm of the total arrival rate of type-2 devices or that of the number of type-1 devices under specific condition. It has also been demonstrated that the proposed systems can significantly outperform the FCFS M/M/Nstatus update system.

Many real-time applications of the Internet of Things (IoT) need to deal with correlated information generated by multiple sensors. The design of efficient status update strategies that minimize the Age of Correlated Information (AoCI) is a key factor. In this paper, we consider an IoT network consisting of sensors equipped with the energy harvesting (EH) capability. We optimize the average AoCI at the data fusion center (DFC) by appropriately managing the energy harvested by sensors, whose true battery states are unobservable during the decision-making process. Particularly, we first formulate the dynamic status update procedure as a partially observable Markov decision process (POMDP), where the environmental dynamics are unknown to the DFC. In order to address the challenges arising from the causality of energy usage, unknown environmental dynamics, unobservability of sensors' true battery states, and large-scale discrete action space, we devise a deep reinforcement learning (DRL)-based dynamic status update algorithm. The algorithm leverages the advantages of the soft actor-critic and long short-term memory techniques. Meanwhile, it incorporates our proposed action decomposition and mapping mechanism. Extensive simulations are conducted to validate the effectiveness of our proposed algorithm by comparing it with available DRL algorithms for POMDPs.

We optimize the Age of Information (AoI) in mobile networks using the age-threshold slotted ALOHA (TSA) protocol. The network comprises multiple source-destination pairs, where each source sends a sequence of status update packets to its destination over a shared spectrum. The TSA protocol stipulates that a source node must remain silent until its AoI reaches a predefined threshold, after which the node accesses the radio channel with a certain probability. Using stochastic geometry tools, we derive analytical expressions for the transmission success probability, mean peak AoI, and time-average AoI. Subsequently, we obtain closed-form expressions for the optimal update rate and age threshold that minimize the mean peak and time-average AoI, respectively. In addition, we establish a scaling law for the mean peak AoI and time-average AoI in mobile networks, revealing that the optimal mean peak AoI and time-average AoI increase linearly with the deployment density. Notably, the growth rate of time-average AoI under TSA is half of that under SA. When considering the optimal mean peak AoI, the TSA protocol exhibits comparable performance to the traditional slotted ALOHA protocol. These findings conclusively affirm the advantage of TSA in reducing higher-order AoI, particularly in densely deployed networks.

This paper focuses on the timely delivery of sensing information in a joint sensing and communication (JSC) system to meet the requirements of emerging applications. Specifically, we investigate the time allocation of a single JSC node equipped with both sensing and communication functions to minimize the long-term average age of estimation information (AoEI) while satisfying the long-term average power constraint. The proposed metric, AoEI, combines radar mutual information (MI) and age of information (AoI) to capture both the passage of time and the accuracy of estimation information, making it more suitable for the JSC system. To solve this problem, we formulate the time allocation problem as a constrained Markov decision process (CMDP) and propose a model-free constrained deep reinforcement learning (CDRL) based algorithm. The simulation results demonstrate that the proposed algorithm can achieve a good trade-off between AoEI and power consumption and converge to a policy that satisfies the constraint in a highly dynamic and uncertain environment.

Federated edge learning is a promising technology to deploy intelligence at the edge of wireless networks in a privacy-preserving manner. Under such a setting, multiple clients collaboratively train a global generic model under the coordination of an edge server. But the training efficiency is often hindered by challenges arising from limited communication and data heterogeneity. In this paper, we present a distributed training paradigm that employs analog over-the-air computation to alleviate the communication bottleneck. Additionally, we leverage a bi-level optimization framework to personalize the federated learning model so as to cope with the data heterogeneity issue. As a result, it enhances the generalization and robustness of each client’s local model. We elaborate on the model training procedure and its advantages over conventional frameworks. We provide a convergence analysis that theoretically demonstrates the training efficiency. We also conduct extensive experiments to validate the efficacy of the proposed framework.

This paper proposes a client selection (CS) method to tackle the communication bottleneck of federated learning (FL) while concurrently coping with FL’s data heterogeneity issue. Specifically, we first analyze the effect of CS in FL and show that FL training can be accelerated by adequately choosing participants to diversify the training dataset in each round of training. Based on this, we lever-age data profiling and determinantal point process (DPP) sampling techniques to develop an algorithm termed Federated Learning with DPP-based Participant Selection (FL-DP3S). This algorithm effectively diversifies the participants’ datasets in each round of training while preserving their data privacy. We conduct extensive experiments to examine the efficacy of our proposed method. The results show that our scheme attains a faster convergence rate, as well as a smaller communication overhead than several baselines.

Emerging interaction-based Internet of Things (IoT) applications have stringent demand for timeliness, imposing critical challenges to the design of status update system. Using redundant devices to update the status of the same process is a promising way to improve timeliness, but this approach can result in out of order update arrivals, making it difficult to analyze timeliness. To that end, the present paper conducts a theoretical study toward the Age of Information (AoI) of a multi-queue status update system where multiple sensors observe one physical process and update a common monitor. Based on the stochastic hybrid systems method, the average AoI of the considered system is derived in closed form. The theoretical results are consistent with the simulation results, verifying the correctness of the theoretical analysis. It is shown that the logarithm of the average AoI is linearly decreasing with the logarithm of the number of sensors.

We propose a frame slotted ALOHA (FSA)-based protocol for source nodes to update status information toward their intended destinations in a random access network. We evaluate the effect of such a protocol on the network’s timeliness performance using the Age of Information (AoI) metric. Specifically, we leverage tools from stochastic geometry to model the geographical positions of the source-destination pairs and capture the entanglement amongst the nodes’ spatial-temporal attributes through the interference they caused to each other. We derive closed-form expressions for the average AoI over a typical transmission link. Our analysis shows that in densely deployed networks, the FSA-based status updating protocol can significantly decrease the average AoI. Furthermore, under the same updating frequency, converting a slotted ALOHA protocol into an FSA-based one always leads to a reduction in the average AoI.

Personalized federated learning (PFL) has been widely investigated to address the challenge of data heterogeneity, especially when a single generic model is inadequate in satisfying the diverse performance requirements of local clients simultaneously. Existing PFL methods are inherently based on the idea that the relations between the generic global and personalized local models are captured by the similarity of model weights. Such a similarity is primarily based on either partitioning the model architecture into generic versus personalized components or modeling client relationships via model weights. To better capture similar (yet distinct) generic versus personalized model representations, we propose $\textit{spectral distillation}$, a novel distillation method based on model spectrum information. Building upon spectral distillation, we also introduce a co-distillation framework that establishes a two-way bridge between generic and personalized model training. Moreover, to utilize the local idle time in conventional PFL, we propose a wait-free local training protocol. Through extensive experiments on multiple datasets over diverse heterogeneous data settings, we demonstrate the outperformance and efficacy of our proposed spectral co-distillation method, as well as our wait-free training protocol.

Data privacy and long-tailed distribution are the norms rather than the exception in many real-world tasks. This paper investigates a federated long-tailed learning (Fed-LT) task in which each client holds a locally heterogeneous dataset; if the datasets can be globally aggregated, they jointly exhibit a long-tailed distribution. Under such a setting, existing federated optimization and/or centralized long-tailed learning methods hardly apply due to challenges in (a) characterizing the global long-tailed distribution under privacy constraints and (b) adjusting the local learning strategy to cope with the head-tail imbalance. In response, we propose a method termed $\texttt{Fed-GraB}$, comprised of a Self-adjusting Gradient Balancer (SGB) module that re-weights clients' gradients in a closed-loop manner, based on the feedback of global long-tailed distribution evaluated by a Direct Prior Analyzer (DPA) module. Using $\texttt{Fed-GraB}$, clients can effectively alleviate the distribution drift caused by data heterogeneity during the model training process and obtain a global model with better performance on the minority classes while maintaining the performance of the majority classes. Extensive experiments demonstrate that $\texttt{Fed-GraB}$ achieves state-of-the-art performance on representative datasets such as CIFAR-10-LT, CIFAR-100-LT, ImageNet-LT, and iNaturalist.

Recommendation algorithm can effectively reduce the difficulty of proactive edge caching prediction by excavating users’ preferences among the massive contents, which has drawn great attentions from both academia and industry. The effectiveness of prediction models depends on big data analysis of user information, however, traditional methods based on centralized learning become more and more impractical due to the growing concern on privacy data protection. Recently, implementing the recommendation algorithm in a federated learning (FL) manner has emerged as a promising approach. In an FL manner, users are allowed to keep their private data local and upload the model parameters learned by local training to the server for collaborative training. In this work, we propose a proactive caching prediction algorithm for mobile vehicle users based on differential privacy and federate learning. Our proposed algorithm not only predicts the popular contents with a strong protection for users’ private data, but also applies to large-scale networks with massive mobile users. In addition, we also investigate the impact of user mobility on the caching prediction accuracy, and propose an attention-based model aggregation mechanism, which assigns different aggregation weights to each vehicle user and edge server to mitigate the performance degradation caused by user movement. The results show that our proposed model can obtain high caching prediction accuracy and strong privacy protection level in vehicular networks.

We develop a theoretical framework to study the training of federated learning models with partial gradients via over-the-air computing. The system consists of an edge server and multiple clients, aiming to collaboratively minimize a global loss function. The clients conduct local training and upload the intermediate parameters (e.g. the gradients) by analog transmissions. Specifically, each client modulates the entries of its local gradient onto a set of common orthogonal waveforms and sends out the signal simultaneously to the edge server; owing to the limited number of orthogonal waveforms, only a subset of the parameters can be selected for uploading during each round of communication. On the server side, it passes the received analog signal to a bank of match filters and obtains a noisy partial gradient vector. The server then uses this partial gradient to update the global parameter and feeds the new model back to all the clients for another round of local training. We derive the convergence rate of such a model training algorithm. We also conduct experiments to investigate the effects of different masking schemes on the convergence performance. The findings advance the understanding of over-the-air federated learning and provide useful insights for system designs.

This work studies the average Age of Information (AoI) of a remote monitoring system in which two sensors observe the same physical process and update the status to a common monitor using orthogonal channels. While using redundant devices to update the status of a process can improve the information timeliness at the monitor, the out-of-order arrivals of updates impose a challenge to the AoI analysis. We first model the system as two parallel M/M/1/1 queues. By leveraging tools from stochastic hybrid systems, we obtain analytically the average AoI of the system. In particular, when the arrival or service rates are the same for the two sensors, the average AoI is given in closed form. Our analysis reveals that the average AoI of the considered system is reduced by 16.44% compared to the single-sensor system when the arrival and service rates are equal to 1. Numerical results show that the considered system outperforms the M/M/2 system in average AoI at high arrival rates.

We investigate the performance of a random access network consisting of source-destination dipoles. The source nodes transmit information packets to their destinations over a shared spectrum. All the transmitters in this network adhere to an age threshold slotted ALOHA (TSA) protocol: every source node remains silent until the age of information (AoI) reaches a threshold, after which the source accesses the radio channel with a certain probability. We derive a tight approximation for the signal-to-interference-plus-noise ratio (SINR) meta distribution and verify its accuracy through simulations. We also obtain analytical expressions for the average AoI. Our analysis reveals that when the network is densely deployed, employing TSA significantly decreases the average AoI. The update rate and age threshold must be jointly optimized to fully exploit the potential of the TSA protocol.

We consider a random access network consisting of source-destination pairs. Each source node generates status updates and transmits this information to its intended destination over a shared spectrum. The goal is to minimize the network-wide Age of Information (AoI). We develop a frame slotted ALOHA (FSA)-based policy for generating and transmitting status updates, where the frame size of each source node is adjusted according to its local environment. The proposed policy is of low complexity and can be implemented in a distributed manner. Additionally, it significantly improves the network AoI performance by (a) equalizing the update generation intervals at each source and (b) reducing interference across the network. Furthermore, we derive an analytical expression for the average network AoI attained for that policy. We evaluate the performance of the proposed scheme through simulations, which demonstrate that the locally adaptive FSA policy achieves a remarkable gain in terms of AoI compared to the slotted ALOHA counterpart, confirming the effectiveness of the proposed method.

Grant-free random access (GFRA) becomes attractive in Internet of Things (IoT) due to its low signaling overhead and short access latency. In this article, we investigate GFRA in a multicel massive multiple-input-multiple-output (MIMO) system. As the IoT device usually has sporadic traffic, we set a packet buffer for each device to describe its temporal traffic, and also use the stochastic geometry to describe the randomness of devices’ spatial locations. With the backoff mechanism, only devices with a nonempty buffer and a successful backoff are activated and allowed to request access. Unlike previous works that regard all devices selecting the same pilot (i.e., the colliding devices) as undetectable, we give a more accurate model that the base station (BS) can detect colliding devices when they locate far away from each other, and we set a unique collision area for each device to quantify the boundary that the collision can be ignored. A tight approximation for the number of packets successfully transmitted at the unit area and time slot, named packet throughput, is derived. Based on it, we obtain the optimal backoff parameter that maximizes the packet throughput under devices’ delay constraints. It is shown that when pilots are insufficient or device packet traffic is heavy, a long backoff time is needed. However, as the pilot grows or the packet traffic turns light, devices should gradually reduce the backoff time. In particular, if pilots are surplus, cheap detectors can be equipped on the BS without an obvious packet throughput reduction.

We propose a frame slotted ALOHA (FSA)-based protocol for a random access network where sources transmit status updates to their intended destinations. We evaluate the effect of such a protocol on the network’s timeliness performance using the Age of Information (AoI) metric. Specifically, we leverage tools from stochastic geometry to model the spatial positions of the source-destination pairs and capture the entanglement amongst the nodes’ spatial-temporal attributes through the interference they caused to each other. We derive analytical expressions for the average and variance of AoI over a typical transmission link in Poisson bipolar and cellular networks, respectively. Our analysis shows that in densely deployed networks, the FSA-based status updating protocol can significantly decrease the average AoI and in addition, stabilizes the age performance by substantially reducing the variance of AoI. Furthermore, under the same updating frequency, converting a slotted ALOHA protocol into an FSA-based one always leads to a reduction in the average AoI. Moreover, implementing FSA in conjunction with power control can further benefit the AoI performance, although the particular values of framesize and power control factor must be adequately tuned to achieve the optimal gain.

Federated learning is a new artificial intelligence technology with which an edge server can orchestrate with multiple end users to train a global model collaboratively. Under this setting, users only upload the locally trained parameters instead of their local data, substantially reducing communication costs and boosting data privacy. Nonetheless, federated learning mainly relies on users’ local training, overlooking the abundant computing resources owned by the edge server. To exploit the edge server’s processing power, we propose a hybrid learning paradigm that consists of centralized and federated learning components. This scheme uploads a portion of users’ data for centralized learning when the local model is trained under federated learning. We derive a theoretical upper bound for the model accuracy, which can be used to assess the performance of the proposed new learning paradigm. To balance the computation and communication resources for a good model accuracy performance, we establish a joint optimization problem of model accuracy, latency, and energy consumption. We also devise the corresponding joint optimization algorithm to solve the problem. Experiment results show that compared with centralized and federated learning, the proposed hybrid learning algorithm can effectively improve the model accuracy and significantly reduce computation and communication resources.

The boom in Internet of Things has spawned many real-time applications, which have stringent requirements for the timeliness of information delivery. As a result, age of information (AoI) has emerged as a metric to evaluate information freshness at the destination and aroused widespread attention from both academia and industry. In this paper, we develop a theoretical framework to evaluate the statistics of AoI, including its average and violation probability, in wireless networks under different types of sources and updating patterns. The analyses account for the randomness that arises from both the spatial deployment and temporal queueing dynamics, and its accuracy is verified through simulations. Based on the analytical results, we design a locally adaptive power control policy to optimize the sum of average AoI of all nodes, which allows each node to assign transmit power according to its local observation. The proposed scheme has low implementation complexity. Numerical results show that the proposed power control policy can significantly improve information freshness. The scheme is well adapted to variants of network environment and heterogeneous source-destination distance. Further, we evaluate the effect of the retransmission mechanism and updating patterns on the AoI performance.

We consider a homogeneous Poisson bipolar network in which the bipoles represent source-destination pairs. The source nodes need to update their destinations about the new status perpetually, and the communications are taken place over a shared spectrum. The common goal of the source nodes is to minimize the network-wide age of information (AoI). We develop a policy by which every source node can adapt its frequency of generating status updates in a local and decentralized manner. At the same time, the network average AoI is minimized by reducing interference amongst transmitters located in geographical proximity. Following this policy, we also derive mathematical expressions to characterize the distribution of the optimal updating rate at each source node, the network average AoI, and the AoI violation probability, i.e., the probability that the AoI of a typical source node exceeds an age threshold. The analytical results are combined with discrete event simulations to provide a detailed evaluation of the performance of the proposed scheme. Particularly, it is shown that our policy is able to adaptively adjust the updating rate of each source node according to the variant of the network topology. In this manner, it is instrumental in decreasing both the network average AoI and AoI violation probability. Additionally, the scheme can maintain the AoI at a low level even when the network grows in size.

We study the average Age of Information (AoI) and peak AoI (PAoI) of a dual-queue status update system that monitors a common stochastic process through two independent channels. Although the double queue parallel transmission is instrumental in reducing AoI, the out of order of data arrivals also imposes a significant challenge to the performance analysis. We consider two settings: the M-M system where the service time of two servers is exponentially distributed; the M-D system in which the service time of one server is exponentially distributed and that of the other is deterministic. For the two dual-queue systems, closed-form expressions of average AoI and PAoI are derived by resorting to the graphic method and state flow graph analysis method. Our analysis reveals that when the two servers have the same service rate, compared with the single-queue system with an exponentially distributed service time, the average PAoI and the average AoI of the M-M system decrease by 33.3% and 37.5%, respectively, and those of the M-D system decrease by 27.7% and 39.7%, respectively. Numerical results show that the two dual-queue systems also outperform the M/M/2 single queue dual-server system with optimized arrival rate in terms of average AoI and PAoI.

Caching at the wireless edge is recognized as a promising solution to accommodate the explosive growth of traffic demand. However, the gain of edge caching is only pronounced given homogeneous user preference. To reap the full potential of caching, recommendation mechanism has emerged as an attractive technology due to its capability of reshaping users’ request distribution. In this work, we propose a joint user-side recommendation and device-to-device (D2D)-assisted offloading strategy, aiming to maximize the operator’s utility. Specifically, we consider that users can recommend their cached contents to encountered users. This strategy takes into account users’ personalized preferences and relative locations, and hence can directly offload the recommended contents through D2D links without burdening cellular links. We then develop a theoretical framework to evaluate the subsequent content transmission, accounting for the randomness of spatial deployment, user mobility, individual delay requirement, incentive, and protection mechanism for existing links. Based on the analytical results, we design a D2D-assisted offloading strategy, which allows the requester to postpone data reception in exchange for discounted service fees. Simulation results show that the operator’s utility can be significantly improved. Particularly, it is found that user mobility facilitates the above process.

Personalized Federated Learning (PFL) is a new Federated Learning (FL) approach to address the heterogeneity issue of the datasets generated by distributed user equipments (UEs). However, most existing PFL implementations rely on synchronous training to ensure good convergence performances, which may lead to a serious straggler problem, where the training time is heavily prolonged by the slowest UE. To address this issue, we propose a semi-synchronous PFL algorithm, termed as Semi-Synchronous Personalized FederatedAveraging (PerFedS2), over mobile edge networks. By jointly optimizing the wireless bandwidth allocation and UE scheduling policy, it not only mitigates the straggler problem but also provides convergent training loss guarantees. We derive an upper bound of the convergence rate of PerFedS2 in terms of the number of participants per global round and the number of rounds. On this basis, the bandwidth allocation problem can be solved using analytical solutions and the UE scheduling policy can be obtained by a greedy algorithm. Experimental results verify the effectiveness of PerFedS2 in saving the training time as well as guaranteeing the convergence of training loss, in contrast to synchronous and asynchronous PFL algorithms.

Personalized Federated Learning (PFL) is a new Federated Learning (FL) paradigm, particularly tackling the heterogeneity issues brought by various mobile user equipments (UEs) in mobile edge computing (MEC) networks. However, due to the ever-increasing number of UEs and the complicated administrative work it brings, it is desirable to switch the PFL algorithm from its conventional two-layer framework to a multiple-layer one. In this paper, we propose hierarchical PFL (HPFL), an algorithm for deploying PFL over massive MEC networks. The UEs in HPFL are divided into multiple clusters, and the UEs in each cluster forward their local updates to the edge server (ES) synchronously for edge model aggregation, while the ESs forward their edge models to the cloud server semi-asynchronously for global model aggregation. The above training manner leads to a tradeoff between the training loss in each round and the round latency. HPFL combines the objectives of training loss minimization and round latency minimization while jointly determining the optimal bandwidth allocation as well as the ES scheduling policy in the hierarchical learning framework. Extensive experiments verify that HPFL not only guarantees convergence in hierarchical aggregation frameworks but also has advantages in round training loss maximization and round latency minimization.

We study the average age of information (AoI) and peak AoI (PAoI) of a dual-queue status update system that monitors a common stochastic process. We capture the state transition characteristics of the considered system by establishing a Markov chain. Using the state flow graph analysis method, we derive closed-form expressions of the average peak age of information (PAoI) and the average age of information (AoI) for the dual-queue update system. The numerical results show that compared with the single-queue update system, the average PAoI of the dual-queue update system is reduced by 33.5% and the average AoI dropped by 37.5%.

Grant-free random access (GFRA) becomes attractive in Internet-of-Things (IoT) due to its low signaling overhead. In this paper, we investigate the GFRA in a multicell massive multiple-input multiple-output (MIMO) system after considering both the spatial and temporal traffic of devices. By introducing the backoff mechanism, only devices with non-empty buffer and a successful backoff can request access. Unlike previous works on GFRA that regard all devices selecting the same pilot as undetectable, we set a unique collision area for each device to quantify the boundary that BS can detect the collision. With tools of stochastic geometry and queueing theory, we derive a tight approximation for the number of packets successfully transmitted at unit area and time slot, named as packet throughput (Tp). Based on it, we find that the range of the collision area has a remarkable effect on Tp. The optimal backoff parameter that maximizes Tpis also obtained, and we find that a long backoff time is needed when the pilot is insufficient or the packet traffic is heavy. Compared with the fully-loaded access, our optimal backoff mechanism can significantly improve Tp, especially for the system with crowded devices.

Federated learning (FL) is an emerging machine learning method that can be applied in mobile edge systems, in which a server and a host of clients collaboratively train a statistical model utilizing the data and computation resources of the clients without directly exposing their privacy-sensitive data. We show that running stochastic gradient descent (SGD) in such a setting can be viewed as adding a momentum-like term to the global aggregation process. Based on this finding, we further analyze the convergence rate of a federated learning system by accounting for the effects of parameter staleness and communication resources. These results advance the understanding of the Federated SGD algorithm, and also forges a link between staleness analysis and federated computing systems, which can be useful for systems designers.

Hierarchical federated edge learning (H-FEEL) has been recently proposed to enhance the federated learning model. Such a system generally consists of three entities, i.e., the server, helpers, and clients. Each helper collects the trained gradients from users nearby, aggregates them, and sends the result to the server for model update. Due to limited communication resources, only a portion of helpers can upload their aggregated gradients to the server, thereby necessitating a well design for helper scheduling and communication resources allocation. In this paper, we develop a training algorithm for H-FEEL which involves local gradient computing, weighted gradient uploading, and model updating phases. By characterizing these phases mathematically and analyzing the one-round convergence bound of the training algorithm, we formulate a problem to achieve the scheduling and resource allocation scheme. To solve the problem, we first transform it into an equivalent problem and then decompose the transformed problem into two subproblems: bit and sub-channel allocation problem and helper scheduling problem. For the first subproblem, we obtain a low-complexity suboptimal solution by using a four-stage method. For the second subproblem, we obtain a stationary point by using the penalty convex-concave procedure. The efficacy of our scheme is demonstrated via simulations, and the analytical framework is shown to provide valuable insights for the design of practical H-FEEL system.

Opening the unlicensed bands provides additional spectrum resources for the next generation wireless network, while severe unfairness and performance degradation occur when one coexists with the incumbent users of these bands. Therefore, plenty of efforts have been made towards fair coexistence, mainly focusing on parameter tuning of listen-before-talk (LBT) and duty-cycle (DC) mechanisms. For better utilization of the unlicensed bands, it is of paramount importance to establish an access mechanism that guarantees the fairness objective among feasible mechanisms. Such access mechanism and the corresponding benchmark, nevertheless, remain largely unknown. To address this issue, this paper considers the coexistence between WiFi and the other unlicensed nodes, and aims to maximize the α-fairness between them. A benchmark is first given by solving the optimization problem. Then we propose a deep reinforcement learning (DRL) mechanism to help the unlicensed nodes make access decisions, such that they coexist with WiFi harmoniously. Extensive simulations have been carried out, and the results show that the DRL mechanism can approach the benchmark.

Analog over-the-air computing enables a swarm of end-user devices to efficiently conduct distributed learning, where the intermediate parameters of users, such as gradients, are modulated and transmitted via a group of orthogonal waveforms, and can be mixed directly at a server without individually detecting the feedback parameters of each user. Nonetheless, the scarcity of orthogonal waveforms, as well as communication resources of the end-user devices, are throttling this paradigm in adopting complex deep learning models. To balance the tradeoff between communication efficiency and accuracy performance, we study model pruning for analog over-the-air distributed learning in this paper. First, a model pruning scheme is proposed to improve the communication efficiency of analog over-the-air training. An importance measure for model parameter pruning is also designed based on the analog over-the-air aggregated gradient, which can characterize the contribution of each parameter without removing channel fading and electromagnetic interference. Second, an analytical expression of the training error upper bound is derived, which shows the proposed scheme is able to converge even when the aggregated gradient is corrupted by heavy-tailed electromagnetic interference with an infinite variance. Finally, several experimental results are provided to show the performance gains achieved by our proposed scheme, and also verify the correctness of analytical results.

The boom in Internet of Things (IoT) has spawned many real-time applications, which have stringent requirements for the timeliness of information delivery. As a result, age of information (AoI) has emerged as a metric to evaluate information freshness at the destination node and aroused widespread attention from both academia and industry. In this paper, we develop a locally adaptive power control policy for wireless ad hoc networks, which adjusts each node’s transmit power according to its local observation so as to optimize the sum of average AoI of all destination nodes. The proposed scheme has a low implementation complexity. Numerical results show that the proposed scheme is well adapted to variants of network environment and can significantly improve information freshness.

With the deluge of wireless data, unmanned aerial vehicles (UAVs) are expected to be deployed as aerial small base stations (SBSs) to relieve the load of ground macro base stations by establishing wireless backhaul connections with them and providing high-quality service to users. Thanks to the emergence of integrated access and backhaul (IAB), the access and backhaul communication links can work on the same millimeter wave (mmWave) band with huge available bandwidth. This paper studies the quality-of-service (QoS) performance of heterogeneous traffic under equal partition and average load partition spectrum allocation strategies for mmWave UAV-assisted IAB cellular networks. Specifically, we develop a theoretical framework to analyze the mean packet throughput (MPT) of users based on stochastic geometry and queueing theory. Simulation results demonstrate that the deployment of UAVs can promote MPT performance compared to ground SBSs and appropriate UAV height, UAV density, and spectrum allocation play significant roles in improving QoS performance of heterogeneous traffic in the network.

The dilemma between the ever-increasing demands for data processing, and the limited capabilities of mobile devices in a wireless communication system calls for the appearance of federated learning (FL). As a distributed machine learning (ML) method, FL executes in an iterative manner by distributing the global model parameters and aggregating the local model parameters, which avoids the transmission of huge raw data and preserves data privacy during the training process. However, since FL cannot control the local training and transmission process, this gives malicious users the opportunity to deteriorate the global aggregation. We adopt a reputation model based on beta distribution function to measure the credibility of local users, and propose a reputation-based scheduling policy with user fairness constraint. By taking into account the impact of wireless channel conditions and malicious attack features, we derive tractable expressions for the convergence rate of FL in a wireless setting. Moreover, we validate the superiority of the proposed reputation-based scheduling policy via numerical analysis and empirical simulations. The results show that the proposed secure wireless FL framework can not only distinguish malicious users from normal users but also effectively defend against several typical attack types featured in attack intensity and attack frequency. The analysis also reveals that the effect of average attack intensity on the convergence performance of FL is dominated by the percentage of malicious user equipments (UEs), and imposes even greater negative effect on the convergence performance of FL as the percentage of malicious UEs increases.

Timeliness is an emerging requirement for many Internet of Things (IoT) applications. In IoT networks with a large number of nodes, severe interference may incur that leads to Age-of-Information (AoI) degradation. It is, therefore, important to study how to optimize the AoI performance. This article focuses on the AoI minimization in random-access Poisson networks. By considering the spatiotemporal interactions amongst the transmitters, an expression of the peak AoI is derived, based on which the optimal peak AoI and the corresponding optimal packet arrival rate and channel access probability are further characterized. The analysis shows that when the channel access probability (resp., the packet arrival rate) is given, the optimal packet arrival rate (resp., the optimal channel access probability) is equal to one when nodes are sparsely deployed, and decreases as the node deployment density increases. With a joint tuning of these two system parameters, the optimal channel access probability always equals one. Moreover, with the sole tuning of the channel access probability, the optimal peak AoI is improved with a smaller packet arrival rate only when the node deployment density is high. In contrast, a higher channel access probability always improves peak AoI performance when the packet arrival rate is solely tuned. The analysis in this article sheds important light on freshness-aware design for large-scale networks.

Age of Information (AoI), a notion that measures the information freshness, is an essential performance measure for time-critical applications in Internet of Things (IoT). With the surge of computing resources at the IoT devices, it is possible to preprocess the information packets that contain the status update before sending them to the destination so as to alleviate the transmission burden. However, the additional time and energy expenditure induced by computing also make the optimal updating a nontrivial problem. In this article, we consider a time-critical IoT system, where the IoT device is capable of preprocessing the status update before the transmission. Particularly, we aim to jointly design the preprocessing and transmission so that the weighted sum of the average AoI of the destination and the energy consumption of the IoT device is minimized. Due to the heterogeneity in transmission and computation capacities, the durations of distinct actions of the IoT device are nonuniform. Therefore, we formulate the status updating problem as an infinite horizon average cost semi-Markov decision process (SMDP) and then transform it into a discrete-time Markov decision process. We demonstrate that the optimal policy is of threshold type with respect to the AoI. Equipped with this, a structure-aware relative policy iteration algorithm is proposed to obtain the optimal policy of the SMDP. Our analysis shows that preprocessing is more beneficial in regimes of high AoIs, given it can reduce the time required for updates. We further prove the switching structure of the optimal policy in a special scenario, where the status updates are transmitted over a reliable channel and derive the optimal threshold. Finally, simulation results demonstrate the efficacy of preprocessing and show that the proposed policy outperforms two baseline policies.

Implementing federated learning (FL) algorithms in wireless networks has garnered a wide range of attention. However, few works have considered the impact of user mobility on the learning performance. To fill this research gap, we develop a theoretical model to characterize the hierarchical federated learning (HFL) algorithm in wireless networks where the mobile users may roam across edge access points (APs), leading to incompletion of inconsistent FL training. We provide the convergence analysis of conventional HFL with user mobility. Our analysis proves that the learning performance of conventional HFL deteriorates drastically with highly-mobile users. And such a decline in the learning performance will be exacerbated with small number of participants and large data distribution divergences among users’ local data. To circumvent these issues, we propose a mobility-aware cluster federated learning (MACFL) algorithm by redesigning the access mechanism, local update rule, and model aggregation scheme. We also conduct experiments to evaluate the learning performance of conventional HFL, a cluster federated learning (CFL) with simple averaging, and our proposed MACFL. The results show that our MACFL can enhance the learning performance, especially for three different cases: ($i$) the case of users with non-independent and identically distributed (non-IID) data, ($ii$) the case of users with high mobility, and ($iii$) the case with a small number of users.

In dense small cell networks, dynamic time-division duplex (D-TDD) technology has emerged as a promising solution to accommodate the fast variants of volatile traffic conditions because it allows each cell to dynamically configure the uplink and downlink transmission directions. However, the flexibility of traffic configuration introduces additional inter-cell interference, which largely deteriorates network throughput. This paper proposes an interference coordination technology for D-TDD small cell networks by integrating fractional frequency reuse (FFR) with cell clustering. To evaluate the system performance, we develop a theoretical framework to analytically characterize the mean packet throughput (MPT) performance by considering the impact of spatio-temporal traffic. The analytical model can be extended to further study the FFR-based D-TDD, clustered D-TDD, and traditional D-TDD networks. We verify the accuracy of our analysis through simulations and whereby explore the effect of different network parameters. Numerical results demonstrate that the proposed scheme outperforms clustered D-TDD and traditional D-TDD for both the downlink and uplink spatially averaged MPT, and can significantly improve the performance in uplink while slightly decreasing that in downlink compared with FFR-based D-TDD. Furthermore, by jointly optimizing network parameters, the spatially averaged MPT can be maximized while enduring MPT per user.

The concept of hierarchical federated edge learning (H-FEEL) has been recently proposed as an enhancement of federated learning model. Such a system generally consists of three entities, i.e., the server, helpers, and clients, in which each helper collects the trained gradients from clients nearby, aggregates them, and sends the result to the server for global model update. Due to limited communication resources, only a portion of helpers can be scheduled to upload their aggregated gradients in each round of the model training. And that necessitates a well-designed scheme for the joint helper scheduling and communication resources allocation. In this paper, we develop a training algorithm for the H-FEEL system which involves local gradient computing, weighted gradient uploading, and machine learning model updating phases. By characterizing these phases mathematically and analyzing one-round convergence bound of the training algorithm, we formulate an optimization problem to achieve the scheduling and resource allocation scheme. The problem simultaneously captures the uncertainty of the wireless channel and the importance of the weighted gradient. To solve the problem, we first transform it into an equivalent problem and then decompose the transformed problem into two subproblems:bit and sub-channel allocationandhelper scheduling, which are mixed integer nonlinear programming and continuous nonlinear problems, respectively. For the first subproblem, we obtain an optimal solution of exponential complexity and a suboptimal solution that has polynomial complexity. For the second subproblem, we obtain a closed-form optimal solution in a special case and a suboptimal solution in the general case. The efficacy of our scheme is amply demonstrated via simulations and the analytical framework is shown to provide valuable design insights for the practical implementation of the H-FEEL system.

Terahertz (THz) communication is capable of providing ultra-wide bandwidth and high data rates. Therefore attracts widespread attention to its applications in next-generation networks. Highly directional antennas are used to compensate for the THz propagation loss, which also incurs beam management challenges. Specifically, caused by node mobility and blockage, frequent beam reselections and beam misalignment greatly degrade THz network performance in terms of reliability and spatial throughput. In this paper, using stochastic geometry, we fill the current research gap in system-level theoretical models for the analysis of beam misalignment and network spatial throughput by considering the effects of beamwidth, mobility, blockage, and molecular absorption. Our analyses show that an increase in nodes density or user mobility often results in severe beam misalignment, which in turn requires more signaling overhead and degrades THz network reliability and throughput. Although using wider beams reduces this impact, it increases THz network sensitivity to molecular absorption. To maximize spatial throughput, optimal beamwidth needs to be adjusted according to communication demand priority and network status. Our work provides useful insights into beamwidth adaptation according to parameters trade-off that helps THz network achieve higher reliability and throughput in different applications.

Recently, the implementation of federated learning (FL) in wireless networks becomes a hotspot due to its flexible collaborative learning methods and privacy-preserving benefits. However, most of the existing works overlook the impact of user mobility on the learning performance, which is critical. Specifically, the mobile users may roam among multiple edge access points (APs) during the local training procedures, leading to incompletion of inconsistent FL training. In this paper, we theoretically study the impact of user mobility on the FL in hierarchical wireless networks. In our system model, the network consists of one cloud server, several edge APs, and multiple mobile users that have their positions vary over time. During the local training process, users may stay in or move out of the coverage area of the originally attached edge AP. In such a practical context, we analyze the convergence rate of the FL algorithm and provide experiments to evaluate the learning performance under different network parameters. Our results provide insights in further improvements of FL in hierarchical wireless networks.

In this article, we develop a framework for the analysis of massive multiple-input-multiple-output (MIMO) systems where multiple types of devices with different configurations and requirements co-exist, by taking into account the randomness of spatial locations and temporal traffic. A tight closed-form approximation of the spatial mean packet throughput, which denotes the average number of packets that are successfully transmitted at any unit time slot and area is derived, by using tools from the stochastic geometry and queuing theory, which captures all the key features of the devices in the Internet of Things (IoT). Based on the analysis, we investigate the optimal scheduling number for each type of devices that maximizes the spatial mean packet throughput while meeting devices' delay constraints. It is found that when the base station (BS) has excessive number of antennas ( M), the BS should schedule all devices under its coverage, regardless of devices' variances on spatiotemporal configurations and demands. However, when M is limited, the BS should have a bias on scheduling devices with heavier traffic, lower decoding threshold, or higher transmit power. On this basis, if the delay constraint of one device becomes stricter, it will be scheduled more often to access the radio channel, which acts more significantly when the ratio of M to the deployment density of devices gets smaller.

We study the performance of wireless links for a class of Poisson networks, in which packets arrive at the transmitters following Bernoulli processes. By combining stochastic geometry with queueing theory, two fundamental measures are analyzed, namely the transmission success probability and the meta distribution of signal-to-interference-plus-noise ratio (SINR). Different from the conventional approaches that assume independent active states across the nodes and use homogeneous point processes to model the locations of interferers, our analysis accounts for the interdependency amongst active states of the transmitters in space and arrives at a non-homogeneous point process for the modeling of interferers' positions, which leads to a more accurate characterization of the SINR. The accuracy of the theoretical results is verified by simulations, and the developed framework is then used to devise design guidelines for the deployment strategies of wireless networks.

This paper considers a scenario in which an access point (AP) is equipped with a server of finite computing power, and serves multiple resource-hungry users by charging users a price. This price helps to regulate users' behavior in offloading jobs to the AP. However, existing works on pricing are based on abstract concave utility functions, giving no dependence on physical layer parameters. To that end, we first introduce a novel utility function, which measures the cost reduction by offloading as compared with executing jobs locally. Based on this utility function we then formulate two offloading games, with one maximizing individuals interest and the other maximizing the overall systems interest. We analyze the structural property of the games and admit in closed-form the Nash Equilibrium and the Social Equilibrium for the homogeneous user case, respectively. The proposed expressions are functions of user parameters such as the weights of time and energy, the distance from the AP, thus constituting an advancement over prior economic works that have considered only abstract functions. Finally, we propose an optimal price-based scheme, with which we prove that the interactive decision-making process with self-interested users converges to a Nash Equilibrium point equal to the Social Equilibrium point.

Optimization of information freshness in wireless networks has usually been performed based on queueing analysis that captures only the temporal traffic dynamics associated with the transmitters and receivers. However, the effect of interference, which is mainly dominated by the interferers' geographic locations, is not well understood. In this paper, we leverage a spatiotemporal model, which allows one to characterize the age of information (AoI) from a joint queueing-geometry perspective, for the design of a decentralized scheduling policy that exploits local observation to make transmission decisions that minimize the AoI. To quantify the performance, we also derive accurate and tractable expressions for the peak AoI. Numerical results reveal that: i) the packet arrival rate directly affects the service process due to queueing interactions, ii) the proposed scheme can adapt to traffic variations and largely reduce the peak AoI, and iii) the proposed scheme scales well as the network grows in size. This is done by adaptively adjusting the radio access probability at each transmitter to the change of the ambient environment.

Caching popular contents at the network edge has been considered as a promising enabler to relieve the pressure on networks due to the fact that a substantial portion of global data traffic is repeatedly requested by many subscribers and thus redundantly generated. Recommendation, on the other hand, has attracted spiraling attention for its capability of reshaping users’ contents demand patterns. In this paper, we examine the practicability of recommendation in boosting the gains of edge caching with uncharted users’ feature information. To this end, we first characterize the average system cost for a generic network model, disclosing its dependence on the recommendation and caching strategies. Then, we formulate the joint caching and recommendation decision oriented cost minimization problem, taking the constraints on each content provider’s cache capacity budget, each individual user’s recommendation size and recommendation quality into account. However, the implicit information regarding users’ preference makes the problem inextricable. To address this issue, a versatile long short term memory (LSTM) network assisted prediction paradigm is proposed to attain the preference schema of users with the assistance of their historical behavior data. Based on that, we rigorously prove the NP-hardness of obtaining the optimal recommendation and caching policies that jointly minimize the system cost. Therewith, an iterative suboptimal algorithm is developed, which has provable polynomial time complexity and convergence guarantee. Extensive simulation results validate the effectiveness of our proposed LSTM enabled feature information prediction approach and the convergence performance of the devised joint decision making methodology. In addition, it is shown that the proposed scheme outperforms numerous benchmarks significantly.

In the Internet of Things (IoT) networks, caching is a promising technique to alleviate energy consumption of sensors by responding to users’ data requests with the data packets cached in the edge caching node (ECN). However, without an efficient status update strategy, the information obtained by users may be stale, which in return would inevitably deteriorate the accuracy and reliability of derived decisions for real-time applications. In this paper, we focus on striking the balance between the information freshness, in terms of age of information (AoI), experienced by users and energy consumed by sensors, by appropriately activating sensors to update their current status. Particularly, we first depict the evolutions of the AoI with each sensor from different users’ perspective with time steps of non-uniform duration, which are determined by both the users’ data requests and the ECN’s status update decision. Then, we formulate a non-uniform time step based dynamic status update optimization problem to minimize the long-term average cost, jointly considering the average AoI and energy consumption. To this end, a Markov Decision Process is formulated and further, a dueling deep R-network based dynamic status update algorithm is devised by combining dueling deep Q-network and tabular R-learning, with which challenges from the curse of dimensionality and unknown of the environmental dynamics can be addressed. Finally, extensive simulations are conducted to validate the effectiveness of our proposed algorithm by comparing it with five baseline deep reinforcement learning algorithms and policies.

The notion of age-of-information (AoI) is investigated in the context of large-scale wireless networks, in which transmitters need to send a sequence of information packets, which are generated as independent Bernoulli processes, to their intended receivers over a shared spectrum. Due to interference, the rate of packet depletion at any given node is entangled with both the spatial configurations, which determine the path loss, and temporal dynamics, which influence the active states, of the other transmitters, resulting in the queues to interact with each other in both space and time over the entire network. To that end, variants in the packet update frequency affect not just the inter-arrival time but also the departure process, and the impact of such phenomena on the AoI is not well understood. In this paper, we establish a theoretical framework to characterize the AoI performance in the aforementioned setting. Particularly, tractable expressions are derived for both the peak and average AoI under two different transmission protocols, namely the first-come-first-serve (FCFS) and the last-come-first-serve with preemption (LCFS-PR). Additionally, our analysis also accounts for the effects of channel access controls such as ALOHA on the AoI. The accuracy of the analysis is verified via simulations, and based on the theoretical outcomes, we find that: i) networks operating under LCFS-PR are able to attain smaller values of peak and average AoI than that under FCFS, whereas the gain is more pronounced when the infrastructure is densely deployed, ii) in sparsely deployed networks, ALOHA with a universally designed channel access probability is not instrumental in reducing the AoI, thus calling for more advanced channel access approaches, and iii) when the infrastructure is densely rolled out, there exists a non-trivial ALOHA channel access probability that minimizes the peak and average AoI under both FCFS and LCFS-PR.

Small cell networks with dynamic time-division duplex (D-TDD) have emerged as a potential solution to address the asymmetric traffic demands in 5G wireless networks. By allowing the dynamic adjustment of cell-specific UL/DL configuration, D-TDD flexibly allocates percentage of subframes to UL and DL transmissions to accommodate the traffic within each cell. However, the unaligned transmissions bring in extra interference which degrades the potential gain achieved by D-TDD. In this work, we propose an analytical framework to study the performance of multi-antenna small cell networks with clustered D-TDD, where cell clustering is employed to mitigate the interference from opposite transmission direction in neighboring cells. With tools from stochastic geometry, we derive explicit expressions and tractable tight upper bounds for success probability and network throughput. The proposed analytical framework allows to quantify the effect of key system parameters, such as UL/DL configuration, cluster size, antenna number, and SINR threshold. Our results show the superiority of the clustered D-TDD over the traditional D-TDD, and reveal the fact that there exists an optimal cluster size for DL performance, while UL performance always benefits from a larger cluster.

We investigate the age-of-information (AoI) in the context of random access networks, in which transmitters need to send a sequence of information packets to intended receivers over shared spectrum. We establish an analytical framework that accounts for the key features of a wireless system, including the fading, path loss, network topology, as well as the spatial interactions amongst the queues. A closed-form expression is derived to quantity the network average AoI and its accuracy is verified via simulations. Our analysis unveils several unconventional behaviors of AoI in such a setting. For instance, even when the packet transmissions are scheduled in a last-come first-serve (LCFS) order whereby the newly incoming packets can replace the undelivered ones, the network average AoI may not monotonically decline with respect to the packet arrival rates, if the infrastructure is densely deployed. Moreover, the ALOHA protocol is shown to be instrumental in reducing the AoI when the packet arrival rates are high, yet it cannot contribute to decreasing the AoI in the regime of infrequent packet arrivals.

Federated learning (FL) is a machine learning model that preserves data privacy in the training process. Specifically, FL brings the model directly to the user equipments (UEs) for local training, where an edge server periodically collects the trained parameters to produce an improved model and sends it back to the UEs. However, since communication usually occurs through a limited spectrum, only a portion of the UEs can update their parameters upon each global aggregation. As such, new scheduling algorithms have to be engineered to facilitate the full implementation of FL. In this paper, based on a metric termed the age of update (AoU), we propose a scheduling policy by jointly accounting for the staleness of the received parameters and the instantaneous channel qualities to improve the running efficiency of FL. The proposed algorithm has low complexity and its effectiveness is demonstrated by Monte Carlo simulations.

Internet of Things (IoT) has emerged as one of the key features of the next-generation wireless networks, where timely delivery of status update packets is essential for many real-time IoT applications. To provide users with context-aware services and lighten the transmission burden, the raw data usually need to be preprocessed before being transmitted to the destination. However, the effect of computing on the overall information freshness is not well understood. In this article, we first develop an analytical framework to investigate the information freshness, in terms of peak age of information (PAoI), of a computing-enabled IoT system with multiple sensors. Specifically, we model the procedure of computing and transmission as a tandem queue and derive the analytical expressions of the average PAoI for different sensors. Based on the theoretical results, we formulate a min-max optimization problem to minimize the maximum average PAoI of different sensors. We further design a derivative-free algorithm to find the optimal updating frequency, with which the complexity for checking the convexity of the formulated problem or obtaining the derivatives of the object function can be largely reduced. The accuracy of our analysis and the effectiveness of the proposed algorithm are verified with extensive simulation results.

Motivated by the increasing computational capacity of wireless user equipments (UEs), e.g., smart phones, tablets, or vehicles, as well as the increasing concerns about sharing private data, a new machine learning model has emerged, namely federated learning (FL), that allows a decoupling of data acquisition and computation at the central unit. Unlike centralized learning taking place in a data center, FL usually operates in a wireless edge network where the communication medium is resource-constrained and unreliable. Due to limited bandwidth, only a portion of UEs can be scheduled for updates at each iteration. Due to the shared nature of the wireless medium, transmissions are subjected to interference and are not guaranteed. The performance of FL system in such a setting is not well understood. In this paper, an analytical model is developed to characterize the performance of FL in wireless networks. Particularly, tractable expressions are derived for the convergence rate of FL in a wireless setting, accounting for effects from both scheduling schemes and inter-cell interference. Using the developed analysis, the effectiveness of three different scheduling policies, i.e., random scheduling (RS), round robin (RR), and proportional fair (PF), are compared in terms of FL convergence rate. It is shown that running FL with PF outperforms RS and RR if the network is operating under a high signal-to-interference-plus-noise ratio (SINR) threshold, while RR is more preferable when the SINR threshold is low. Moreover, the FL convergence rate decreases rapidly as the SINR threshold increases, thus confirming the importance of compression and quantization of the update parameters. The analysis also reveals a trade-off between the number of scheduled UEs and subchannel bandwidth under a fixed amount of available spectrum.

Federated learning (FL), as a type of distributed machine learning, is capable of significantly preserving clients’ private data from being exposed to adversaries. Nevertheless, private information can still be divulged by analyzing uploaded parameters from clients, e.g., weights trained in deep neural networks. In this paper, to effectively prevent information leakage, we propose a novel framework based on the concept of differential privacy (DP), in which artificial noise is added to parameters at the clients’ side before aggregating, namely, noising before model aggregation FL (NbAFL). First, we prove that the NbAFL can satisfy DP under distinct protection levels by properly adapting different variances of artificial noise. Then we develop a theoretical convergence bound on the loss function of the trained FL model in the NbAFL. Specifically, the theoretical bound reveals the following three key properties: 1) there is a tradeoff between convergence performance and privacy protection levels, i.e., better convergence performance leads to a lower protection level; 2) given a fixed privacy protection level, increasing the number$N$of overall clients participating in FL can improve the convergence performance; and 3) there is an optimal number aggregation times (communication rounds) in terms of convergence performance for a given protection level. Furthermore, we propose a$K$-client random scheduling strategy, where$K$($1\leq K< N$) clients are randomly selected from the$N$overall clients to participate in each aggregation. We also develop a corresponding convergence bound for the loss function in this case and the$K$-client random scheduling strategy also retains the above three properties. Moreover, we find that there is an optimal$K$that achieves the best convergence performance at a fixed privacy level. Evaluations demonstrate that our theoretical results are consistent with simulations, thereby facilitating the design of various privacy-preserving FL algorithms with different tradeoff requirements on convergence performance and privacy levels.

By exploiting the computing power and local data of distributed clients, federated learning (FL) features ubiquitous properties such as reduction of communication overhead and preserving data privacy. In each communication round of FL, the clients update local models based on their own data and upload their local updates via wireless channels. However, latency caused by hundreds to thousands of communication rounds remains a bottleneck in FL. To minimize the training latency, this work provides a multi-armed bandit-based framework for online client scheduling (CS) in FL without knowing wireless channel state information and statistical characteristics of clients. Firstly, we propose a CS algorithm based on the upper confidence bound policy (CS-UCB) for ideal scenarios where local datasets of clients are independent and identically distributed (i.i.d.) and balanced. An upper bound of the expected performance regret of the proposed CS-UCB algorithm is provided, which indicates that the regret grows logarithmically over communication rounds. Then, to address non-ideal scenarios with non-i.i.d. and unbalanced properties of local datasets and varying availability of clients, we further propose a CS algorithm based on the UCB policy and virtual queue technique (CS-UCB-Q). An upper bound is also derived, which shows that the expected performance regret of the proposed CS-UCB-Q algorithm can have a sub-linear growth over communication rounds under certain conditions. Besides, the convergence performance of FL training is also analyzed. Finally, simulation results validate the efficiency of the proposed algorithms.

Optimization of information freshness in wireless networks has usually been performed based on queueing analysis that captures only the temporal traffic dynamics associated with the transmitters and receivers. However, the effect of interference, which is mainly dominated by the interferersa geographic locations, is not well understood. In this paper, we leverage a spatiotemporal model, which allows one to characterize the age of information (AoI) from a joint queueing-geometry perspective, and design a decentralized scheduling policy that exploits local observation to make transmission decisions that minimize the AoI. Simulation results reveal that the proposed scheme not only largely reduces the peak AoI but also scales well with the network size.

In this paper, we develop a mathematical framework for the analysis of SINR in small cell networks with temporal traffic variation. Our model captures the interplay between the spatial location of small access points, which determines the magnitude of mutual interference, and their temporal traffic dynamic. We derive a tractable expression for the SINR meta distribution, and verify its accuracy via simulations. Based on our analysis, we find that i) under the same configuration, when traffic condition changes from light to heavy, the corresponding SINR requirement can differ by more than 10 dB for the network to maintain coverage, ii) the SINR coverage probability varies largely with traffic fluctuation in the sub-medium load regime, and iii) spatial densification can boost up the cell-edge rate almost linearly in the dense regime, thus demonstrates the effectiveness from cell densification.

Internet of Things (IoT) has been emerging as one of the use cases permeating our daily lives in 5th Generation wireless networks, where status update packages are usually required to be timely delivered for many IoT based intelligent applications. Enabling the collected raw data to be preprocessed before transmitted to the destination can provide users with better context-aware services and lighten the transmission burden. However, the effect from data preprocessing on the overall information freshness is an essential yet unrevealed issue. In this work we study the joint effect of data preprocessing and transmission procedures on information freshness measured by peak age of information (PAoI). Particularity, we formulate the considered multi-source preprocessing and transmission enabled IoT system as a tandem queue where a priority M/G/1 queue is followed by a G/G/1 queue. Then, we respectively derive the closed-form and an information theoretic approximation of the expectations of waiting time for the formulated processing queue and transmission queue, and further get the analytical expression of the average PAoI for packages from different sources. Finally, the accuracy of our analysis is verified with simulation results.

In Internet of Things, wireless access networks are required to support a large number of user equipments (UEs) in real time. With UEs' frequently arrival and departure, plenty of packet collisions can occur. Successive decoding thus becomes a promising technique to support the massive connectivity as it is capable of recovering packets from mixed received signals. In this paper, we propose an adaptive framing with successive decoding (AFSD) frame structure to deal with the fluidity of UEs. In the AFSD structure, the frame length is adaptively adjusted based on the present UE number, and successive decoding is adopted to alleviate the packet loss caused by packet collision. To quantify its performance, we provide exact as well as asymptotic results for the average delay. The analytical results are validated by simulations, and further extended to practical systems with estimated UE number, which shows that knowledge of active UE number is useful for delay reduction. Optimizations on UE transmission probability and frame length are also presented. In particular, the numerical studies reveal that by using optimized parameters, the average delay is reduced significantly for a wide range of arrival rates, which validates the effectiveness of the AFSD structure.

While the growth of mobile applications brings enjoyment to daily life, it also imposes more complicated traffic situations on wireless networks. A complete understanding of the impact from the traffic profile is thus essential for network operators to deploy next generation small cell networks. In this paper, based on stochastic geometry and queuing theory, we develop a mathematical framework that captures the interplay between the spatial location of small access points, which determines the magnitude of mutual interference, and their temporal traffic dynamic. We derive a tractable expression for the SINR distribution, and verify its accuracy via simulations. Based on our analysis, we find that: 1) under the same configuration, when traffic condition changes from light to heavy, the corresponding SINR requirement can differ by more than 10 dB for the network to maintain coverage; 2) the SINR coverage probability varies largely with traffic fluctuation in the sub-medium load regime, whereas in a scenario with a very light traffic load, the SINR outage probability increases linearly with the packet arrival rate; and 3) spatial densification can boost the cell-edge rate almost linearly in the dense regime, and thus confirms the appeal for ultra-dense deployment of small cells in the next generation wireless network.

We propose a downlink cell-edge-aware zero forcing (CEA-ZF) and block diagonalization (BD) cooperative precoding scheme to reduce the downlink interference caused by the small cell range expansion in a heterogeneous network (HetNet). The CEA-ZF precoding algorithm adopted in the downlink transmission of macro base station (MBS) exploits the spatial degrees of freedom from large antenna array to suppress the inter-cell interference, and the BD precoding algorithm is introduced in the small access point (SAP) to eliminate the multi-user interference. Simulation results demonstrate the benefits of the proposed downlink precoding scheme over the alternative approach, and verify the proposed scheme as a more effective interference mitigation scheme for the downlink massive multiple-input multiple-output (MIMO) HetNet. Moreover, the optimal range expansion bias (REB) of small cell range expansion is obtained, and we also give a lower bound for the UE sum-rate of the proposed precoding scheme.

Dynamic time-division duplex (D-TDD) transmission in small cell networks have been emerged as one of the promising solutions to support the asymmetric traffic requirements in the next generation cellular mobile communication systems. On the other hand, the backhaul, which carries the traffic between base stations (BSs) and the core network, has great influence on providing reliable and timely connectivity. In this work, we study a two-tier D-TDD network taking the random locations of devices, packet arrival process, scheduling, interference, and backhaul into consideration to understand the influence from backhaul to uplink (UL) and downlink (DL) mean packet throughput per UE (MPT for short). We use an approximate method to derive the interference and successful transmission probability via stochastic geometry, and then achieve the mathematical derivation of DL and UL MPT with different kinds of backhaul using queueing theory tools. Based on the simulation results, we verify the accuracy of our analysis and explore the impact of the service parameters and backhaul.

Capacity of energy harvesting communications with deterministic energy arrival and finite battery size is investigated. An abstraction of the physical layer is considered, where binary sequences are transmitted through a binary symmetric channel, and a cost function is associated with the transmission of each symbol. Upper and lower bounds on the channel capacity are derived for the general case by studying the normalized exponent of the cardinality of the set of feasible input sequences. Several upper bounds on the exponent are proposed by studying supersets of the feasible set. Lower bounds are derived by applying the binary entropy-power inequality and by using specific signaling schemes based on a save-and-transmit strategy. Numerical results are presented for several values of the energy arrival rate and battery size, validating the usefulness of the capacity bounds established for the energy harvesting channels.

Small cell networks under dynamic time-division duplex (D-TDD) transmission have emerged as a promising solution to accommodate the varied uplink (UL) and downlink (DL) traffic in next generation cellular mobile communication networks. By allowing each cell to individually configure its communication direction, D-TDD allocates resources to accommodate whichever transmission direction needs it most. However, with unaligned transmissions, the interference increases and limits the performance of mean packet throughput (MPT). In this paper, we study the small cell networks under D-TDD with cell clustering being the interference mitigation technique (clustered D-TDD). By leveraging stochastic geometry and queuing theory, we develop an analytical framework that captures both spatial and temporal randomness. We study the MPT whose analytical expression is verified via simulation, and based on the analysis, we explore the impact from different network and service parameters. In particular, numerical results show that there is an optimal cluster size for DL MPT, while UL MPT always benefits from increasing cluster size. By grouping cells into clusters, the clustered D-TDD can provide the flexible service compared with static time-division duplex (S-TDD), and provide significant improvement over a traditional D-TDD in terms of UL MPT at a small cost of DL MPT.

In this paper, we develop a framework using tools from stochastic geometry and queuing theory to evaluate the flow-level performance of downlink cellular networks with spatio-temporal traffic. Under this framework, we first obtain the mean service rate and the non-empty probability of a scheduled user queue by solving a fixed-point equation, which captures the inherent correlation between the interference and the queue status. By leveraging these results, we then derive closed-form expressions for the mean packet throughput and its bounds, defined as the mean number of packets that can be delivered during a given time duration. Simulation results validate the accuracy of the presented analysis, which can provide useful insight on the design of cellular networks while incorporating the spatial and temporal fluctuations of traffic.

We develop a framework for downlink heteroge- neous cellular networks with line-of-sight (LoS) and non-line- of- sight (NLoS) transmissions. Using stochastic geometry, we derive a tight approximation of the achievable downlink rate that enables us to compare the performance between densifying small cells and expanding base station (BS) antenna arrays. Interestingly, we find that adding small cells into a sparse network improves the achievable rate much faster than expanding antenna arrays at the macro BS. However, when the small cell density exceeds a critical threshold, the spacial densification will lose its benefits and further impair the network capacity. To this end, we present the optimal small cell density that maximizes the rate as practical deployment guidance. In contrast, expanding macro BS antenna array can always increase the capacity until reaching an upper bound caused by pilot contamination, and this bound also surpasses the peak rate obtained from deployment of small cells. Therefore, small cells are preferred for low rate requirements due to the rapid rate gain, and the massive MIMO is preferred for higher rate requirements due to better achievable rate.

We investigate the capacity of energy harvesting binary symmetric channels with deterministic energy arrival process and finite battery size. Using an abstraction of the physical layer, binary symbols are transmitted. A cost function is associated with each transmitted symbol. Upper and lower bounds on the channel capacity are derived as functions of the normalized exponent of the cardinality of the set of feasible input sequences. Upper and lower bounds on the normalized exponent are established by studying supersets defined by relaxed constraints and employing a harvest-and-transmit signaling scheme, respectively. Numerical results validate that bounds on the exponent imply effective bounds on the channel capacity.

In Internet of Things, wireless access networks need to support a large number of user equipments (UEs) in real time, where UEs frequently arrive and leave and plenty of packet collisions occur. In this paper, we propose a protocol termed adaptive framing with successive decoding (AFSD), that not only supports vast connectivity but also involves little signalling overhead. By using AFSD, the frame length can be adaptively adjusted based on the present UE number to deal with the fluidity of UEs, and successive decoding is adopted to recover signal from the mixed received signal to alleviate the packet collision. We explicitly analyze the average delay of our protocol. Numerical results verify the accuracy of our analysis, as well as the effectiveness of the proposed protocol on handling massive access.

We develop a framework for downlink heterogeneous cellular networks with line-of-sight (LoS) and non-LoS transmissions. Using stochastic geometry, we derive tight approximation of average downlink rate that enables us to compare the performance between densifying small cells and expanding base station (BS) antenna arrays. Interestingly, we find that adding small cells into the network improves the downlink rate much faster than expanding antenna arrays at the macro BS. However, when the small cell density exceeds a critical threshold, the spatial densification will lose its benefits and further impair the network capacity. To this end, we provide the optimal small cell density that maximizes the rate via numerical results for practical deployment guidance. In contrast, expanding macro BS antenna array can always benefit the capacity until an upper bound caused by pilot contamination, and this bound also surpasses the peak rate obtained from the deployment of small cells. Furthermore, we find that allocating part of antennas to distributed small cell BSs works better than centralizing all antennas at the macro BS, and the optimal allocation proportion is also given numerically for practical configuration reference. In summary, this paper provides a further understanding on how to leverage small cells and massive MIMO in future heterogeneous cellular networks deployment.

In this paper, we develop a general framework to analyze the rate performance of a two-tier MIMO heterogeneous network (HetNet) with wireless backhaul under spatial multiplexing. We consider linear precoding and receive filtering in the presence of interference from uplink and downlink transmissions. We find that the sum rate per area of the HetNet is sensitive to the network load, i.e., the number of users served by each base station. We show that a two-tier HetNet with wireless backhaul can achieve higher sum rate per area than a one-tier cellular network. However, this requires the bandwidth division between radio access links and wireless backhaul to be optimally designed according to the load conditions.

Dense and heterogeneous networks (HetNets) are being deployed to provide better coverage and throughput, thus improving the quality of experience at mobile users. Besides the important implications for energy consumption, the trend towards densification calls for more and more wireless links to forward a massive backhaul traffic into the core network. It is critically important to take into account the presence of a wireless backhaul for the energy-efficient design of HetNets. In this paper, we provide a general framework to analyze the energy efficiency of a two-tier MIMO heterogeneous network with wireless backhaul under spatial multiplexing and dynamic time division duplex. We find that a two-tier HetNet with wireless backhaul can be significantly more energy efficient than a one-tier cellular network. However, this requires the backhaul bandwidth to be carefully allocated according to the network load conditions.

As future networks aim to meet the ever-increasing requirements of high-data rate applications, dense, and heterogeneous networks (HetNets) will be deployed to provide better coverage and throughput. Besides the important implications for energy consumption, the trend toward densification calls for more and more wireless links to forward a massive backhaul traffic into the core network. It is critically important to take into account the presence of a wireless backhaul for the energy-efficient design of HetNets. In this paper, we provide a general framework to analyze the energy efficiency of a two-tier MIMO heterogeneous network with wireless backhaul in the presence of both uplink and downlink transmissions. We find that under spatial multiplexing the energy efficiency of a HetNet is sensitive to the network load, and it should be taken into account when controlling the number of users served by each base station. We show that a two-tier HetNet with wireless backhaul can be significantly more energy efficient than a one-tier cellular network. However, this requires the bandwidth division between radio access links and wireless backhaul to be optimally designed according to the load conditions.

The concept of mobile user equipment (UE) relay (UER) has been introduced to support device-to-device (D2D) communications for enhancing communication reliability. However, as the UER needs to use its own power for other UE's data transmission, relaying information in D2D communication may be undesirable for the UER. To overcome this issue, motivated by the recent advances in energy harvesting (EH) techniques, we propose a D2D communication provided EH heterogeneous cellular network (D2D-EHHN), where UERs harvest energy from an access point (AP) and use the harvested energy for D2D communication. We develop a framework for the design and analysis of D2D-EHHN by introducing the EH region (EHR) and modeling the status of harvested energy using Markov chain. The UER distribution is derived, and a transmission mode selection scheme including the efficient UER selection method is proposed. The network outage probability is derived in close form to measure the performance of D2D-EHHN. Based on our analysis results, we explore the effects of network parameters on the outage probability and the optimal offloading bias in terms of the outage probability. Particularly, we show that having a high EH efficiency enhances the performance of D2D-EHHN, but can also degrade, especially for dense network.

MOTIVATION: Genomic imprinting plays an important role in both normal development and diseases. Abnormal imprinting is strongly associated with several human diseases including cancers. Most of the imprinted genes were discovered in the neighborhood of the known imprinted genes. This approach is difficult to extend to analyze the whole genome. We have decided to take a computational approach to systematically search the whole genome for the presence of mono-allelic expressed genes and imprinted genes in human genome. RESULTS: A computational method was developed to identify novel imprinted or mono-allelic genes. Individuals represented in human cDNA libraries were genotyped using Bayesian statistics, and differential expression of polymorphic alleles was identified. A significant reduction in the number of libraries that expressed both alleles, measured by Z-statistics, is a strong indicator for an imprinted or a mono-allelic gene. AVAILABILITY: The data sets are available at http://leelab.nci.nih.gov/leelab/jsp/IGDM/IGDM.html